==================================================Ascend ============================= test session starts ============================== platform linux -- Python 3.9.19, pytest-6.2.5, py-1.11.0, pluggy-1.5.0 rootdir: /home/jenkins/mindspore/testcases/testcases/tests/st/infer/ops/test_internal_ops, configfile: ../../../../../../../../sault/virtual_test/virtualenv_003/sault/config/pytest.ini plugins: mock-3.14.0, hydra-core-1.3.2, forked-1.6.0, anyio-4.9.0, xdist-1.32.0 collected 16 items test_matmul_split.py [WARNING] ME(168319:281472866918192,MainProcess):2026-01-29-17:37:46.873.232 [mindspore/context.py:1334] For 'context.set_context', the parameter 'device_target' will be deprecated and removed in a future version. Please use the api mindspore.set_device() instead. TotalTime = 2.88351, [21] [bootstrap]: 0.00061398 [type_inference]: 2.60385 [event_method]: 2.323e-05 [auto_monad]: 0.0153667 [graph_reusing]: 2.225e-05 [inline]: 3.88999e-06 [add_attr]: 0.0750013, [1] [add_attr_with_inline]: 0.0749795, [1] [Cycle 1]: 0.00010801, [2] [tag_attr]: 2.408e-05 [meta_addattr_fg_expand]: 5.12e-06 [parallel-infer-symbol]: 3.96001e-06 [pre_auto_parallel]: 4.646e-05 [insert-virtual-dataset]: 2.78e-06 [parallel-infer-symbol-second]: 9.00007e-07 [dataset_repeat_opt]: 2.71999e-06 [pipeline_split]: 2.40002e-06 [optimize]: 0.168166, [53] [py_interpret_to_execute]: 3.561e-05 [rewriter_before_opt_a]: 9.485e-05 [opt_a]: 0.132838, [2] [Cycle 1]: 0.118628, [45] [expand_dump_flag]: 3.28998e-06 [switch_simplify]: 3.445e-05 [loop_unroll]: 2.041e-05 [a_1]: 0.00720765 [with_stream_mark]: 3.635e-05 [recompute_prepare]: 2.369e-05 [updatestate_depend_eliminate]: 1.096e-05 [updatestate_assign_eliminate]: 1.21e-05 [updatestate_loads_eliminate]: 1.112e-05 [parameter_eliminate]: 2.19999e-06 [a_2]: 0.00023917 [accelerated_algorithm]: 4.772e-05 [shard]: 3.01001e-06 [meta_shard_fg_expand]: 1.877e-05 [shard_inline]: 5.865e-05 [merge_send_recv]: 4.416e-05 [auto_parallel]: 2.359e-05 [parallel]: 6.494e-05 [flash_sp]: 2.47e-05 [merge_comm]: 9.24998e-06 [allreduce_fusion]: 8.62998e-06 [matmul_add_comm_reduction]: 1.968e-05 [allreduce_slice_to_reducescatter]: 1.02e-06 [virtual_shard_identity]: 2.33e-05 [virtual_dataset]: 1.793e-05 [get_grad_eliminate_]: 1.643e-05 [virtual_output]: 1.755e-05 [merge_forward]: 8.84003e-06 [cell_reuse_recompute_pass]: 3.46999e-06 [offload_activation]: 2.012e-05 [cell_reuse_handle_not_recompute_node_pass]: 4.728e-05 [merge_recompute_call_nodes]: 1.42999e-06 [before_grad]: 2.876e-05 [set_forward_comm_id_for_comm_node_pass]: 9.34e-06 [meta_fg_expand]: 6.76e-06 [flash_sp_send_recv_attached]: 6.16e-06 [receive_attached]: 1.234e-05 [after_resolve]: 2.518e-05 [a_after_grad]: 2.74e-05 [renormalize]: 0.0943844 [add_forward_monad_depend]: 1.939e-05 [auto_monad_grad]: 0.00651263 [auto_monad_eliminator]: 0.00011504 [cse]: 0.00266787 [a_3]: 0.00016392 [Cycle 2]: 0.0141922, [45] [expand_dump_flag]: 4.42998e-06 [switch_simplify]: 2.082e-05 [loop_unroll]: 1.877e-05 [a_1]: 0.00614238 [with_stream_mark]: 4.649e-05 [recompute_prepare]: 2.787e-05 [updatestate_depend_eliminate]: 1.078e-05 [updatestate_assign_eliminate]: 8.31002e-06 [updatestate_loads_eliminate]: 1.17e-05 [parameter_eliminate]: 2.81e-06 [a_2]: 0.00022967 [accelerated_algorithm]: 2.355e-05 [shard]: 3.76999e-06 [meta_shard_fg_expand]: 7.01001e-06 [shard_inline]: 1.572e-05 [merge_send_recv]: 1.493e-05 [auto_parallel]: 1.498e-05 [parallel]: 1.175e-05 [flash_sp]: 5.03002e-06 [merge_comm]: 8.84003e-06 [allreduce_fusion]: 8.78001e-06 [matmul_add_comm_reduction]: 1.746e-05 [allreduce_slice_to_reducescatter]: 0.00652539 [virtual_shard_identity]: 7.092e-05 [virtual_dataset]: 1.781e-05 [get_grad_eliminate_]: 1.694e-05 [virtual_output]: 1.589e-05 [merge_forward]: 2.305e-05 [cell_reuse_recompute_pass]: 3.53e-06 [offload_activation]: 3.013e-05 [cell_reuse_handle_not_recompute_node_pass]: 4.481e-05 [merge_recompute_call_nodes]: 3.18e-06 [before_grad]: 2.666e-05 [set_forward_comm_id_for_comm_node_pass]: 1.453e-05 [meta_fg_expand]: 9.71e-06 [flash_sp_send_recv_attached]: 2.87002e-06 [receive_attached]: 2.81e-06 [after_resolve]: 2.405e-05 [a_after_grad]: 2.789e-05 [renormalize]: 8.9989e-08 [add_forward_monad_depend]: 6.91999e-06 [auto_monad_grad]: 3.26001e-06 [auto_monad_eliminator]: 6.961e-05 [cse]: 0.00013505 [a_3]: 0.00010637 [py_interpret_to_execute_after_opt_a]: 2.776e-05 [slice_cell_reuse_recomputed_activation]: 2.81e-06 [rewriter_after_opt_a]: 0.00469453 [convert_after_rewriter]: 4.342e-05 [order_py_execute_after_rewriter]: 1.23e-05 [mutable_eliminate]: 0.00654035 [opt_b]: 0.00654705, [1] [Cycle 1]: 0.00653221, [7] [b_1]: 0.00622874 [b_2]: 3.303e-05 [updatestate_depend_eliminate]: 2.6e-05 [updatestate_assign_eliminate]: 1.245e-05 [updatestate_loads_eliminate]: 1.547e-05 [renormalize]: 6.99976e-07 [cse]: 0.00012328 [optimize_parallel_all_gather_comm]: 3.898e-05 [overlap_param_gather]: 6.66999e-06 [cconv]: 4.409e-05 [loop_unroll]: 0.00998553 [opt_after_cconv]: 0.00036652, [1] [Cycle 1]: 0.00035311, [7] [c_1]: 0.00014929 [parameter_eliminate]: 7.08e-06 [updatestate_depend_eliminate]: 2.383e-05 [updatestate_assign_eliminate]: 1.022e-05 [updatestate_loads_eliminate]: 1.412e-05 [cse]: 9.95e-05 [renormalize]: 1.35999e-06 [remove_dup_value]: 8.422e-05 [tuple_transform]: 0.00019593, [1] [Cycle 1]: 0.00018826, [4] [d_1]: 0.00013927 [none_parameter_eliminate]: 2.62001e-06 [renormalize]: 2.89991e-07 [switch_simplify]: 2.036e-05 [partial_unused_args_eliminate]: 2.64001e-06 [add_recomputation]: 0.00561775 [cse_after_recomputation]: 0.00014479, [1] [Cycle 1]: 0.00012861, [1] [cse]: 0.00010487 [environ_conv]: 3.887e-05 [swap_dp_allreduce_reducescatter]: 1.553e-05 [bias_add_comm_swap]: 4.15e-06 [label_micro_interleaved_index]: 7.73001e-06 [label_fine_grained_interleaved_index]: 2.89999e-06 [merge_cast_opt]: 1.81e-06 [slice_recompute_activation]: 2.42001e-06 [micro_interleaved_order_control]: 2.34001e-06 [assign_add_opt]: 1.82999e-06 [ForceFp32Comm]: 1.17999e-06 [remove_cast_before_assign_add]: 1.12e-06 [full_micro_interleaved_order_control]: 2.18002e-06 [reorder_send_recv_between_fp_bp]: 3.02002e-06 [comm_op_add_attrs]: 1.24003e-06 [add_comm_op_reuse_tag]: 1.09e-06 [interleave_split_concat_branches]: 1.25001e-06 [interleave_parallel_branches]: 1.11997e-06 [overlap_opt_shard_in_pipeline]: 2.811e-05 [overlap_opt_shard_grad_in_pipeline]: 1.82999e-06 [control_data_broadcast_order]: 3.309e-05 [grouped_pairwise_exchange_alltoall]: 1.92001e-06 [offloading_packed_experts]: 8.59e-06 [overlap_recompute_and_grad_model_parallel]: 8.30999e-06 [overlap_grad_matmul_and_grad_allreduce]: 1.34e-06 [overlap_recompute_allgather_and_fa_grad]: 1.34e-06 [overlap_recompute_comm]: 2.76e-06 [overlap_grad_ring_attention]: 7.95998e-06 [overlap_grad_flash_sp]: 4.441e-05 [begin_end_overlap_inline]: 5.79981e-07 [split_matmul_comm_elemetwise]: 2.45002e-06 [split_layernorm_comm]: 1.71e-06 [handle_group_info]: 1.05001e-06 [symbol_engine_optimizer]: 0.00020511, [1] [Cycle 1]: 0.00019819, [6] [build]: 1.863e-05 [elim_shapecalc]: 4.346e-05 [elim_not_effective]: 4.071e-05 [opt_reshape]: 2.449e-05 [fold_const_symbol]: 2.766e-05 [renormalize]: 5.19998e-07 [detach_backward]: 2.92002e-06 [pipeline_parallel_scheduler]: 1.99e-06 [auto_monad_reorder]: 6.371e-05 [get_jit_bprop_graph]: 0.00552104 [rewriter_after_jit_bprop_graph]: 2.051e-05 [opt_after_jit_grad]: 0.00782332 [validate]: 0.00014344 Sums bootstrap : 0.000614s : 0.02% type_inference : 2.603851s : 93.19% event_method : 0.000023s : 0.00% auto_monad : 0.015367s : 0.55% graph_reusing : 0.000022s : 0.00% inline : 0.000004s : 0.00% add_attr.add_attr_with_inline.tag_attr : 0.000024s : 0.00% add_attr.add_attr_with_inline.meta_addattr_fg_expand : 0.000005s : 0.00% parallel-infer-symbol : 0.000004s : 0.00% pre_auto_parallel : 0.000046s : 0.00% insert-virtual-dataset : 0.000003s : 0.00% parallel-infer-symbol-second : 0.000001s : 0.00% dataset_repeat_opt : 0.000003s : 0.00% pipeline_split : 0.000002s : 0.00% optimize.py_interpret_to_execute : 0.000036s : 0.00% optimize.rewriter_before_opt_a : 0.000095s : 0.00% optimize.opt_a.expand_dump_flag : 0.000008s : 0.00% optimize.opt_a.switch_simplify : 0.000055s : 0.00% optimize.opt_a.loop_unroll : 0.000039s : 0.00% optimize.opt_a.a_1 : 0.013350s : 0.48% optimize.opt_a.with_stream_mark : 0.000083s : 0.00% optimize.opt_a.recompute_prepare : 0.000052s : 0.00% optimize.opt_a.updatestate_depend_eliminate : 0.000022s : 0.00% optimize.opt_a.updatestate_assign_eliminate : 0.000020s : 0.00% optimize.opt_a.updatestate_loads_eliminate : 0.000023s : 0.00% optimize.opt_a.parameter_eliminate : 0.000005s : 0.00% optimize.opt_a.a_2 : 0.000469s : 0.02% optimize.opt_a.accelerated_algorithm : 0.000071s : 0.00% optimize.opt_a.shard : 0.000007s : 0.00% optimize.opt_a.meta_shard_fg_expand : 0.000026s : 0.00% optimize.opt_a.shard_inline : 0.000074s : 0.00% optimize.opt_a.merge_send_recv : 0.000059s : 0.00% optimize.opt_a.auto_parallel : 0.000039s : 0.00% optimize.opt_a.parallel : 0.000077s : 0.00% optimize.opt_a.flash_sp : 0.000030s : 0.00% optimize.opt_a.merge_comm : 0.000018s : 0.00% optimize.opt_a.allreduce_fusion : 0.000017s : 0.00% optimize.opt_a.matmul_add_comm_reduction : 0.000037s : 0.00% optimize.opt_a.allreduce_slice_to_reducescatter : 0.006526s : 0.23% optimize.opt_a.virtual_shard_identity : 0.000094s : 0.00% optimize.opt_a.virtual_dataset : 0.000036s : 0.00% optimize.opt_a.get_grad_eliminate_ : 0.000033s : 0.00% optimize.opt_a.virtual_output : 0.000033s : 0.00% optimize.opt_a.merge_forward : 0.000032s : 0.00% optimize.opt_a.cell_reuse_recompute_pass : 0.000007s : 0.00% optimize.opt_a.offload_activation : 0.000050s : 0.00% optimize.opt_a.cell_reuse_handle_not_recompute_node_pass : 0.000092s : 0.00% optimize.opt_a.merge_recompute_call_nodes : 0.000005s : 0.00% optimize.opt_a.before_grad : 0.000055s : 0.00% optimize.opt_a.set_forward_comm_id_for_comm_node_pass : 0.000024s : 0.00% optimize.opt_a.meta_fg_expand : 0.000016s : 0.00% optimize.opt_a.flash_sp_send_recv_attached : 0.000009s : 0.00% optimize.opt_a.receive_attached : 0.000015s : 0.00% optimize.opt_a.after_resolve : 0.000049s : 0.00% optimize.opt_a.a_after_grad : 0.000055s : 0.00% optimize.opt_a.renormalize : 0.094385s : 3.38% optimize.opt_a.add_forward_monad_depend : 0.000026s : 0.00% optimize.opt_a.auto_monad_grad : 0.006516s : 0.23% optimize.opt_a.auto_monad_eliminator : 0.000185s : 0.01% optimize.opt_a.cse : 0.002803s : 0.10% optimize.opt_a.a_3 : 0.000270s : 0.01% optimize.py_interpret_to_execute_after_opt_a : 0.000028s : 0.00% optimize.slice_cell_reuse_recomputed_activation : 0.000003s : 0.00% optimize.rewriter_after_opt_a : 0.004695s : 0.17% optimize.convert_after_rewriter : 0.000043s : 0.00% optimize.order_py_execute_after_rewriter : 0.000012s : 0.00% optimize.mutable_eliminate : 0.006540s : 0.23% optimize.opt_b.b_1 : 0.006229s : 0.22% optimize.opt_b.b_2 : 0.000033s : 0.00% optimize.opt_b.updatestate_depend_eliminate : 0.000026s : 0.00% optimize.opt_b.updatestate_assign_eliminate : 0.000012s : 0.00% optimize.opt_b.updatestate_loads_eliminate : 0.000015s : 0.00% optimize.opt_b.renormalize : 0.000001s : 0.00% optimize.opt_b.cse : 0.000123s : 0.00% optimize.optimize_parallel_all_gather_comm : 0.000039s : 0.00% optimize.overlap_param_gather : 0.000007s : 0.00% optimize.cconv : 0.000044s : 0.00% optimize.loop_unroll : 0.009986s : 0.36% optimize.opt_after_cconv.c_1 : 0.000149s : 0.01% optimize.opt_after_cconv.parameter_eliminate : 0.000007s : 0.00% optimize.opt_after_cconv.updatestate_depend_eliminate : 0.000024s : 0.00% optimize.opt_after_cconv.updatestate_assign_eliminate : 0.000010s : 0.00% optimize.opt_after_cconv.updatestate_loads_eliminate : 0.000014s : 0.00% optimize.opt_after_cconv.cse : 0.000099s : 0.00% optimize.opt_after_cconv.renormalize : 0.000001s : 0.00% optimize.remove_dup_value : 0.000084s : 0.00% optimize.tuple_transform.d_1 : 0.000139s : 0.00% optimize.tuple_transform.none_parameter_eliminate : 0.000003s : 0.00% optimize.tuple_transform.renormalize : 0.000000s : 0.00% optimize.tuple_transform.switch_simplify : 0.000020s : 0.00% optimize.partial_unused_args_eliminate : 0.000003s : 0.00% optimize.add_recomputation : 0.005618s : 0.20% optimize.cse_after_recomputation.cse : 0.000105s : 0.00% optimize.environ_conv : 0.000039s : 0.00% optimize.swap_dp_allreduce_reducescatter : 0.000016s : 0.00% optimize.bias_add_comm_swap : 0.000004s : 0.00% optimize.label_micro_interleaved_index : 0.000008s : 0.00% optimize.label_fine_grained_interleaved_index : 0.000003s : 0.00% optimize.merge_cast_opt : 0.000002s : 0.00% optimize.slice_recompute_activation : 0.000002s : 0.00% optimize.micro_interleaved_order_control : 0.000002s : 0.00% optimize.assign_add_opt : 0.000002s : 0.00% optimize.ForceFp32Comm : 0.000001s : 0.00% optimize.remove_cast_before_assign_add : 0.000001s : 0.00% optimize.full_micro_interleaved_order_control : 0.000002s : 0.00% optimize.reorder_send_recv_between_fp_bp : 0.000003s : 0.00% optimize.comm_op_add_attrs : 0.000001s : 0.00% optimize.add_comm_op_reuse_tag : 0.000001s : 0.00% optimize.interleave_split_concat_branches : 0.000001s : 0.00% optimize.interleave_parallel_branches : 0.000001s : 0.00% optimize.overlap_opt_shard_in_pipeline : 0.000028s : 0.00% optimize.overlap_opt_shard_grad_in_pipeline : 0.000002s : 0.00% optimize.control_data_broadcast_order : 0.000033s : 0.00% optimize.grouped_pairwise_exchange_alltoall : 0.000002s : 0.00% optimize.offloading_packed_experts : 0.000009s : 0.00% optimize.overlap_recompute_and_grad_model_parallel : 0.000008s : 0.00% optimize.overlap_grad_matmul_and_grad_allreduce : 0.000001s : 0.00% optimize.overlap_recompute_allgather_and_fa_grad : 0.000001s : 0.00% optimize.overlap_recompute_comm : 0.000003s : 0.00% optimize.overlap_grad_ring_attention : 0.000008s : 0.00% optimize.overlap_grad_flash_sp : 0.000044s : 0.00% optimize.begin_end_overlap_inline : 0.000001s : 0.00% optimize.split_matmul_comm_elemetwise : 0.000002s : 0.00% optimize.split_layernorm_comm : 0.000002s : 0.00% optimize.handle_group_info : 0.000001s : 0.00% optimize.symbol_engine_optimizer.build : 0.000019s : 0.00% optimize.symbol_engine_optimizer.elim_shapecalc : 0.000043s : 0.00% optimize.symbol_engine_optimizer.elim_not_effective : 0.000041s : 0.00% optimize.symbol_engine_optimizer.opt_reshape : 0.000024s : 0.00% optimize.symbol_engine_optimizer.fold_const_symbol : 0.000028s : 0.00% optimize.symbol_engine_optimizer.renormalize : 0.000001s : 0.00% detach_backward : 0.000003s : 0.00% pipeline_parallel_scheduler : 0.000002s : 0.00% auto_monad_reorder : 0.000064s : 0.00% get_jit_bprop_graph : 0.005521s : 0.20% rewriter_after_jit_bprop_graph : 0.000021s : 0.00% opt_after_jit_grad : 0.007823s : 0.28% validate : 0.000143s : 0.01% Time group info: ------[substitution.] 0.000345 96 3.37% : 0.000012s : 2: substitution.depend_value_elim 1.38% : 0.000005s : 8: substitution.elim_not_effective 1.13% : 0.000004s : 8: substitution.fold_const_symbol 4.26% : 0.000015s : 13: substitution.graph_param_transform 38.82% : 0.000134s : 1: substitution.inline 2.81% : 0.000010s : 16: substitution.j_node_and_user_rematch 7.95% : 0.000027s : 2: substitution.less_batch_normalization 2.33% : 0.000008s : 6: substitution.load_eliminater 0.77% : 0.000003s : 2: substitution.opt_reshape 4.14% : 0.000014s : 16: substitution.remove_not_recompute_node 2.79% : 0.000010s : 4: substitution.replace_old_param 12.58% : 0.000043s : 4: substitution.reshape_eliminate 10.19% : 0.000035s : 6: substitution.updatestate_pure_node_eliminater 7.47% : 0.000026s : 8: substitution.updatestate_useless_node_eliminater ------[type_inference.] 2.603710 2 98.32% : 2.559872s : 1: type_inference.infer 1.68% : 0.043838s : 1: type_inference.specialize ------[replace.] 0.006461 1 100.00% : 0.006461s : 1: replace.inline ------[match.] 0.000133 1 100.00% : 0.000133s : 1: match.inline ------[predicate.] 0.000537 3002 0.89% : 0.000005s : 27: predicate.accumulaten_eliminater 1.71% : 0.000009s : 13: predicate.ad_related_special_op_eliminate 0.69% : 0.000004s : 26: predicate.addn_check_dump 0.76% : 0.000004s : 27: predicate.addn_zero_filter 0.71% : 0.000004s : 27: predicate.adjust_all_reduce_mul_add 2.14% : 0.000011s : 53: predicate.arithmetic_simplify 0.89% : 0.000005s : 27: predicate.cast_eliminate 0.73% : 0.000004s : 26: predicate.check_bprop_eliminate 0.64% : 0.000003s : 26: predicate.compare_switch_simplify 0.39% : 0.000002s : 13: predicate.const_output_eliminate 0.70% : 0.000004s : 26: predicate.depend_value_elim 0.75% : 0.000004s : 27: predicate.dict_get_item_const_eliminator 0.83% : 0.000004s : 27: predicate.dict_get_item_eliminator 0.75% : 0.000004s : 27: predicate.dict_set_item_eliminator 1.33% : 0.000007s : 26: predicate.dumpgradient_eliminate 0.37% : 0.000002s : 13: predicate.elim_not_effective 0.58% : 0.000003s : 13: predicate.elim_shapecalc_of_broadcastargs 1.06% : 0.000006s : 40: predicate.environ_add_const_eliminate 0.98% : 0.000005s : 40: predicate.environ_get_add_eliminate 0.95% : 0.000005s : 40: predicate.environ_get_depend_swap 1.70% : 0.000009s : 66: predicate.environ_get_eliminate 0.95% : 0.000005s : 40: predicate.environ_get_set_eliminate 0.69% : 0.000004s : 28: predicate.exchange_switch_depend_value 1.66% : 0.000009s : 28: predicate.float_depend_g_call 0.63% : 0.000003s : 26: predicate.float_environ_get_switch 0.93% : 0.000005s : 39: predicate.float_tuple_getitem_switch 0.20% : 0.000001s : 13: predicate.fold_const_symbol 0.74% : 0.000004s : 26: predicate.get_grad_eliminate 0.26% : 0.000001s : 13: predicate.graph_param_transform 0.63% : 0.000003s : 26: predicate.incorporate_call 0.56% : 0.000003s : 26: predicate.incorporate_call_switch 6.21% : 0.000033s : 133: predicate.inline 0.94% : 0.000005s : 26: predicate.inline_without_move 0.37% : 0.000002s : 26: predicate.j_node_and_user_rematch 1.18% : 0.000006s : 27: predicate.less_batch_normalization 1.63% : 0.000009s : 53: predicate.list_to_tuple_eliminator_ 2.18% : 0.000012s : 80: predicate.load_eliminater 2.26% : 0.000012s : 13: predicate.loop_unroll_after_grad 0.86% : 0.000005s : 31: predicate.loop_unroll_before_grad 2.15% : 0.000012s : 53: predicate.make_slice_get_slice_eliminator 0.65% : 0.000003s : 26: predicate.merge_addn 0.70% : 0.000004s : 26: predicate.micro_step_allgather_replace 0.72% : 0.000004s : 26: predicate.mini_step_allgather_replace 0.65% : 0.000004s : 27: predicate.minmaximum_grad 1.78% : 0.000010s : 13: predicate.mutable_eliminate 0.43% : 0.000002s : 13: predicate.opt_reshape 0.42% : 0.000002s : 13: predicate.parallel_virtual_node 2.39% : 0.000013s : 28: predicate.partial_defer_inline 1.09% : 0.000006s : 40: predicate.partial_eliminate 0.71% : 0.000004s : 27: predicate.print_const_string_wrapper 0.66% : 0.000004s : 26: predicate.reduce_all_const_elim 1.13% : 0.000006s : 27: predicate.reduce_eliminate 2.04% : 0.000011s : 80: predicate.redundant_stop_gradient_eliminater 0.54% : 0.000003s : 26: predicate.remove_not_recompute_node 1.33% : 0.000007s : 53: predicate.replace_applicator 0.53% : 0.000003s : 26: predicate.replace_old_param 0.38% : 0.000002s : 13: predicate.reset_defer_inline 1.10% : 0.000006s : 27: predicate.reshape_eliminate 0.72% : 0.000004s : 26: predicate.row_tensor_add_zeros_like 0.71% : 0.000004s : 13: predicate.row_tensor_eliminate 1.14% : 0.000006s : 26: predicate.same_eliminate 0.49% : 0.000003s : 26: predicate.set_cell_output_no_recompute 2.86% : 0.000015s : 26: predicate.shard_identity_eliminate 0.78% : 0.000004s : 26: predicate.special_op_eliminate 0.77% : 0.000004s : 26: predicate.specialize_transform 1.21% : 0.000007s : 26: predicate.split_environ_get_set_with_tuple_value 0.80% : 0.000004s : 26: predicate.stack_unstack_eliminate 0.41% : 0.000002s : 13: predicate.switch_call_monad_eliminater 0.94% : 0.000005s : 28: predicate.switch_defer_inline 1.56% : 0.000008s : 54: predicate.switch_layer_defer_inline 2.85% : 0.000015s : 98: predicate.switch_simplify 0.88% : 0.000005s : 27: predicate.tile_eliminate 0.90% : 0.000005s : 27: predicate.transpose_eliminate 1.56% : 0.000008s : 53: predicate.tuple_list_convert_item_index_to_positive 1.65% : 0.000009s : 53: predicate.tuple_list_get_item_const_eliminator 1.46% : 0.000008s : 53: predicate.tuple_list_get_item_depend_reorder 3.18% : 0.000017s : 79: predicate.tuple_list_get_item_eliminator 1.56% : 0.000008s : 53: predicate.tuple_list_get_set_item_eliminator 2.37% : 0.000013s : 79: predicate.tuple_list_set_item_eliminator 1.54% : 0.000008s : 53: predicate.tuple_to_list_eliminator_ 1.97% : 0.000011s : 80: predicate.updatestate_pure_node_eliminater 3.02% : 0.000016s : 106: predicate.updatestate_useless_node_eliminater 0.38% : 0.000002s : 13: predicate.value_based_eliminate 0.80% : 0.000004s : 26: predicate.virtual_dataset_eliminate 0.76% : 0.000004s : 26: predicate.virtual_output_eliminate 0.41% : 0.000002s : 13: predicate.virtual_view_grad_eliminate 0.47% : 0.000002s : 13: predicate.zero_like_fill_zero ------[func_graph_cloner_run.] 0.126282 58 77.52% : 0.097892s : 55: func_graph_cloner_run.FuncGraphClonerGraph 22.48% : 0.028390s : 3: func_graph_cloner_run.FuncGraphSpecializer ------[meta_graph.] 0.000000 0 ------[manager.] 0.000000 0 ------[pynative] 0.000000 0 ------[others.] 3.236475 192 0.00% : 0.000004s : 1: ForceFp32Comm 2.32% : 0.075009s : 1: add_attr 2.32% : 0.074985s : 1: add_attr_with_inline 0.00% : 0.000004s : 1: add_comm_op_reuse_tag 0.17% : 0.005662s : 1: add_recomputation 0.00% : 0.000005s : 1: assign_add_opt 0.68% : 0.021863s : 1: auto_monad 0.00% : 0.000070s : 1: auto_monad_reorder 0.00% : 0.000004s : 1: begin_end_overlap_inline 0.00% : 0.000007s : 1: bias_add_comm_swap 0.02% : 0.000651s : 1: bootstrap 0.00% : 0.000049s : 1: cconv 0.00% : 0.000005s : 1: comm_op_add_attrs 0.00% : 0.000037s : 1: control_data_broadcast_order 0.00% : 0.000051s : 1: convert_after_rewriter 0.00% : 0.000151s : 1: cse_after_recomputation 0.00% : 0.000006s : 1: dataset_repeat_opt 0.00% : 0.000006s : 1: detach_backward 0.00% : 0.000044s : 1: environ_conv 0.00% : 0.000034s : 1: event_method 0.00% : 0.000005s : 1: full_micro_interleaved_order_control 0.17% : 0.005588s : 1: get_jit_bprop_graph 0.00% : 0.000043s : 1: graph_reusing 0.00% : 0.000005s : 1: grouped_pairwise_exchange_alltoall 0.00% : 0.000004s : 1: handle_group_info 0.00% : 0.000008s : 1: inline 0.00% : 0.000007s : 1: insert-virtual-dataset 0.00% : 0.000004s : 1: interleave_parallel_branches 0.00% : 0.000004s : 1: interleave_split_concat_branches 0.00% : 0.000006s : 1: label_fine_grained_interleaved_index 0.00% : 0.000011s : 1: label_micro_interleaved_index 0.31% : 0.010014s : 1: loop_unroll 0.00% : 0.000005s : 1: merge_cast_opt 0.00% : 0.000005s : 1: micro_interleaved_order_control 0.20% : 0.006565s : 1: mutable_eliminate 0.00% : 0.000012s : 1: offloading_packed_experts 0.00% : 0.000069s : 1: opt.transform.loop_unroll_optimizer 0.00% : 0.000063s : 1: opt.transform.mutable_eliminate 0.45% : 0.014665s : 78: opt.transform.opt_a 0.00% : 0.000147s : 1: opt.transform.opt_after_cconv 0.00% : 0.000100s : 1: opt.transform.opt_after_jit_grad 0.01% : 0.000421s : 28: opt.transform.opt_b 0.00% : 0.000156s : 2: opt.transform.opt_trans_graph 0.00% : 0.000128s : 4: opt.transform.symbol_engine_opt 4.10% : 0.132841s : 1: opt_a 0.01% : 0.000372s : 1: opt_after_cconv 0.24% : 0.007850s : 1: opt_after_jit_grad 0.20% : 0.006553s : 1: opt_b 5.20% : 0.168174s : 1: optimize 0.00% : 0.000043s : 1: optimize_parallel_all_gather_comm 0.00% : 0.000017s : 1: order_py_execute_after_rewriter 0.00% : 0.000049s : 1: overlap_grad_flash_sp 0.00% : 0.000004s : 1: overlap_grad_matmul_and_grad_allreduce 0.00% : 0.000012s : 1: overlap_grad_ring_attention 0.00% : 0.000005s : 1: overlap_opt_shard_grad_in_pipeline 0.00% : 0.000032s : 1: overlap_opt_shard_in_pipeline 0.00% : 0.000010s : 1: overlap_param_gather 0.00% : 0.000004s : 1: overlap_recompute_allgather_and_fa_grad 0.00% : 0.000012s : 1: overlap_recompute_and_grad_model_parallel 0.00% : 0.000006s : 1: overlap_recompute_comm 0.00% : 0.000009s : 1: parallel-infer-symbol 0.00% : 0.000004s : 1: parallel-infer-symbol-second 0.00% : 0.000007s : 1: partial_unused_args_eliminate 0.00% : 0.000005s : 1: pipeline_parallel_scheduler 0.00% : 0.000005s : 1: pipeline_split 0.00% : 0.000051s : 1: pre_auto_parallel 0.00% : 0.000040s : 1: py_interpret_to_execute 0.00% : 0.000033s : 1: py_interpret_to_execute_after_opt_a 0.00% : 0.000004s : 1: remove_cast_before_assign_add 0.00% : 0.000091s : 1: remove_dup_value 2.03% : 0.065619s : 1: renormalize.infer 0.89% : 0.028746s : 1: renormalize.specialize 0.00% : 0.000006s : 1: reorder_send_recv_between_fp_bp 0.00% : 0.000030s : 1: rewriter_after_jit_bprop_graph 0.15% : 0.004719s : 1: rewriter_after_opt_a 0.00% : 0.000100s : 1: rewriter_before_opt_a 0.00% : 0.000006s : 1: slice_cell_reuse_recomputed_activation 0.00% : 0.000005s : 1: slice_recompute_activation 0.00% : 0.000005s : 1: split_layernorm_comm 0.00% : 0.000005s : 1: split_matmul_comm_elemetwise 0.00% : 0.000019s : 1: swap_dp_allreduce_reducescatter 0.01% : 0.000209s : 1: symbol_engine_optimizer 0.01% : 0.000199s : 1: tuple_transform 80.45% : 2.603893s : 1: type_inference . [hook] pytest_runtest_teardown:test_matmul_qkv_bf16_1[False-mstype0-k_n_shape0-256] tests/st/infer/ops/test_internal_ops/test_matmul_split.py::test_matmul_qkv_bf16_1[False-mstype0-k_n_shape0-256],max_mem:124.0M [WARNING] ME(168319:281472866918192,MainProcess):2026-01-29-17:38:37.667.105 [mindspore/context.py:1334] For 'context.set_context', the parameter 'device_target' will be deprecated and removed in a future version. Please use the api mindspore.set_device() instead. TotalTime = 4.58818, [21] [bootstrap]: 0.00065945 [type_inference]: 4.0849 [event_method]: 2.234e-05 [auto_monad]: 0.00038223 [graph_reusing]: 6.83e-06 [inline]: 3.09999e-06 [add_attr]: 0.00844257, [1] [add_attr_with_inline]: 0.00842365, [1] [Cycle 1]: 0.00011356, [2] [tag_attr]: 5.293e-05 [meta_addattr_fg_expand]: 5.27999e-06 [parallel-infer-symbol]: 5.80002e-06 [pre_auto_parallel]: 3.963e-05 [insert-virtual-dataset]: 3.97e-06 [parallel-infer-symbol-second]: 8.00006e-07 [dataset_repeat_opt]: 2.29001e-06 [pipeline_split]: 1.57001e-06 [optimize]: 0.492586, [53] [py_interpret_to_execute]: 4.117e-05 [rewriter_before_opt_a]: 9.657e-05 [opt_a]: 0.488691, [2] [Cycle 1]: 0.254082, [45] [expand_dump_flag]: 3.01001e-06 [switch_simplify]: 3.999e-05 [loop_unroll]: 1.962e-05 [a_1]: 0.239488 [with_stream_mark]: 5.539e-05 [recompute_prepare]: 3.088e-05 [updatestate_depend_eliminate]: 1.345e-05 [updatestate_assign_eliminate]: 9.03002e-06 [updatestate_loads_eliminate]: 1.129e-05 [parameter_eliminate]: 3.16999e-06 [a_2]: 0.0002366 [accelerated_algorithm]: 4.869e-05 [shard]: 3.11999e-06 [meta_shard_fg_expand]: 7.43e-06 [shard_inline]: 1.795e-05 [merge_send_recv]: 1.642e-05 [auto_parallel]: 1.55e-05 [parallel]: 3.705e-05 [flash_sp]: 1.654e-05 [merge_comm]: 9.61e-06 [allreduce_fusion]: 7.97e-06 [matmul_add_comm_reduction]: 2.817e-05 [allreduce_slice_to_reducescatter]: 9.30013e-07 [virtual_shard_identity]: 2.025e-05 [virtual_dataset]: 1.579e-05 [get_grad_eliminate_]: 1.653e-05 [virtual_output]: 1.62e-05 [merge_forward]: 8.78001e-06 [cell_reuse_recompute_pass]: 3.53e-06 [offload_activation]: 1.904e-05 [cell_reuse_handle_not_recompute_node_pass]: 3.036e-05 [merge_recompute_call_nodes]: 1.76e-06 [before_grad]: 2.753e-05 [set_forward_comm_id_for_comm_node_pass]: 8.62998e-06 [meta_fg_expand]: 7.16001e-06 [flash_sp_send_recv_attached]: 7.61001e-06 [receive_attached]: 2.58e-06 [after_resolve]: 2.481e-05 [a_after_grad]: 2.605e-05 [renormalize]: 0.0122177 [add_forward_monad_depend]: 2.556e-05 [auto_monad_grad]: 3.31999e-06 [auto_monad_eliminator]: 7.281e-05 [cse]: 0.00085681 [a_3]: 0.00014736 [Cycle 2]: 0.234591, [45] [expand_dump_flag]: 4.31002e-06 [switch_simplify]: 2.037e-05 [loop_unroll]: 1.706e-05 [a_1]: 0.00052516 [with_stream_mark]: 3.711e-05 [recompute_prepare]: 2.104e-05 [updatestate_depend_eliminate]: 1.156e-05 [updatestate_assign_eliminate]: 8.34002e-06 [updatestate_loads_eliminate]: 1.182e-05 [parameter_eliminate]: 3.34001e-06 [a_2]: 0.00023265 [accelerated_algorithm]: 2.537e-05 [shard]: 3.35998e-06 [meta_shard_fg_expand]: 6.36e-06 [shard_inline]: 1.729e-05 [merge_send_recv]: 1.817e-05 [auto_parallel]: 1.735e-05 [parallel]: 1.097e-05 [flash_sp]: 4.57e-06 [merge_comm]: 1.063e-05 [allreduce_fusion]: 9.74999e-06 [matmul_add_comm_reduction]: 2.091e-05 [allreduce_slice_to_reducescatter]: 1.09998e-06 [virtual_shard_identity]: 2.161e-05 [virtual_dataset]: 1.614e-05 [get_grad_eliminate_]: 1.739e-05 [virtual_output]: 1.559e-05 [merge_forward]: 1.048e-05 [cell_reuse_recompute_pass]: 4.37e-06 [offload_activation]: 1.922e-05 [cell_reuse_handle_not_recompute_node_pass]: 3.387e-05 [merge_recompute_call_nodes]: 1.74998e-06 [before_grad]: 2.688e-05 [set_forward_comm_id_for_comm_node_pass]: 1.084e-05 [meta_fg_expand]: 7.09001e-06 [flash_sp_send_recv_attached]: 2.01e-06 [receive_attached]: 2.89999e-06 [after_resolve]: 2.291e-05 [a_after_grad]: 9.565e-05 [renormalize]: 1.10012e-07 [add_forward_monad_depend]: 1.048e-05 [auto_monad_grad]: 3.42002e-06 [auto_monad_eliminator]: 8.857e-05 [cse]: 0.0001941 [a_3]: 0.00012117 [py_interpret_to_execute_after_opt_a]: 3.211e-05 [slice_cell_reuse_recomputed_activation]: 2.83998e-06 [rewriter_after_opt_a]: 0.0002539 [convert_after_rewriter]: 1.373e-05 [order_py_execute_after_rewriter]: 1.155e-05 [mutable_eliminate]: 0.00080292 [opt_b]: 0.00054595, [1] [Cycle 1]: 0.00053708, [7] [b_1]: 0.00037225 [b_2]: 1.816e-05 [updatestate_depend_eliminate]: 1.425e-05 [updatestate_assign_eliminate]: 8.08999e-06 [updatestate_loads_eliminate]: 1.107e-05 [renormalize]: 9.20001e-07 [cse]: 7.02e-05 [optimize_parallel_all_gather_comm]: 3.209e-05 [overlap_param_gather]: 2.49999e-06 [cconv]: 4.073e-05 [loop_unroll]: 0.00053417 [opt_after_cconv]: 0.00023102, [1] [Cycle 1]: 0.00022325, [7] [c_1]: 0.00010317 [parameter_eliminate]: 4.75001e-06 [updatestate_depend_eliminate]: 1.356e-05 [updatestate_assign_eliminate]: 7.21999e-06 [updatestate_loads_eliminate]: 8.33001e-06 [cse]: 4.767e-05 [renormalize]: 6.29982e-07 [remove_dup_value]: 7.77e-05 [tuple_transform]: 0.00016332, [1] [Cycle 1]: 0.00015835, [4] [d_1]: 0.00011634 [none_parameter_eliminate]: 2.11998e-06 [renormalize]: 4.00003e-07 [switch_simplify]: 1.721e-05 [partial_unused_args_eliminate]: 2.48002e-06 [add_recomputation]: 0.00011675 [cse_after_recomputation]: 5.326e-05, [1] [Cycle 1]: 4.814e-05, [1] [cse]: 4.201e-05 [environ_conv]: 2.015e-05 [swap_dp_allreduce_reducescatter]: 1.277e-05 [bias_add_comm_swap]: 3.41999e-06 [label_micro_interleaved_index]: 5.81e-06 [label_fine_grained_interleaved_index]: 3.61001e-06 [merge_cast_opt]: 1.54e-06 [slice_recompute_activation]: 2.54999e-06 [micro_interleaved_order_control]: 2.68e-06 [assign_add_opt]: 1.45999e-06 [ForceFp32Comm]: 8.79983e-07 [remove_cast_before_assign_add]: 1.42999e-06 [full_micro_interleaved_order_control]: 2.22001e-06 [reorder_send_recv_between_fp_bp]: 2.92002e-06 [comm_op_add_attrs]: 1.10999e-06 [add_comm_op_reuse_tag]: 1.14e-06 [interleave_split_concat_branches]: 1.62001e-06 [interleave_parallel_branches]: 1.14e-06 [overlap_opt_shard_in_pipeline]: 1.77999e-06 [overlap_opt_shard_grad_in_pipeline]: 2.01998e-06 [control_data_broadcast_order]: 2.802e-05 [grouped_pairwise_exchange_alltoall]: 1.67001e-06 [offloading_packed_experts]: 8.13001e-06 [overlap_recompute_and_grad_model_parallel]: 7.90998e-06 [overlap_grad_matmul_and_grad_allreduce]: 1.41002e-06 [overlap_recompute_allgather_and_fa_grad]: 1.45999e-06 [overlap_recompute_comm]: 2.60997e-06 [overlap_grad_ring_attention]: 7.65e-06 [overlap_grad_flash_sp]: 3.958e-05 [begin_end_overlap_inline]: 5.50004e-07 [split_matmul_comm_elemetwise]: 2.45002e-06 [split_layernorm_comm]: 1.82001e-06 [handle_group_info]: 1.07e-06 [symbol_engine_optimizer]: 0.00038838, [1] [Cycle 1]: 0.00038128, [6] [build]: 1.978e-05 [elim_shapecalc]: 2.619e-05 [elim_not_effective]: 3.015e-05 [opt_reshape]: 2.354e-05 [fold_const_symbol]: 0.00023712 [renormalize]: 4.00003e-07 [detach_backward]: 3.08998e-06 [pipeline_parallel_scheduler]: 1.82001e-06 [auto_monad_reorder]: 5.671e-05 [get_jit_bprop_graph]: 2.73e-06 [rewriter_after_jit_bprop_graph]: 7.64002e-06 [opt_after_jit_grad]: 0.000705 [validate]: 0.00012704 Sums bootstrap : 0.000659s : 0.02% type_inference : 4.084897s : 94.00% event_method : 0.000022s : 0.00% auto_monad : 0.000382s : 0.01% graph_reusing : 0.000007s : 0.00% inline : 0.000003s : 0.00% add_attr.add_attr_with_inline.tag_attr : 0.000053s : 0.00% add_attr.add_attr_with_inline.meta_addattr_fg_expand : 0.000005s : 0.00% parallel-infer-symbol : 0.000006s : 0.00% pre_auto_parallel : 0.000040s : 0.00% insert-virtual-dataset : 0.000004s : 0.00% parallel-infer-symbol-second : 0.000001s : 0.00% dataset_repeat_opt : 0.000002s : 0.00% pipeline_split : 0.000002s : 0.00% optimize.py_interpret_to_execute : 0.000041s : 0.00% optimize.rewriter_before_opt_a : 0.000097s : 0.00% optimize.opt_a.expand_dump_flag : 0.000007s : 0.00% optimize.opt_a.switch_simplify : 0.000060s : 0.00% optimize.opt_a.loop_unroll : 0.000037s : 0.00% optimize.opt_a.a_1 : 0.240013s : 5.52% optimize.opt_a.with_stream_mark : 0.000093s : 0.00% optimize.opt_a.recompute_prepare : 0.000052s : 0.00% optimize.opt_a.updatestate_depend_eliminate : 0.000025s : 0.00% optimize.opt_a.updatestate_assign_eliminate : 0.000017s : 0.00% optimize.opt_a.updatestate_loads_eliminate : 0.000023s : 0.00% optimize.opt_a.parameter_eliminate : 0.000007s : 0.00% optimize.opt_a.a_2 : 0.000469s : 0.01% optimize.opt_a.accelerated_algorithm : 0.000074s : 0.00% optimize.opt_a.shard : 0.000006s : 0.00% optimize.opt_a.meta_shard_fg_expand : 0.000014s : 0.00% optimize.opt_a.shard_inline : 0.000035s : 0.00% optimize.opt_a.merge_send_recv : 0.000035s : 0.00% optimize.opt_a.auto_parallel : 0.000033s : 0.00% optimize.opt_a.parallel : 0.000048s : 0.00% optimize.opt_a.flash_sp : 0.000021s : 0.00% optimize.opt_a.merge_comm : 0.000020s : 0.00% optimize.opt_a.allreduce_fusion : 0.000018s : 0.00% optimize.opt_a.matmul_add_comm_reduction : 0.000049s : 0.00% optimize.opt_a.allreduce_slice_to_reducescatter : 0.000002s : 0.00% optimize.opt_a.virtual_shard_identity : 0.000042s : 0.00% optimize.opt_a.virtual_dataset : 0.000032s : 0.00% optimize.opt_a.get_grad_eliminate_ : 0.000034s : 0.00% optimize.opt_a.virtual_output : 0.000032s : 0.00% optimize.opt_a.merge_forward : 0.000019s : 0.00% optimize.opt_a.cell_reuse_recompute_pass : 0.000008s : 0.00% optimize.opt_a.offload_activation : 0.000038s : 0.00% optimize.opt_a.cell_reuse_handle_not_recompute_node_pass : 0.000064s : 0.00% optimize.opt_a.merge_recompute_call_nodes : 0.000004s : 0.00% optimize.opt_a.before_grad : 0.000054s : 0.00% optimize.opt_a.set_forward_comm_id_for_comm_node_pass : 0.000019s : 0.00% optimize.opt_a.meta_fg_expand : 0.000014s : 0.00% optimize.opt_a.flash_sp_send_recv_attached : 0.000010s : 0.00% optimize.opt_a.receive_attached : 0.000005s : 0.00% optimize.opt_a.after_resolve : 0.000048s : 0.00% optimize.opt_a.a_after_grad : 0.000122s : 0.00% optimize.opt_a.renormalize : 0.012218s : 0.28% optimize.opt_a.add_forward_monad_depend : 0.000036s : 0.00% optimize.opt_a.auto_monad_grad : 0.000007s : 0.00% optimize.opt_a.auto_monad_eliminator : 0.000161s : 0.00% optimize.opt_a.cse : 0.001051s : 0.02% optimize.opt_a.a_3 : 0.000269s : 0.01% optimize.py_interpret_to_execute_after_opt_a : 0.000032s : 0.00% optimize.slice_cell_reuse_recomputed_activation : 0.000003s : 0.00% optimize.rewriter_after_opt_a : 0.000254s : 0.01% optimize.convert_after_rewriter : 0.000014s : 0.00% optimize.order_py_execute_after_rewriter : 0.000012s : 0.00% optimize.mutable_eliminate : 0.000803s : 0.02% optimize.opt_b.b_1 : 0.000372s : 0.01% optimize.opt_b.b_2 : 0.000018s : 0.00% optimize.opt_b.updatestate_depend_eliminate : 0.000014s : 0.00% optimize.opt_b.updatestate_assign_eliminate : 0.000008s : 0.00% optimize.opt_b.updatestate_loads_eliminate : 0.000011s : 0.00% optimize.opt_b.renormalize : 0.000001s : 0.00% optimize.opt_b.cse : 0.000070s : 0.00% optimize.optimize_parallel_all_gather_comm : 0.000032s : 0.00% optimize.overlap_param_gather : 0.000002s : 0.00% optimize.cconv : 0.000041s : 0.00% optimize.loop_unroll : 0.000534s : 0.01% optimize.opt_after_cconv.c_1 : 0.000103s : 0.00% optimize.opt_after_cconv.parameter_eliminate : 0.000005s : 0.00% optimize.opt_after_cconv.updatestate_depend_eliminate : 0.000014s : 0.00% optimize.opt_after_cconv.updatestate_assign_eliminate : 0.000007s : 0.00% optimize.opt_after_cconv.updatestate_loads_eliminate : 0.000008s : 0.00% optimize.opt_after_cconv.cse : 0.000048s : 0.00% optimize.opt_after_cconv.renormalize : 0.000001s : 0.00% optimize.remove_dup_value : 0.000078s : 0.00% optimize.tuple_transform.d_1 : 0.000116s : 0.00% optimize.tuple_transform.none_parameter_eliminate : 0.000002s : 0.00% optimize.tuple_transform.renormalize : 0.000000s : 0.00% optimize.tuple_transform.switch_simplify : 0.000017s : 0.00% optimize.partial_unused_args_eliminate : 0.000002s : 0.00% optimize.add_recomputation : 0.000117s : 0.00% optimize.cse_after_recomputation.cse : 0.000042s : 0.00% optimize.environ_conv : 0.000020s : 0.00% optimize.swap_dp_allreduce_reducescatter : 0.000013s : 0.00% optimize.bias_add_comm_swap : 0.000003s : 0.00% optimize.label_micro_interleaved_index : 0.000006s : 0.00% optimize.label_fine_grained_interleaved_index : 0.000004s : 0.00% optimize.merge_cast_opt : 0.000002s : 0.00% optimize.slice_recompute_activation : 0.000003s : 0.00% optimize.micro_interleaved_order_control : 0.000003s : 0.00% optimize.assign_add_opt : 0.000001s : 0.00% optimize.ForceFp32Comm : 0.000001s : 0.00% optimize.remove_cast_before_assign_add : 0.000001s : 0.00% optimize.full_micro_interleaved_order_control : 0.000002s : 0.00% optimize.reorder_send_recv_between_fp_bp : 0.000003s : 0.00% optimize.comm_op_add_attrs : 0.000001s : 0.00% optimize.add_comm_op_reuse_tag : 0.000001s : 0.00% optimize.interleave_split_concat_branches : 0.000002s : 0.00% optimize.interleave_parallel_branches : 0.000001s : 0.00% optimize.overlap_opt_shard_in_pipeline : 0.000002s : 0.00% optimize.overlap_opt_shard_grad_in_pipeline : 0.000002s : 0.00% optimize.control_data_broadcast_order : 0.000028s : 0.00% optimize.grouped_pairwise_exchange_alltoall : 0.000002s : 0.00% optimize.offloading_packed_experts : 0.000008s : 0.00% optimize.overlap_recompute_and_grad_model_parallel : 0.000008s : 0.00% optimize.overlap_grad_matmul_and_grad_allreduce : 0.000001s : 0.00% optimize.overlap_recompute_allgather_and_fa_grad : 0.000001s : 0.00% optimize.overlap_recompute_comm : 0.000003s : 0.00% optimize.overlap_grad_ring_attention : 0.000008s : 0.00% optimize.overlap_grad_flash_sp : 0.000040s : 0.00% optimize.begin_end_overlap_inline : 0.000001s : 0.00% optimize.split_matmul_comm_elemetwise : 0.000002s : 0.00% optimize.split_layernorm_comm : 0.000002s : 0.00% optimize.handle_group_info : 0.000001s : 0.00% optimize.symbol_engine_optimizer.build : 0.000020s : 0.00% optimize.symbol_engine_optimizer.elim_shapecalc : 0.000026s : 0.00% optimize.symbol_engine_optimizer.elim_not_effective : 0.000030s : 0.00% optimize.symbol_engine_optimizer.opt_reshape : 0.000024s : 0.00% optimize.symbol_engine_optimizer.fold_const_symbol : 0.000237s : 0.01% optimize.symbol_engine_optimizer.renormalize : 0.000000s : 0.00% detach_backward : 0.000003s : 0.00% pipeline_parallel_scheduler : 0.000002s : 0.00% auto_monad_reorder : 0.000057s : 0.00% get_jit_bprop_graph : 0.000003s : 0.00% rewriter_after_jit_bprop_graph : 0.000008s : 0.00% opt_after_jit_grad : 0.000705s : 0.02% validate : 0.000127s : 0.00% Time group info: ------[substitution.] 0.000338 96 3.51% : 0.000012s : 2: substitution.depend_value_elim 1.16% : 0.000004s : 8: substitution.elim_not_effective 1.08% : 0.000004s : 8: substitution.fold_const_symbol 3.51% : 0.000012s : 13: substitution.graph_param_transform 41.66% : 0.000141s : 1: substitution.inline 2.83% : 0.000010s : 16: substitution.j_node_and_user_rematch 8.84% : 0.000030s : 2: substitution.less_batch_normalization 1.86% : 0.000006s : 6: substitution.load_eliminater 0.85% : 0.000003s : 2: substitution.opt_reshape 3.73% : 0.000013s : 16: substitution.remove_not_recompute_node 2.56% : 0.000009s : 4: substitution.replace_old_param 14.68% : 0.000050s : 4: substitution.reshape_eliminate 6.78% : 0.000023s : 6: substitution.updatestate_pure_node_eliminater 6.95% : 0.000023s : 8: substitution.updatestate_useless_node_eliminater ------[type_inference.] 4.084776 2 94.28% : 3.851148s : 1: type_inference.infer 5.72% : 0.233628s : 1: type_inference.specialize ------[replace.] 0.000022 1 100.00% : 0.000022s : 1: replace.inline ------[match.] 0.000139 1 100.00% : 0.000139s : 1: match.inline ------[predicate.] 0.000489 3002 1.01% : 0.000005s : 27: predicate.accumulaten_eliminater 0.77% : 0.000004s : 13: predicate.ad_related_special_op_eliminate 0.67% : 0.000003s : 26: predicate.addn_check_dump 0.86% : 0.000004s : 27: predicate.addn_zero_filter 0.78% : 0.000004s : 27: predicate.adjust_all_reduce_mul_add 2.20% : 0.000011s : 53: predicate.arithmetic_simplify 0.83% : 0.000004s : 27: predicate.cast_eliminate 0.78% : 0.000004s : 26: predicate.check_bprop_eliminate 0.73% : 0.000004s : 26: predicate.compare_switch_simplify 0.23% : 0.000001s : 13: predicate.const_output_eliminate 0.78% : 0.000004s : 26: predicate.depend_value_elim 0.79% : 0.000004s : 27: predicate.dict_get_item_const_eliminator 0.92% : 0.000004s : 27: predicate.dict_get_item_eliminator 0.76% : 0.000004s : 27: predicate.dict_set_item_eliminator 1.05% : 0.000005s : 26: predicate.dumpgradient_eliminate 0.23% : 0.000001s : 13: predicate.elim_not_effective 0.46% : 0.000002s : 13: predicate.elim_shapecalc_of_broadcastargs 1.30% : 0.000006s : 40: predicate.environ_add_const_eliminate 0.99% : 0.000005s : 40: predicate.environ_get_add_eliminate 1.08% : 0.000005s : 40: predicate.environ_get_depend_swap 1.82% : 0.000009s : 66: predicate.environ_get_eliminate 1.03% : 0.000005s : 40: predicate.environ_get_set_eliminate 0.73% : 0.000004s : 28: predicate.exchange_switch_depend_value 1.45% : 0.000007s : 28: predicate.float_depend_g_call 0.72% : 0.000004s : 26: predicate.float_environ_get_switch 1.10% : 0.000005s : 39: predicate.float_tuple_getitem_switch 0.24% : 0.000001s : 13: predicate.fold_const_symbol 0.76% : 0.000004s : 26: predicate.get_grad_eliminate 0.24% : 0.000001s : 13: predicate.graph_param_transform 0.75% : 0.000004s : 26: predicate.incorporate_call 0.69% : 0.000003s : 26: predicate.incorporate_call_switch 5.29% : 0.000026s : 133: predicate.inline 3.26% : 0.000016s : 26: predicate.inline_without_move 0.42% : 0.000002s : 26: predicate.j_node_and_user_rematch 1.13% : 0.000006s : 27: predicate.less_batch_normalization 1.67% : 0.000008s : 53: predicate.list_to_tuple_eliminator_ 2.15% : 0.000011s : 80: predicate.load_eliminater 0.80% : 0.000004s : 13: predicate.loop_unroll_after_grad 0.98% : 0.000005s : 31: predicate.loop_unroll_before_grad 1.57% : 0.000008s : 53: predicate.make_slice_get_slice_eliminator 0.75% : 0.000004s : 26: predicate.merge_addn 0.74% : 0.000004s : 26: predicate.micro_step_allgather_replace 0.82% : 0.000004s : 26: predicate.mini_step_allgather_replace 0.75% : 0.000004s : 27: predicate.minmaximum_grad 1.05% : 0.000005s : 13: predicate.mutable_eliminate 0.47% : 0.000002s : 13: predicate.opt_reshape 0.42% : 0.000002s : 13: predicate.parallel_virtual_node 1.14% : 0.000006s : 28: predicate.partial_defer_inline 1.62% : 0.000008s : 40: predicate.partial_eliminate 0.93% : 0.000005s : 27: predicate.print_const_string_wrapper 0.70% : 0.000003s : 26: predicate.reduce_all_const_elim 1.12% : 0.000005s : 27: predicate.reduce_eliminate 2.25% : 0.000011s : 80: predicate.redundant_stop_gradient_eliminater 0.49% : 0.000002s : 26: predicate.remove_not_recompute_node 1.53% : 0.000007s : 53: predicate.replace_applicator 0.57% : 0.000003s : 26: predicate.replace_old_param 0.28% : 0.000001s : 13: predicate.reset_defer_inline 1.09% : 0.000005s : 27: predicate.reshape_eliminate 0.76% : 0.000004s : 26: predicate.row_tensor_add_zeros_like 0.45% : 0.000002s : 13: predicate.row_tensor_eliminate 1.33% : 0.000007s : 26: predicate.same_eliminate 0.62% : 0.000003s : 26: predicate.set_cell_output_no_recompute 1.02% : 0.000005s : 26: predicate.shard_identity_eliminate 0.81% : 0.000004s : 26: predicate.special_op_eliminate 0.81% : 0.000004s : 26: predicate.specialize_transform 1.28% : 0.000006s : 26: predicate.split_environ_get_set_with_tuple_value 1.47% : 0.000007s : 26: predicate.stack_unstack_eliminate 0.46% : 0.000002s : 13: predicate.switch_call_monad_eliminater 0.84% : 0.000004s : 28: predicate.switch_defer_inline 1.53% : 0.000007s : 54: predicate.switch_layer_defer_inline 3.30% : 0.000016s : 98: predicate.switch_simplify 0.85% : 0.000004s : 27: predicate.tile_eliminate 0.87% : 0.000004s : 27: predicate.transpose_eliminate 1.70% : 0.000008s : 53: predicate.tuple_list_convert_item_index_to_positive 1.67% : 0.000008s : 53: predicate.tuple_list_get_item_const_eliminator 1.44% : 0.000007s : 53: predicate.tuple_list_get_item_depend_reorder 2.97% : 0.000015s : 79: predicate.tuple_list_get_item_eliminator 1.64% : 0.000008s : 53: predicate.tuple_list_get_set_item_eliminator 2.51% : 0.000012s : 79: predicate.tuple_list_set_item_eliminator 1.50% : 0.000007s : 53: predicate.tuple_to_list_eliminator_ 3.44% : 0.000017s : 80: predicate.updatestate_pure_node_eliminater 3.06% : 0.000015s : 106: predicate.updatestate_useless_node_eliminater 0.50% : 0.000002s : 13: predicate.value_based_eliminate 0.80% : 0.000004s : 26: predicate.virtual_dataset_eliminate 0.76% : 0.000004s : 26: predicate.virtual_output_eliminate 0.39% : 0.000002s : 13: predicate.virtual_view_grad_eliminate 0.44% : 0.000002s : 13: predicate.zero_like_fill_zero ------[func_graph_cloner_run.] 0.089503 58 95.36% : 0.085350s : 55: func_graph_cloner_run.FuncGraphClonerGraph 4.64% : 0.004152s : 3: func_graph_cloner_run.FuncGraphSpecializer ------[meta_graph.] 0.000000 0 ------[manager.] 0.000000 0 ------[pynative] 0.000000 0 ------[others.] 5.343462 192 0.00% : 0.000004s : 1: ForceFp32Comm 0.16% : 0.008451s : 1: add_attr 0.16% : 0.008428s : 1: add_attr_with_inline 0.00% : 0.000004s : 1: add_comm_op_reuse_tag 0.00% : 0.000122s : 1: add_recomputation 0.00% : 0.000004s : 1: assign_add_opt 0.01% : 0.000396s : 1: auto_monad 0.00% : 0.000062s : 1: auto_monad_reorder 0.00% : 0.000003s : 1: begin_end_overlap_inline 0.00% : 0.000006s : 1: bias_add_comm_swap 0.01% : 0.000698s : 1: bootstrap 0.00% : 0.000045s : 1: cconv 0.00% : 0.000005s : 1: comm_op_add_attrs 0.00% : 0.000031s : 1: control_data_broadcast_order 0.00% : 0.000018s : 1: convert_after_rewriter 0.00% : 0.000056s : 1: cse_after_recomputation 0.00% : 0.000005s : 1: dataset_repeat_opt 0.00% : 0.000006s : 1: detach_backward 0.00% : 0.000024s : 1: environ_conv 0.00% : 0.000032s : 1: event_method 0.00% : 0.000005s : 1: full_micro_interleaved_order_control 0.00% : 0.000007s : 1: get_jit_bprop_graph 0.00% : 0.000013s : 1: graph_reusing 0.00% : 0.000005s : 1: grouped_pairwise_exchange_alltoall 0.00% : 0.000004s : 1: handle_group_info 0.00% : 0.000007s : 1: inline 0.00% : 0.000009s : 1: insert-virtual-dataset 0.00% : 0.000004s : 1: interleave_parallel_branches 0.00% : 0.000004s : 1: interleave_split_concat_branches 0.00% : 0.000007s : 1: label_fine_grained_interleaved_index 0.00% : 0.000009s : 1: label_micro_interleaved_index 0.01% : 0.000545s : 1: loop_unroll 0.00% : 0.000005s : 1: merge_cast_opt 0.00% : 0.000006s : 1: micro_interleaved_order_control 0.02% : 0.000817s : 1: mutable_eliminate 0.00% : 0.000012s : 1: offloading_packed_experts 0.00% : 0.000027s : 1: opt.transform.loop_unroll_optimizer 0.00% : 0.000034s : 1: opt.transform.mutable_eliminate 4.52% : 0.241288s : 78: opt.transform.opt_a 0.00% : 0.000101s : 1: opt.transform.opt_after_cconv 0.00% : 0.000072s : 1: opt.transform.opt_after_jit_grad 0.01% : 0.000351s : 28: opt.transform.opt_b 0.00% : 0.000131s : 2: opt.transform.opt_trans_graph 0.01% : 0.000313s : 4: opt.transform.symbol_engine_opt 9.15% : 0.488695s : 1: opt_a 0.00% : 0.000235s : 1: opt_after_cconv 0.01% : 0.000719s : 1: opt_after_jit_grad 0.01% : 0.000550s : 1: opt_b 9.22% : 0.492593s : 1: optimize 0.00% : 0.000036s : 1: optimize_parallel_all_gather_comm 0.00% : 0.000015s : 1: order_py_execute_after_rewriter 0.00% : 0.000043s : 1: overlap_grad_flash_sp 0.00% : 0.000004s : 1: overlap_grad_matmul_and_grad_allreduce 0.00% : 0.000012s : 1: overlap_grad_ring_attention 0.00% : 0.000005s : 1: overlap_opt_shard_grad_in_pipeline 0.00% : 0.000005s : 1: overlap_opt_shard_in_pipeline 0.00% : 0.000007s : 1: overlap_param_gather 0.00% : 0.000006s : 1: overlap_recompute_allgather_and_fa_grad 0.00% : 0.000011s : 1: overlap_recompute_and_grad_model_parallel 0.00% : 0.000005s : 1: overlap_recompute_comm 0.00% : 0.000010s : 1: parallel-infer-symbol 0.00% : 0.000004s : 1: parallel-infer-symbol-second 0.00% : 0.000006s : 1: partial_unused_args_eliminate 0.00% : 0.000005s : 1: pipeline_parallel_scheduler 0.00% : 0.000006s : 1: pipeline_split 0.00% : 0.000044s : 1: pre_auto_parallel 0.00% : 0.000046s : 1: py_interpret_to_execute 0.00% : 0.000036s : 1: py_interpret_to_execute_after_opt_a 0.00% : 0.000004s : 1: remove_cast_before_assign_add 0.00% : 0.000082s : 1: remove_dup_value 0.14% : 0.007518s : 1: renormalize.infer 0.09% : 0.004680s : 1: renormalize.specialize 0.00% : 0.000006s : 1: reorder_send_recv_between_fp_bp 0.00% : 0.000011s : 1: rewriter_after_jit_bprop_graph 0.00% : 0.000260s : 1: rewriter_after_opt_a 0.00% : 0.000102s : 1: rewriter_before_opt_a 0.00% : 0.000006s : 1: slice_cell_reuse_recomputed_activation 0.00% : 0.000005s : 1: slice_recompute_activation 0.00% : 0.000006s : 1: split_layernorm_comm 0.00% : 0.000005s : 1: split_matmul_comm_elemetwise 0.00% : 0.000016s : 1: swap_dp_allreduce_reducescatter 0.01% : 0.000392s : 1: symbol_engine_optimizer 0.00% : 0.000166s : 1: tuple_transform 76.45% : 4.084929s : 1: type_inference . [hook] pytest_runtest_teardown:test_matmul_qkv_bf16_1[False-mstype0-k_n_shape0-1024] tests/st/infer/ops/test_internal_ops/test_matmul_split.py::test_matmul_qkv_bf16_1[False-mstype0-k_n_shape0-1024],max_mem:142.0M [WARNING] ME(168319:281472866918192,MainProcess):2026-01-29-17:39:29.801.076 [mindspore/context.py:1334] For 'context.set_context', the parameter 'device_target' will be deprecated and removed in a future version. Please use the api mindspore.set_device() instead. TotalTime = 3.81263, [21] [bootstrap]: 0.00086465 [type_inference]: 3.45241 [event_method]: 1.99e-05 [auto_monad]: 0.00040461 [graph_reusing]: 6.51999e-06 [inline]: 2.84999e-06 [add_attr]: 0.00701538, [1] [add_attr_with_inline]: 0.00699567, [1] [Cycle 1]: 7.859e-05, [2] [tag_attr]: 2.303e-05 [meta_addattr_fg_expand]: 4e-06 [parallel-infer-symbol]: 4.74998e-06 [pre_auto_parallel]: 3.837e-05 [insert-virtual-dataset]: 2.82002e-06 [parallel-infer-symbol-second]: 7.89994e-07 [dataset_repeat_opt]: 2.44999e-06 [pipeline_split]: 1.54998e-06 [optimize]: 0.350692, [53] [py_interpret_to_execute]: 3.705e-05 [rewriter_before_opt_a]: 8.88e-05 [opt_a]: 0.116108, [2] [Cycle 1]: 0.113847, [45] [expand_dump_flag]: 3.28e-06 [switch_simplify]: 3.674e-05 [loop_unroll]: 1.989e-05 [a_1]: 0.00095244 [with_stream_mark]: 3.016e-05 [recompute_prepare]: 2.514e-05 [updatestate_depend_eliminate]: 1.172e-05 [updatestate_assign_eliminate]: 9.92001e-06 [updatestate_loads_eliminate]: 1.142e-05 [parameter_eliminate]: 2.83998e-06 [a_2]: 0.0002556 [accelerated_algorithm]: 4.566e-05 [shard]: 2.73e-06 [meta_shard_fg_expand]: 4.28999e-06 [shard_inline]: 1.703e-05 [merge_send_recv]: 1.472e-05 [auto_parallel]: 1.292e-05 [parallel]: 3.95e-05 [flash_sp]: 1.39e-05 [merge_comm]: 9.39e-06 [allreduce_fusion]: 8.30999e-06 [matmul_add_comm_reduction]: 1.823e-05 [allreduce_slice_to_reducescatter]: 6.69999e-07 [virtual_shard_identity]: 2.187e-05 [virtual_dataset]: 1.878e-05 [get_grad_eliminate_]: 1.963e-05 [virtual_output]: 1.691e-05 [merge_forward]: 9.24e-06 [cell_reuse_recompute_pass]: 2.82002e-06 [offload_activation]: 2.207e-05 [cell_reuse_handle_not_recompute_node_pass]: 4.473e-05 [merge_recompute_call_nodes]: 1.91998e-06 [before_grad]: 2.965e-05 [set_forward_comm_id_for_comm_node_pass]: 9.52999e-06 [meta_fg_expand]: 7.65e-06 [flash_sp_send_recv_attached]: 7.82e-06 [receive_attached]: 2.83e-06 [after_resolve]: 2.61e-05 [a_after_grad]: 3.857e-05 [renormalize]: 0.110635 [add_forward_monad_depend]: 1.987e-05 [auto_monad_grad]: 3.11001e-06 [auto_monad_eliminator]: 6.509e-05 [cse]: 0.00079503 [a_3]: 0.00014568 [Cycle 2]: 0.00224504, [45] [expand_dump_flag]: 4.37e-06 [switch_simplify]: 1.961e-05 [loop_unroll]: 1.662e-05 [a_1]: 0.00049557 [with_stream_mark]: 3.088e-05 [recompute_prepare]: 2.05e-05 [updatestate_depend_eliminate]: 1.037e-05 [updatestate_assign_eliminate]: 8.15e-06 [updatestate_loads_eliminate]: 1.11e-05 [parameter_eliminate]: 2.84999e-06 [a_2]: 0.00026986 [accelerated_algorithm]: 2.442e-05 [shard]: 2.68e-06 [meta_shard_fg_expand]: 4.94e-06 [shard_inline]: 1.59e-05 [merge_send_recv]: 1.634e-05 [auto_parallel]: 1.651e-05 [parallel]: 1.092e-05 [flash_sp]: 4.90999e-06 [merge_comm]: 9.10999e-06 [allreduce_fusion]: 9.37001e-06 [matmul_add_comm_reduction]: 1.951e-05 [allreduce_slice_to_reducescatter]: 6.69999e-07 [virtual_shard_identity]: 2.17e-05 [virtual_dataset]: 1.625e-05 [get_grad_eliminate_]: 1.557e-05 [virtual_output]: 1.486e-05 [merge_forward]: 1.102e-05 [cell_reuse_recompute_pass]: 3.5e-06 [offload_activation]: 1.888e-05 [cell_reuse_handle_not_recompute_node_pass]: 3.167e-05 [merge_recompute_call_nodes]: 1.74e-06 [before_grad]: 2.59e-05 [set_forward_comm_id_for_comm_node_pass]: 1.012e-05 [meta_fg_expand]: 6.34001e-06 [flash_sp_send_recv_attached]: 1.97001e-06 [receive_attached]: 2.98e-06 [after_resolve]: 2.449e-05 [a_after_grad]: 2.584e-05 [renormalize]: 9.00181e-08 [add_forward_monad_depend]: 4.07e-06 [auto_monad_grad]: 3.15998e-06 [auto_monad_eliminator]: 7.254e-05 [cse]: 0.00050926 [a_3]: 0.00012043 [py_interpret_to_execute_after_opt_a]: 3.277e-05 [slice_cell_reuse_recomputed_activation]: 3.14999e-06 [rewriter_after_opt_a]: 0.00030707 [convert_after_rewriter]: 1.824e-05 [order_py_execute_after_rewriter]: 1.068e-05 [mutable_eliminate]: 0.00082627 [opt_b]: 0.230994, [1] [Cycle 1]: 0.230981, [7] [b_1]: 0.230718 [b_2]: 2.737e-05 [updatestate_depend_eliminate]: 2.88e-05 [updatestate_assign_eliminate]: 9.41e-06 [updatestate_loads_eliminate]: 1.216e-05 [renormalize]: 4.2998e-07 [cse]: 0.00010371 [optimize_parallel_all_gather_comm]: 4.208e-05 [overlap_param_gather]: 3.13e-06 [cconv]: 4.344e-05 [loop_unroll]: 0.00083173 [opt_after_cconv]: 0.00026832, [1] [Cycle 1]: 0.00025708, [7] [c_1]: 0.00010947 [parameter_eliminate]: 6.51999e-06 [updatestate_depend_eliminate]: 1.752e-05 [updatestate_assign_eliminate]: 7.5e-06 [updatestate_loads_eliminate]: 1.014e-05 [cse]: 6.357e-05 [renormalize]: 5.69999e-07 [remove_dup_value]: 7.93e-05 [tuple_transform]: 0.00016997, [1] [Cycle 1]: 0.00016333, [4] [d_1]: 0.0001206 [none_parameter_eliminate]: 2.44999e-06 [renormalize]: 3.69997e-07 [switch_simplify]: 1.669e-05 [partial_unused_args_eliminate]: 2.37999e-06 [add_recomputation]: 0.00012404 [cse_after_recomputation]: 6.042e-05, [1] [Cycle 1]: 5.463e-05, [1] [cse]: 4.709e-05 [environ_conv]: 2.113e-05 [swap_dp_allreduce_reducescatter]: 1.236e-05 [bias_add_comm_swap]: 3.93999e-06 [label_micro_interleaved_index]: 6.83e-06 [label_fine_grained_interleaved_index]: 2.99999e-06 [merge_cast_opt]: 1.69e-06 [slice_recompute_activation]: 2.27999e-06 [micro_interleaved_order_control]: 2.48002e-06 [assign_add_opt]: 1.52001e-06 [ForceFp32Comm]: 1.04e-06 [remove_cast_before_assign_add]: 1.24e-06 [full_micro_interleaved_order_control]: 2.46998e-06 [reorder_send_recv_between_fp_bp]: 2.78e-06 [comm_op_add_attrs]: 1.12999e-06 [add_comm_op_reuse_tag]: 1.02e-06 [interleave_split_concat_branches]: 1.56998e-06 [interleave_parallel_branches]: 1.22999e-06 [overlap_opt_shard_in_pipeline]: 3.8e-06 [overlap_opt_shard_grad_in_pipeline]: 1.95001e-06 [control_data_broadcast_order]: 2.765e-05 [grouped_pairwise_exchange_alltoall]: 2.37999e-06 [offloading_packed_experts]: 7.42002e-06 [overlap_recompute_and_grad_model_parallel]: 7.62998e-06 [overlap_grad_matmul_and_grad_allreduce]: 1.23002e-06 [overlap_recompute_allgather_and_fa_grad]: 1.31002e-06 [overlap_recompute_comm]: 2.43e-06 [overlap_grad_ring_attention]: 8.72e-06 [overlap_grad_flash_sp]: 4.087e-05 [begin_end_overlap_inline]: 8.2e-07 [split_matmul_comm_elemetwise]: 2.27001e-06 [split_layernorm_comm]: 1.97001e-06 [handle_group_info]: 1.14998e-06 [symbol_engine_optimizer]: 0.00016184, [1] [Cycle 1]: 0.00015511, [6] [build]: 1.65e-05 [elim_shapecalc]: 2.398e-05 [elim_not_effective]: 3.067e-05 [opt_reshape]: 2.304e-05 [fold_const_symbol]: 2.551e-05 [renormalize]: 2.00002e-07 [detach_backward]: 2.53998e-06 [pipeline_parallel_scheduler]: 1.80001e-06 [auto_monad_reorder]: 4.928e-05 [get_jit_bprop_graph]: 2.19999e-06 [rewriter_after_jit_bprop_graph]: 6.51999e-06 [opt_after_jit_grad]: 0.00073117 [validate]: 0.00011004 Sums bootstrap : 0.000865s : 0.02% type_inference : 3.452409s : 90.75% event_method : 0.000020s : 0.00% auto_monad : 0.000405s : 0.01% graph_reusing : 0.000007s : 0.00% inline : 0.000003s : 0.00% add_attr.add_attr_with_inline.tag_attr : 0.000023s : 0.00% add_attr.add_attr_with_inline.meta_addattr_fg_expand : 0.000004s : 0.00% parallel-infer-symbol : 0.000005s : 0.00% pre_auto_parallel : 0.000038s : 0.00% insert-virtual-dataset : 0.000003s : 0.00% parallel-infer-symbol-second : 0.000001s : 0.00% dataset_repeat_opt : 0.000002s : 0.00% pipeline_split : 0.000002s : 0.00% optimize.py_interpret_to_execute : 0.000037s : 0.00% optimize.rewriter_before_opt_a : 0.000089s : 0.00% optimize.opt_a.expand_dump_flag : 0.000008s : 0.00% optimize.opt_a.switch_simplify : 0.000056s : 0.00% optimize.opt_a.loop_unroll : 0.000037s : 0.00% optimize.opt_a.a_1 : 0.001448s : 0.04% optimize.opt_a.with_stream_mark : 0.000061s : 0.00% optimize.opt_a.recompute_prepare : 0.000046s : 0.00% optimize.opt_a.updatestate_depend_eliminate : 0.000022s : 0.00% optimize.opt_a.updatestate_assign_eliminate : 0.000018s : 0.00% optimize.opt_a.updatestate_loads_eliminate : 0.000023s : 0.00% optimize.opt_a.parameter_eliminate : 0.000006s : 0.00% optimize.opt_a.a_2 : 0.000525s : 0.01% optimize.opt_a.accelerated_algorithm : 0.000070s : 0.00% optimize.opt_a.shard : 0.000005s : 0.00% optimize.opt_a.meta_shard_fg_expand : 0.000009s : 0.00% optimize.opt_a.shard_inline : 0.000033s : 0.00% optimize.opt_a.merge_send_recv : 0.000031s : 0.00% optimize.opt_a.auto_parallel : 0.000029s : 0.00% optimize.opt_a.parallel : 0.000050s : 0.00% optimize.opt_a.flash_sp : 0.000019s : 0.00% optimize.opt_a.merge_comm : 0.000018s : 0.00% optimize.opt_a.allreduce_fusion : 0.000018s : 0.00% optimize.opt_a.matmul_add_comm_reduction : 0.000038s : 0.00% optimize.opt_a.allreduce_slice_to_reducescatter : 0.000001s : 0.00% optimize.opt_a.virtual_shard_identity : 0.000044s : 0.00% optimize.opt_a.virtual_dataset : 0.000035s : 0.00% optimize.opt_a.get_grad_eliminate_ : 0.000035s : 0.00% optimize.opt_a.virtual_output : 0.000032s : 0.00% optimize.opt_a.merge_forward : 0.000020s : 0.00% optimize.opt_a.cell_reuse_recompute_pass : 0.000006s : 0.00% optimize.opt_a.offload_activation : 0.000041s : 0.00% optimize.opt_a.cell_reuse_handle_not_recompute_node_pass : 0.000076s : 0.00% optimize.opt_a.merge_recompute_call_nodes : 0.000004s : 0.00% optimize.opt_a.before_grad : 0.000056s : 0.00% optimize.opt_a.set_forward_comm_id_for_comm_node_pass : 0.000020s : 0.00% optimize.opt_a.meta_fg_expand : 0.000014s : 0.00% optimize.opt_a.flash_sp_send_recv_attached : 0.000010s : 0.00% optimize.opt_a.receive_attached : 0.000006s : 0.00% optimize.opt_a.after_resolve : 0.000051s : 0.00% optimize.opt_a.a_after_grad : 0.000064s : 0.00% optimize.opt_a.renormalize : 0.110635s : 2.91% optimize.opt_a.add_forward_monad_depend : 0.000024s : 0.00% optimize.opt_a.auto_monad_grad : 0.000006s : 0.00% optimize.opt_a.auto_monad_eliminator : 0.000138s : 0.00% optimize.opt_a.cse : 0.001304s : 0.03% optimize.opt_a.a_3 : 0.000266s : 0.01% optimize.py_interpret_to_execute_after_opt_a : 0.000033s : 0.00% optimize.slice_cell_reuse_recomputed_activation : 0.000003s : 0.00% optimize.rewriter_after_opt_a : 0.000307s : 0.01% optimize.convert_after_rewriter : 0.000018s : 0.00% optimize.order_py_execute_after_rewriter : 0.000011s : 0.00% optimize.mutable_eliminate : 0.000826s : 0.02% optimize.opt_b.b_1 : 0.230718s : 6.06% optimize.opt_b.b_2 : 0.000027s : 0.00% optimize.opt_b.updatestate_depend_eliminate : 0.000029s : 0.00% optimize.opt_b.updatestate_assign_eliminate : 0.000009s : 0.00% optimize.opt_b.updatestate_loads_eliminate : 0.000012s : 0.00% optimize.opt_b.renormalize : 0.000000s : 0.00% optimize.opt_b.cse : 0.000104s : 0.00% optimize.optimize_parallel_all_gather_comm : 0.000042s : 0.00% optimize.overlap_param_gather : 0.000003s : 0.00% optimize.cconv : 0.000043s : 0.00% optimize.loop_unroll : 0.000832s : 0.02% optimize.opt_after_cconv.c_1 : 0.000109s : 0.00% optimize.opt_after_cconv.parameter_eliminate : 0.000007s : 0.00% optimize.opt_after_cconv.updatestate_depend_eliminate : 0.000018s : 0.00% optimize.opt_after_cconv.updatestate_assign_eliminate : 0.000007s : 0.00% optimize.opt_after_cconv.updatestate_loads_eliminate : 0.000010s : 0.00% optimize.opt_after_cconv.cse : 0.000064s : 0.00% optimize.opt_after_cconv.renormalize : 0.000001s : 0.00% optimize.remove_dup_value : 0.000079s : 0.00% optimize.tuple_transform.d_1 : 0.000121s : 0.00% optimize.tuple_transform.none_parameter_eliminate : 0.000002s : 0.00% optimize.tuple_transform.renormalize : 0.000000s : 0.00% optimize.tuple_transform.switch_simplify : 0.000017s : 0.00% optimize.partial_unused_args_eliminate : 0.000002s : 0.00% optimize.add_recomputation : 0.000124s : 0.00% optimize.cse_after_recomputation.cse : 0.000047s : 0.00% optimize.environ_conv : 0.000021s : 0.00% optimize.swap_dp_allreduce_reducescatter : 0.000012s : 0.00% optimize.bias_add_comm_swap : 0.000004s : 0.00% optimize.label_micro_interleaved_index : 0.000007s : 0.00% optimize.label_fine_grained_interleaved_index : 0.000003s : 0.00% optimize.merge_cast_opt : 0.000002s : 0.00% optimize.slice_recompute_activation : 0.000002s : 0.00% optimize.micro_interleaved_order_control : 0.000002s : 0.00% optimize.assign_add_opt : 0.000002s : 0.00% optimize.ForceFp32Comm : 0.000001s : 0.00% optimize.remove_cast_before_assign_add : 0.000001s : 0.00% optimize.full_micro_interleaved_order_control : 0.000002s : 0.00% optimize.reorder_send_recv_between_fp_bp : 0.000003s : 0.00% optimize.comm_op_add_attrs : 0.000001s : 0.00% optimize.add_comm_op_reuse_tag : 0.000001s : 0.00% optimize.interleave_split_concat_branches : 0.000002s : 0.00% optimize.interleave_parallel_branches : 0.000001s : 0.00% optimize.overlap_opt_shard_in_pipeline : 0.000004s : 0.00% optimize.overlap_opt_shard_grad_in_pipeline : 0.000002s : 0.00% optimize.control_data_broadcast_order : 0.000028s : 0.00% optimize.grouped_pairwise_exchange_alltoall : 0.000002s : 0.00% optimize.offloading_packed_experts : 0.000007s : 0.00% optimize.overlap_recompute_and_grad_model_parallel : 0.000008s : 0.00% optimize.overlap_grad_matmul_and_grad_allreduce : 0.000001s : 0.00% optimize.overlap_recompute_allgather_and_fa_grad : 0.000001s : 0.00% optimize.overlap_recompute_comm : 0.000002s : 0.00% optimize.overlap_grad_ring_attention : 0.000009s : 0.00% optimize.overlap_grad_flash_sp : 0.000041s : 0.00% optimize.begin_end_overlap_inline : 0.000001s : 0.00% optimize.split_matmul_comm_elemetwise : 0.000002s : 0.00% optimize.split_layernorm_comm : 0.000002s : 0.00% optimize.handle_group_info : 0.000001s : 0.00% optimize.symbol_engine_optimizer.build : 0.000016s : 0.00% optimize.symbol_engine_optimizer.elim_shapecalc : 0.000024s : 0.00% optimize.symbol_engine_optimizer.elim_not_effective : 0.000031s : 0.00% optimize.symbol_engine_optimizer.opt_reshape : 0.000023s : 0.00% optimize.symbol_engine_optimizer.fold_const_symbol : 0.000026s : 0.00% optimize.symbol_engine_optimizer.renormalize : 0.000000s : 0.00% detach_backward : 0.000003s : 0.00% pipeline_parallel_scheduler : 0.000002s : 0.00% auto_monad_reorder : 0.000049s : 0.00% get_jit_bprop_graph : 0.000002s : 0.00% rewriter_after_jit_bprop_graph : 0.000007s : 0.00% opt_after_jit_grad : 0.000731s : 0.02% validate : 0.000110s : 0.00% Time group info: ------[substitution.] 0.000303 96 3.67% : 0.000011s : 2: substitution.depend_value_elim 1.28% : 0.000004s : 8: substitution.elim_not_effective 1.18% : 0.000004s : 8: substitution.fold_const_symbol 4.06% : 0.000012s : 13: substitution.graph_param_transform 41.86% : 0.000127s : 1: substitution.inline 2.95% : 0.000009s : 16: substitution.j_node_and_user_rematch 7.51% : 0.000023s : 2: substitution.less_batch_normalization 1.99% : 0.000006s : 6: substitution.load_eliminater 0.98% : 0.000003s : 2: substitution.opt_reshape 4.32% : 0.000013s : 16: substitution.remove_not_recompute_node 2.91% : 0.000009s : 4: substitution.replace_old_param 11.19% : 0.000034s : 4: substitution.reshape_eliminate 8.77% : 0.000027s : 6: substitution.updatestate_pure_node_eliminater 7.32% : 0.000022s : 8: substitution.updatestate_useless_node_eliminater ------[type_inference.] 3.452292 2 96.27% : 3.323471s : 1: type_inference.infer 3.73% : 0.128821s : 1: type_inference.specialize ------[replace.] 0.000019 1 100.00% : 0.000019s : 1: replace.inline ------[match.] 0.000126 1 100.00% : 0.000126s : 1: match.inline ------[predicate.] 0.000480 3002 0.75% : 0.000004s : 27: predicate.accumulaten_eliminater 1.01% : 0.000005s : 13: predicate.ad_related_special_op_eliminate 0.69% : 0.000003s : 26: predicate.addn_check_dump 0.74% : 0.000004s : 27: predicate.addn_zero_filter 0.67% : 0.000003s : 27: predicate.adjust_all_reduce_mul_add 1.96% : 0.000009s : 53: predicate.arithmetic_simplify 0.77% : 0.000004s : 27: predicate.cast_eliminate 0.74% : 0.000004s : 26: predicate.check_bprop_eliminate 0.70% : 0.000003s : 26: predicate.compare_switch_simplify 0.22% : 0.000001s : 13: predicate.const_output_eliminate 0.76% : 0.000004s : 26: predicate.depend_value_elim 0.76% : 0.000004s : 27: predicate.dict_get_item_const_eliminator 0.83% : 0.000004s : 27: predicate.dict_get_item_eliminator 0.77% : 0.000004s : 27: predicate.dict_set_item_eliminator 1.17% : 0.000006s : 26: predicate.dumpgradient_eliminate 0.25% : 0.000001s : 13: predicate.elim_not_effective 0.44% : 0.000002s : 13: predicate.elim_shapecalc_of_broadcastargs 1.07% : 0.000005s : 40: predicate.environ_add_const_eliminate 1.03% : 0.000005s : 40: predicate.environ_get_add_eliminate 1.03% : 0.000005s : 40: predicate.environ_get_depend_swap 1.78% : 0.000009s : 66: predicate.environ_get_eliminate 1.06% : 0.000005s : 40: predicate.environ_get_set_eliminate 0.71% : 0.000003s : 28: predicate.exchange_switch_depend_value 1.25% : 0.000006s : 28: predicate.float_depend_g_call 0.69% : 0.000003s : 26: predicate.float_environ_get_switch 1.62% : 0.000008s : 39: predicate.float_tuple_getitem_switch 0.21% : 0.000001s : 13: predicate.fold_const_symbol 0.88% : 0.000004s : 26: predicate.get_grad_eliminate 0.29% : 0.000001s : 13: predicate.graph_param_transform 0.69% : 0.000003s : 26: predicate.incorporate_call 0.69% : 0.000003s : 26: predicate.incorporate_call_switch 5.47% : 0.000026s : 133: predicate.inline 0.90% : 0.000004s : 26: predicate.inline_without_move 0.40% : 0.000002s : 26: predicate.j_node_and_user_rematch 1.19% : 0.000006s : 27: predicate.less_batch_normalization 1.56% : 0.000008s : 53: predicate.list_to_tuple_eliminator_ 2.27% : 0.000011s : 80: predicate.load_eliminater 1.01% : 0.000005s : 13: predicate.loop_unroll_after_grad 0.93% : 0.000004s : 31: predicate.loop_unroll_before_grad 3.48% : 0.000017s : 53: predicate.make_slice_get_slice_eliminator 0.72% : 0.000003s : 26: predicate.merge_addn 0.73% : 0.000003s : 26: predicate.micro_step_allgather_replace 0.71% : 0.000003s : 26: predicate.mini_step_allgather_replace 0.65% : 0.000003s : 27: predicate.minmaximum_grad 0.97% : 0.000005s : 13: predicate.mutable_eliminate 0.50% : 0.000002s : 13: predicate.opt_reshape 0.49% : 0.000002s : 13: predicate.parallel_virtual_node 1.03% : 0.000005s : 28: predicate.partial_defer_inline 1.16% : 0.000006s : 40: predicate.partial_eliminate 0.74% : 0.000004s : 27: predicate.print_const_string_wrapper 0.73% : 0.000004s : 26: predicate.reduce_all_const_elim 1.02% : 0.000005s : 27: predicate.reduce_eliminate 2.11% : 0.000010s : 80: predicate.redundant_stop_gradient_eliminater 0.50% : 0.000002s : 26: predicate.remove_not_recompute_node 1.27% : 0.000006s : 53: predicate.replace_applicator 0.59% : 0.000003s : 26: predicate.replace_old_param 0.50% : 0.000002s : 13: predicate.reset_defer_inline 4.89% : 0.000023s : 27: predicate.reshape_eliminate 0.76% : 0.000004s : 26: predicate.row_tensor_add_zeros_like 0.63% : 0.000003s : 13: predicate.row_tensor_eliminate 1.10% : 0.000005s : 26: predicate.same_eliminate 0.58% : 0.000003s : 26: predicate.set_cell_output_no_recompute 0.93% : 0.000004s : 26: predicate.shard_identity_eliminate 1.00% : 0.000005s : 26: predicate.special_op_eliminate 0.85% : 0.000004s : 26: predicate.specialize_transform 1.30% : 0.000006s : 26: predicate.split_environ_get_set_with_tuple_value 0.89% : 0.000004s : 26: predicate.stack_unstack_eliminate 0.44% : 0.000002s : 13: predicate.switch_call_monad_eliminater 0.84% : 0.000004s : 28: predicate.switch_defer_inline 1.52% : 0.000007s : 54: predicate.switch_layer_defer_inline 3.10% : 0.000015s : 98: predicate.switch_simplify 0.89% : 0.000004s : 27: predicate.tile_eliminate 0.81% : 0.000004s : 27: predicate.transpose_eliminate 1.65% : 0.000008s : 53: predicate.tuple_list_convert_item_index_to_positive 1.74% : 0.000008s : 53: predicate.tuple_list_get_item_const_eliminator 1.58% : 0.000008s : 53: predicate.tuple_list_get_item_depend_reorder 2.86% : 0.000014s : 79: predicate.tuple_list_get_item_eliminator 1.74% : 0.000008s : 53: predicate.tuple_list_get_set_item_eliminator 2.44% : 0.000012s : 79: predicate.tuple_list_set_item_eliminator 1.65% : 0.000008s : 53: predicate.tuple_to_list_eliminator_ 2.08% : 0.000010s : 80: predicate.updatestate_pure_node_eliminater 3.07% : 0.000015s : 106: predicate.updatestate_useless_node_eliminater 0.47% : 0.000002s : 13: predicate.value_based_eliminate 0.79% : 0.000004s : 26: predicate.virtual_dataset_eliminate 0.77% : 0.000004s : 26: predicate.virtual_output_eliminate 0.36% : 0.000002s : 13: predicate.virtual_view_grad_eliminate 0.44% : 0.000002s : 13: predicate.zero_like_fill_zero ------[func_graph_cloner_run.] 0.014150 58 76.69% : 0.010852s : 55: func_graph_cloner_run.FuncGraphClonerGraph 23.31% : 0.003298s : 3: func_graph_cloner_run.FuncGraphSpecializer ------[meta_graph.] 0.000000 0 ------[manager.] 0.000000 0 ------[pynative] 0.000000 0 ------[others.] 4.514597 192 0.00% : 0.000004s : 1: ForceFp32Comm 0.16% : 0.007022s : 1: add_attr 0.16% : 0.007000s : 1: add_attr_with_inline 0.00% : 0.000004s : 1: add_comm_op_reuse_tag 0.00% : 0.000131s : 1: add_recomputation 0.00% : 0.000004s : 1: assign_add_opt 0.01% : 0.000417s : 1: auto_monad 0.00% : 0.000055s : 1: auto_monad_reorder 0.00% : 0.000004s : 1: begin_end_overlap_inline 0.00% : 0.000007s : 1: bias_add_comm_swap 0.02% : 0.000939s : 1: bootstrap 0.00% : 0.000048s : 1: cconv 0.00% : 0.000005s : 1: comm_op_add_attrs 0.00% : 0.000032s : 1: control_data_broadcast_order 0.00% : 0.000023s : 1: convert_after_rewriter 0.00% : 0.000064s : 1: cse_after_recomputation 0.00% : 0.000005s : 1: dataset_repeat_opt 0.00% : 0.000006s : 1: detach_backward 0.00% : 0.000026s : 1: environ_conv 0.00% : 0.000029s : 1: event_method 0.00% : 0.000005s : 1: full_micro_interleaved_order_control 0.00% : 0.000007s : 1: get_jit_bprop_graph 0.00% : 0.000011s : 1: graph_reusing 0.00% : 0.000005s : 1: grouped_pairwise_exchange_alltoall 0.00% : 0.000004s : 1: handle_group_info 0.00% : 0.000007s : 1: inline 0.00% : 0.000008s : 1: insert-virtual-dataset 0.00% : 0.000005s : 1: interleave_parallel_branches 0.00% : 0.000004s : 1: interleave_split_concat_branches 0.00% : 0.000006s : 1: label_fine_grained_interleaved_index 0.00% : 0.000010s : 1: label_micro_interleaved_index 0.02% : 0.000847s : 1: loop_unroll 0.00% : 0.000005s : 1: merge_cast_opt 0.00% : 0.000005s : 1: micro_interleaved_order_control 0.02% : 0.000841s : 1: mutable_eliminate 0.00% : 0.000010s : 1: offloading_packed_experts 0.00% : 0.000036s : 1: opt.transform.loop_unroll_optimizer 0.00% : 0.000036s : 1: opt.transform.mutable_eliminate 0.06% : 0.002736s : 78: opt.transform.opt_a 0.00% : 0.000108s : 1: opt.transform.opt_after_cconv 0.00% : 0.000062s : 1: opt.transform.opt_after_jit_grad 5.11% : 0.230684s : 28: opt.transform.opt_b 0.00% : 0.000134s : 2: opt.transform.opt_trans_graph 0.00% : 0.000099s : 4: opt.transform.symbol_engine_opt 2.57% : 0.116112s : 1: opt_a 0.01% : 0.000272s : 1: opt_after_cconv 0.02% : 0.000747s : 1: opt_after_jit_grad 5.12% : 0.230999s : 1: opt_b 7.77% : 0.350700s : 1: optimize 0.00% : 0.000046s : 1: optimize_parallel_all_gather_comm 0.00% : 0.000014s : 1: order_py_execute_after_rewriter 0.00% : 0.000045s : 1: overlap_grad_flash_sp 0.00% : 0.000004s : 1: overlap_grad_matmul_and_grad_allreduce 0.00% : 0.000013s : 1: overlap_grad_ring_attention 0.00% : 0.000005s : 1: overlap_opt_shard_grad_in_pipeline 0.00% : 0.000007s : 1: overlap_opt_shard_in_pipeline 0.00% : 0.000008s : 1: overlap_param_gather 0.00% : 0.000005s : 1: overlap_recompute_allgather_and_fa_grad 0.00% : 0.000011s : 1: overlap_recompute_and_grad_model_parallel 0.00% : 0.000005s : 1: overlap_recompute_comm 0.00% : 0.000009s : 1: parallel-infer-symbol 0.00% : 0.000004s : 1: parallel-infer-symbol-second 0.00% : 0.000006s : 1: partial_unused_args_eliminate 0.00% : 0.000005s : 1: pipeline_parallel_scheduler 0.00% : 0.000006s : 1: pipeline_split 0.00% : 0.000043s : 1: pre_auto_parallel 0.00% : 0.000041s : 1: py_interpret_to_execute 0.00% : 0.000038s : 1: py_interpret_to_execute_after_opt_a 0.00% : 0.000005s : 1: remove_cast_before_assign_add 0.00% : 0.000086s : 1: remove_dup_value 2.37% : 0.106872s : 1: renormalize.infer 0.08% : 0.003743s : 1: renormalize.specialize 0.00% : 0.000006s : 1: reorder_send_recv_between_fp_bp 0.00% : 0.000010s : 1: rewriter_after_jit_bprop_graph 0.01% : 0.000317s : 1: rewriter_after_opt_a 0.00% : 0.000094s : 1: rewriter_before_opt_a 0.00% : 0.000006s : 1: slice_cell_reuse_recomputed_activation 0.00% : 0.000006s : 1: slice_recompute_activation 0.00% : 0.000006s : 1: split_layernorm_comm 0.00% : 0.000005s : 1: split_matmul_comm_elemetwise 0.00% : 0.000015s : 1: swap_dp_allreduce_reducescatter 0.00% : 0.000165s : 1: symbol_engine_optimizer 0.00% : 0.000173s : 1: tuple_transform 76.47% : 3.452445s : 1: type_inference . [hook] pytest_runtest_teardown:test_matmul_qkv_bf16_1[False-mstype0-k_n_shape1-256] tests/st/infer/ops/test_internal_ops/test_matmul_split.py::test_matmul_qkv_bf16_1[False-mstype0-k_n_shape1-256],max_mem:146.0M [WARNING] ME(168319:281472866918192,MainProcess):2026-01-29-17:40:24.578.542 [mindspore/context.py:1334] For 'context.set_context', the parameter 'device_target' will be deprecated and removed in a future version. Please use the api mindspore.set_device() instead. TotalTime = 3.51943, [21] [bootstrap]: 0.00079985 [type_inference]: 3.11384 [event_method]: 2.037e-05 [auto_monad]: 0.00036314 [graph_reusing]: 6.97002e-06 [inline]: 3.37997e-06 [add_attr]: 0.234851, [1] [add_attr_with_inline]: 0.234832, [1] [Cycle 1]: 7.082e-05, [2] [tag_attr]: 2.436e-05 [meta_addattr_fg_expand]: 4.2e-06 [parallel-infer-symbol]: 6.46999e-06 [pre_auto_parallel]: 3.617e-05 [insert-virtual-dataset]: 3.41001e-06 [parallel-infer-symbol-second]: 1.39e-06 [dataset_repeat_opt]: 1.84e-06 [pipeline_split]: 1.74e-06 [optimize]: 0.168192, [53] [py_interpret_to_execute]: 3.4e-05 [rewriter_before_opt_a]: 8.153e-05 [opt_a]: 0.0165229, [2] [Cycle 1]: 0.0146241, [45] [expand_dump_flag]: 2.96999e-06 [switch_simplify]: 3.721e-05 [loop_unroll]: 1.996e-05 [a_1]: 0.00085636 [with_stream_mark]: 2.42e-05 [recompute_prepare]: 2.169e-05 [updatestate_depend_eliminate]: 1.128e-05 [updatestate_assign_eliminate]: 8.1e-06 [updatestate_loads_eliminate]: 1.134e-05 [parameter_eliminate]: 2.24001e-06 [a_2]: 0.00025166 [accelerated_algorithm]: 4.989e-05 [shard]: 2.23002e-06 [meta_shard_fg_expand]: 3.53999e-06 [shard_inline]: 1.695e-05 [merge_send_recv]: 1.444e-05 [auto_parallel]: 1.301e-05 [parallel]: 3.35e-05 [flash_sp]: 1.301e-05 [merge_comm]: 9.51e-06 [allreduce_fusion]: 8.23999e-06 [matmul_add_comm_reduction]: 1.716e-05 [allreduce_slice_to_reducescatter]: 6.59988e-07 [virtual_shard_identity]: 2.015e-05 [virtual_dataset]: 1.86e-05 [get_grad_eliminate_]: 2.003e-05 [virtual_output]: 1.811e-05 [merge_forward]: 8.69e-06 [cell_reuse_recompute_pass]: 2.58e-06 [offload_activation]: 1.744e-05 [cell_reuse_handle_not_recompute_node_pass]: 3.505e-05 [merge_recompute_call_nodes]: 1.67001e-06 [before_grad]: 2.91e-05 [set_forward_comm_id_for_comm_node_pass]: 8.87999e-06 [meta_fg_expand]: 6.48998e-06 [flash_sp_send_recv_attached]: 6.76e-06 [receive_attached]: 3.08e-06 [after_resolve]: 2.488e-05 [a_after_grad]: 2.994e-05 [renormalize]: 0.0116842 [add_forward_monad_depend]: 2.264e-05 [auto_monad_grad]: 3.65e-06 [auto_monad_eliminator]: 6.359e-05 [cse]: 0.00068475 [a_3]: 0.00014538 [Cycle 2]: 0.00188095, [45] [expand_dump_flag]: 3.42002e-06 [switch_simplify]: 5.605e-05 [loop_unroll]: 1.86e-05 [a_1]: 0.00049984 [with_stream_mark]: 3.746e-05 [recompute_prepare]: 2.398e-05 [updatestate_depend_eliminate]: 1.084e-05 [updatestate_assign_eliminate]: 8.99e-06 [updatestate_loads_eliminate]: 1.067e-05 [parameter_eliminate]: 3.21001e-06 [a_2]: 0.00023872 [accelerated_algorithm]: 2.654e-05 [shard]: 2.36e-06 [meta_shard_fg_expand]: 5.35999e-06 [shard_inline]: 1.744e-05 [merge_send_recv]: 1.816e-05 [auto_parallel]: 1.708e-05 [parallel]: 1.116e-05 [flash_sp]: 5.20999e-06 [merge_comm]: 1.102e-05 [allreduce_fusion]: 9.36e-06 [matmul_add_comm_reduction]: 1.884e-05 [allreduce_slice_to_reducescatter]: 1.21997e-06 [virtual_shard_identity]: 2.554e-05 [virtual_dataset]: 1.702e-05 [get_grad_eliminate_]: 1.592e-05 [virtual_output]: 1.597e-05 [merge_forward]: 1.049e-05 [cell_reuse_recompute_pass]: 3.61999e-06 [offload_activation]: 1.912e-05 [cell_reuse_handle_not_recompute_node_pass]: 3.674e-05 [merge_recompute_call_nodes]: 1.76e-06 [before_grad]: 2.858e-05 [set_forward_comm_id_for_comm_node_pass]: 1.095e-05 [meta_fg_expand]: 6.84001e-06 [flash_sp_send_recv_attached]: 2.49999e-06 [receive_attached]: 3.19001e-06 [after_resolve]: 2.445e-05 [a_after_grad]: 2.655e-05 [renormalize]: 7.00238e-08 [add_forward_monad_depend]: 4.88001e-06 [auto_monad_grad]: 3.63999e-06 [auto_monad_eliminator]: 6.66e-05 [cse]: 7.289e-05 [a_3]: 0.00010907 [py_interpret_to_execute_after_opt_a]: 3.234e-05 [slice_cell_reuse_recomputed_activation]: 2.77002e-06 [rewriter_after_opt_a]: 0.147573 [convert_after_rewriter]: 3.446e-05 [order_py_execute_after_rewriter]: 1.306e-05 [mutable_eliminate]: 0.00090091 [opt_b]: 0.00061541, [1] [Cycle 1]: 0.00060374, [7] [b_1]: 0.00038924 [b_2]: 1.908e-05 [updatestate_depend_eliminate]: 1.919e-05 [updatestate_assign_eliminate]: 8.11002e-06 [updatestate_loads_eliminate]: 1.159e-05 [renormalize]: 9.79984e-07 [cse]: 0.00010184 [optimize_parallel_all_gather_comm]: 3.69e-05 [overlap_param_gather]: 1.96e-06 [cconv]: 4.829e-05 [loop_unroll]: 0.0007025 [opt_after_cconv]: 0.00038239, [1] [Cycle 1]: 0.00037022, [7] [c_1]: 0.00021022 [parameter_eliminate]: 7.25003e-06 [updatestate_depend_eliminate]: 1.946e-05 [updatestate_assign_eliminate]: 8.05e-06 [updatestate_loads_eliminate]: 1.139e-05 [cse]: 6.833e-05 [renormalize]: 8.2e-07 [remove_dup_value]: 8.377e-05 [tuple_transform]: 0.00018864, [1] [Cycle 1]: 0.00018109, [4] [d_1]: 0.00013364 [none_parameter_eliminate]: 2.83e-06 [renormalize]: 2.00002e-07 [switch_simplify]: 1.832e-05 [partial_unused_args_eliminate]: 2.20002e-06 [add_recomputation]: 0.00013133 [cse_after_recomputation]: 6.837e-05, [1] [Cycle 1]: 6.191e-05, [1] [cse]: 5.193e-05 [environ_conv]: 2.413e-05 [swap_dp_allreduce_reducescatter]: 1.328e-05 [bias_add_comm_swap]: 4.05e-06 [label_micro_interleaved_index]: 8.03999e-06 [label_fine_grained_interleaved_index]: 3.26001e-06 [merge_cast_opt]: 1.72999e-06 [slice_recompute_activation]: 2.43998e-06 [micro_interleaved_order_control]: 3.35998e-06 [assign_add_opt]: 1.97001e-06 [ForceFp32Comm]: 1.29998e-06 [remove_cast_before_assign_add]: 1.14e-06 [full_micro_interleaved_order_control]: 2.30002e-06 [reorder_send_recv_between_fp_bp]: 2.93e-06 [comm_op_add_attrs]: 1.27999e-06 [add_comm_op_reuse_tag]: 1.50001e-06 [interleave_split_concat_branches]: 1.54e-06 [interleave_parallel_branches]: 1.14003e-06 [overlap_opt_shard_in_pipeline]: 1.23002e-06 [overlap_opt_shard_grad_in_pipeline]: 1.82999e-06 [control_data_broadcast_order]: 3.211e-05 [grouped_pairwise_exchange_alltoall]: 1.79e-06 [offloading_packed_experts]: 8e-06 [overlap_recompute_and_grad_model_parallel]: 7.77e-06 [overlap_grad_matmul_and_grad_allreduce]: 1.25999e-06 [overlap_recompute_allgather_and_fa_grad]: 1.54e-06 [overlap_recompute_comm]: 2.34999e-06 [overlap_grad_ring_attention]: 8.74e-06 [overlap_grad_flash_sp]: 4.073e-05 [begin_end_overlap_inline]: 5.59987e-07 [split_matmul_comm_elemetwise]: 2.81999e-06 [split_layernorm_comm]: 2.66999e-06 [handle_group_info]: 1.40999e-06 [symbol_engine_optimizer]: 0.00018761, [1] [Cycle 1]: 0.0001805, [6] [build]: 2.366e-05 [elim_shapecalc]: 2.995e-05 [elim_not_effective]: 3.391e-05 [opt_reshape]: 2.511e-05 [fold_const_symbol]: 2.579e-05 [renormalize]: 5.19998e-07 [detach_backward]: 2.88003e-06 [pipeline_parallel_scheduler]: 1.91998e-06 [auto_monad_reorder]: 5.63e-05 [get_jit_bprop_graph]: 1.88002e-06 [rewriter_after_jit_bprop_graph]: 7.63001e-06 [opt_after_jit_grad]: 0.00083172 [validate]: 0.00012167 Sums bootstrap : 0.000800s : 0.02% type_inference : 3.113842s : 94.85% event_method : 0.000020s : 0.00% auto_monad : 0.000363s : 0.01% graph_reusing : 0.000007s : 0.00% inline : 0.000003s : 0.00% add_attr.add_attr_with_inline.tag_attr : 0.000024s : 0.00% add_attr.add_attr_with_inline.meta_addattr_fg_expand : 0.000004s : 0.00% parallel-infer-symbol : 0.000006s : 0.00% pre_auto_parallel : 0.000036s : 0.00% insert-virtual-dataset : 0.000003s : 0.00% parallel-infer-symbol-second : 0.000001s : 0.00% dataset_repeat_opt : 0.000002s : 0.00% pipeline_split : 0.000002s : 0.00% optimize.py_interpret_to_execute : 0.000034s : 0.00% optimize.rewriter_before_opt_a : 0.000082s : 0.00% optimize.opt_a.expand_dump_flag : 0.000006s : 0.00% optimize.opt_a.switch_simplify : 0.000093s : 0.00% optimize.opt_a.loop_unroll : 0.000039s : 0.00% optimize.opt_a.a_1 : 0.001356s : 0.04% optimize.opt_a.with_stream_mark : 0.000062s : 0.00% optimize.opt_a.recompute_prepare : 0.000046s : 0.00% optimize.opt_a.updatestate_depend_eliminate : 0.000022s : 0.00% optimize.opt_a.updatestate_assign_eliminate : 0.000017s : 0.00% optimize.opt_a.updatestate_loads_eliminate : 0.000022s : 0.00% optimize.opt_a.parameter_eliminate : 0.000005s : 0.00% optimize.opt_a.a_2 : 0.000490s : 0.01% optimize.opt_a.accelerated_algorithm : 0.000076s : 0.00% optimize.opt_a.shard : 0.000005s : 0.00% optimize.opt_a.meta_shard_fg_expand : 0.000009s : 0.00% optimize.opt_a.shard_inline : 0.000034s : 0.00% optimize.opt_a.merge_send_recv : 0.000033s : 0.00% optimize.opt_a.auto_parallel : 0.000030s : 0.00% optimize.opt_a.parallel : 0.000045s : 0.00% optimize.opt_a.flash_sp : 0.000018s : 0.00% optimize.opt_a.merge_comm : 0.000021s : 0.00% optimize.opt_a.allreduce_fusion : 0.000018s : 0.00% optimize.opt_a.matmul_add_comm_reduction : 0.000036s : 0.00% optimize.opt_a.allreduce_slice_to_reducescatter : 0.000002s : 0.00% optimize.opt_a.virtual_shard_identity : 0.000046s : 0.00% optimize.opt_a.virtual_dataset : 0.000036s : 0.00% optimize.opt_a.get_grad_eliminate_ : 0.000036s : 0.00% optimize.opt_a.virtual_output : 0.000034s : 0.00% optimize.opt_a.merge_forward : 0.000019s : 0.00% optimize.opt_a.cell_reuse_recompute_pass : 0.000006s : 0.00% optimize.opt_a.offload_activation : 0.000037s : 0.00% optimize.opt_a.cell_reuse_handle_not_recompute_node_pass : 0.000072s : 0.00% optimize.opt_a.merge_recompute_call_nodes : 0.000003s : 0.00% optimize.opt_a.before_grad : 0.000058s : 0.00% optimize.opt_a.set_forward_comm_id_for_comm_node_pass : 0.000020s : 0.00% optimize.opt_a.meta_fg_expand : 0.000013s : 0.00% optimize.opt_a.flash_sp_send_recv_attached : 0.000009s : 0.00% optimize.opt_a.receive_attached : 0.000006s : 0.00% optimize.opt_a.after_resolve : 0.000049s : 0.00% optimize.opt_a.a_after_grad : 0.000056s : 0.00% optimize.opt_a.renormalize : 0.011684s : 0.36% optimize.opt_a.add_forward_monad_depend : 0.000028s : 0.00% optimize.opt_a.auto_monad_grad : 0.000007s : 0.00% optimize.opt_a.auto_monad_eliminator : 0.000130s : 0.00% optimize.opt_a.cse : 0.000758s : 0.02% optimize.opt_a.a_3 : 0.000254s : 0.01% optimize.py_interpret_to_execute_after_opt_a : 0.000032s : 0.00% optimize.slice_cell_reuse_recomputed_activation : 0.000003s : 0.00% optimize.rewriter_after_opt_a : 0.147573s : 4.49% optimize.convert_after_rewriter : 0.000034s : 0.00% optimize.order_py_execute_after_rewriter : 0.000013s : 0.00% optimize.mutable_eliminate : 0.000901s : 0.03% optimize.opt_b.b_1 : 0.000389s : 0.01% optimize.opt_b.b_2 : 0.000019s : 0.00% optimize.opt_b.updatestate_depend_eliminate : 0.000019s : 0.00% optimize.opt_b.updatestate_assign_eliminate : 0.000008s : 0.00% optimize.opt_b.updatestate_loads_eliminate : 0.000012s : 0.00% optimize.opt_b.renormalize : 0.000001s : 0.00% optimize.opt_b.cse : 0.000102s : 0.00% optimize.optimize_parallel_all_gather_comm : 0.000037s : 0.00% optimize.overlap_param_gather : 0.000002s : 0.00% optimize.cconv : 0.000048s : 0.00% optimize.loop_unroll : 0.000702s : 0.02% optimize.opt_after_cconv.c_1 : 0.000210s : 0.01% optimize.opt_after_cconv.parameter_eliminate : 0.000007s : 0.00% optimize.opt_after_cconv.updatestate_depend_eliminate : 0.000019s : 0.00% optimize.opt_after_cconv.updatestate_assign_eliminate : 0.000008s : 0.00% optimize.opt_after_cconv.updatestate_loads_eliminate : 0.000011s : 0.00% optimize.opt_after_cconv.cse : 0.000068s : 0.00% optimize.opt_after_cconv.renormalize : 0.000001s : 0.00% optimize.remove_dup_value : 0.000084s : 0.00% optimize.tuple_transform.d_1 : 0.000134s : 0.00% optimize.tuple_transform.none_parameter_eliminate : 0.000003s : 0.00% optimize.tuple_transform.renormalize : 0.000000s : 0.00% optimize.tuple_transform.switch_simplify : 0.000018s : 0.00% optimize.partial_unused_args_eliminate : 0.000002s : 0.00% optimize.add_recomputation : 0.000131s : 0.00% optimize.cse_after_recomputation.cse : 0.000052s : 0.00% optimize.environ_conv : 0.000024s : 0.00% optimize.swap_dp_allreduce_reducescatter : 0.000013s : 0.00% optimize.bias_add_comm_swap : 0.000004s : 0.00% optimize.label_micro_interleaved_index : 0.000008s : 0.00% optimize.label_fine_grained_interleaved_index : 0.000003s : 0.00% optimize.merge_cast_opt : 0.000002s : 0.00% optimize.slice_recompute_activation : 0.000002s : 0.00% optimize.micro_interleaved_order_control : 0.000003s : 0.00% optimize.assign_add_opt : 0.000002s : 0.00% optimize.ForceFp32Comm : 0.000001s : 0.00% optimize.remove_cast_before_assign_add : 0.000001s : 0.00% optimize.full_micro_interleaved_order_control : 0.000002s : 0.00% optimize.reorder_send_recv_between_fp_bp : 0.000003s : 0.00% optimize.comm_op_add_attrs : 0.000001s : 0.00% optimize.add_comm_op_reuse_tag : 0.000002s : 0.00% optimize.interleave_split_concat_branches : 0.000002s : 0.00% optimize.interleave_parallel_branches : 0.000001s : 0.00% optimize.overlap_opt_shard_in_pipeline : 0.000001s : 0.00% optimize.overlap_opt_shard_grad_in_pipeline : 0.000002s : 0.00% optimize.control_data_broadcast_order : 0.000032s : 0.00% optimize.grouped_pairwise_exchange_alltoall : 0.000002s : 0.00% optimize.offloading_packed_experts : 0.000008s : 0.00% optimize.overlap_recompute_and_grad_model_parallel : 0.000008s : 0.00% optimize.overlap_grad_matmul_and_grad_allreduce : 0.000001s : 0.00% optimize.overlap_recompute_allgather_and_fa_grad : 0.000002s : 0.00% optimize.overlap_recompute_comm : 0.000002s : 0.00% optimize.overlap_grad_ring_attention : 0.000009s : 0.00% optimize.overlap_grad_flash_sp : 0.000041s : 0.00% optimize.begin_end_overlap_inline : 0.000001s : 0.00% optimize.split_matmul_comm_elemetwise : 0.000003s : 0.00% optimize.split_layernorm_comm : 0.000003s : 0.00% optimize.handle_group_info : 0.000001s : 0.00% optimize.symbol_engine_optimizer.build : 0.000024s : 0.00% optimize.symbol_engine_optimizer.elim_shapecalc : 0.000030s : 0.00% optimize.symbol_engine_optimizer.elim_not_effective : 0.000034s : 0.00% optimize.symbol_engine_optimizer.opt_reshape : 0.000025s : 0.00% optimize.symbol_engine_optimizer.fold_const_symbol : 0.000026s : 0.00% optimize.symbol_engine_optimizer.renormalize : 0.000001s : 0.00% detach_backward : 0.000003s : 0.00% pipeline_parallel_scheduler : 0.000002s : 0.00% auto_monad_reorder : 0.000056s : 0.00% get_jit_bprop_graph : 0.000002s : 0.00% rewriter_after_jit_bprop_graph : 0.000008s : 0.00% opt_after_jit_grad : 0.000832s : 0.03% validate : 0.000122s : 0.00% Time group info: ------[substitution.] 0.000298 96 4.08% : 0.000012s : 2: substitution.depend_value_elim 1.40% : 0.000004s : 8: substitution.elim_not_effective 1.21% : 0.000004s : 8: substitution.fold_const_symbol 4.24% : 0.000013s : 13: substitution.graph_param_transform 41.13% : 0.000122s : 1: substitution.inline 3.04% : 0.000009s : 16: substitution.j_node_and_user_rematch 8.05% : 0.000024s : 2: substitution.less_batch_normalization 2.28% : 0.000007s : 6: substitution.load_eliminater 1.46% : 0.000004s : 2: substitution.opt_reshape 4.43% : 0.000013s : 16: substitution.remove_not_recompute_node 2.83% : 0.000008s : 4: substitution.replace_old_param 10.33% : 0.000031s : 4: substitution.reshape_eliminate 8.77% : 0.000026s : 6: substitution.updatestate_pure_node_eliminater 6.75% : 0.000020s : 8: substitution.updatestate_useless_node_eliminater ------[type_inference.] 3.113733 2 99.82% : 3.108221s : 1: type_inference.infer 0.18% : 0.005512s : 1: type_inference.specialize ------[replace.] 0.000019 1 100.00% : 0.000019s : 1: replace.inline ------[match.] 0.000121 1 100.00% : 0.000121s : 1: match.inline ------[predicate.] 0.000560 3002 0.67% : 0.000004s : 27: predicate.accumulaten_eliminater 1.07% : 0.000006s : 13: predicate.ad_related_special_op_eliminate 0.65% : 0.000004s : 26: predicate.addn_check_dump 0.61% : 0.000003s : 27: predicate.addn_zero_filter 0.65% : 0.000004s : 27: predicate.adjust_all_reduce_mul_add 1.90% : 0.000011s : 53: predicate.arithmetic_simplify 0.68% : 0.000004s : 27: predicate.cast_eliminate 0.65% : 0.000004s : 26: predicate.check_bprop_eliminate 0.67% : 0.000004s : 26: predicate.compare_switch_simplify 0.22% : 0.000001s : 13: predicate.const_output_eliminate 0.75% : 0.000004s : 26: predicate.depend_value_elim 0.69% : 0.000004s : 27: predicate.dict_get_item_const_eliminator 0.73% : 0.000004s : 27: predicate.dict_get_item_eliminator 0.66% : 0.000004s : 27: predicate.dict_set_item_eliminator 1.00% : 0.000006s : 26: predicate.dumpgradient_eliminate 0.25% : 0.000001s : 13: predicate.elim_not_effective 0.45% : 0.000003s : 13: predicate.elim_shapecalc_of_broadcastargs 0.96% : 0.000005s : 40: predicate.environ_add_const_eliminate 0.94% : 0.000005s : 40: predicate.environ_get_add_eliminate 1.15% : 0.000006s : 40: predicate.environ_get_depend_swap 1.56% : 0.000009s : 66: predicate.environ_get_eliminate 0.87% : 0.000005s : 40: predicate.environ_get_set_eliminate 0.60% : 0.000003s : 28: predicate.exchange_switch_depend_value 1.09% : 0.000006s : 28: predicate.float_depend_g_call 0.68% : 0.000004s : 26: predicate.float_environ_get_switch 0.92% : 0.000005s : 39: predicate.float_tuple_getitem_switch 0.18% : 0.000001s : 13: predicate.fold_const_symbol 0.81% : 0.000005s : 26: predicate.get_grad_eliminate 0.29% : 0.000002s : 13: predicate.graph_param_transform 0.61% : 0.000003s : 26: predicate.incorporate_call 0.55% : 0.000003s : 26: predicate.incorporate_call_switch 4.57% : 0.000026s : 133: predicate.inline 0.85% : 0.000005s : 26: predicate.inline_without_move 0.34% : 0.000002s : 26: predicate.j_node_and_user_rematch 1.12% : 0.000006s : 27: predicate.less_batch_normalization 1.40% : 0.000008s : 53: predicate.list_to_tuple_eliminator_ 18.70% : 0.000105s : 80: predicate.load_eliminater 1.13% : 0.000006s : 13: predicate.loop_unroll_after_grad 0.81% : 0.000005s : 31: predicate.loop_unroll_before_grad 1.46% : 0.000008s : 53: predicate.make_slice_get_slice_eliminator 0.64% : 0.000004s : 26: predicate.merge_addn 0.60% : 0.000003s : 26: predicate.micro_step_allgather_replace 0.60% : 0.000003s : 26: predicate.mini_step_allgather_replace 0.59% : 0.000003s : 27: predicate.minmaximum_grad 1.08% : 0.000006s : 13: predicate.mutable_eliminate 0.43% : 0.000002s : 13: predicate.opt_reshape 0.38% : 0.000002s : 13: predicate.parallel_virtual_node 0.88% : 0.000005s : 28: predicate.partial_defer_inline 1.00% : 0.000006s : 40: predicate.partial_eliminate 0.68% : 0.000004s : 27: predicate.print_const_string_wrapper 0.63% : 0.000004s : 26: predicate.reduce_all_const_elim 0.87% : 0.000005s : 27: predicate.reduce_eliminate 1.88% : 0.000011s : 80: predicate.redundant_stop_gradient_eliminater 0.48% : 0.000003s : 26: predicate.remove_not_recompute_node 1.22% : 0.000007s : 53: predicate.replace_applicator 0.49% : 0.000003s : 26: predicate.replace_old_param 0.40% : 0.000002s : 13: predicate.reset_defer_inline 0.81% : 0.000005s : 27: predicate.reshape_eliminate 0.72% : 0.000004s : 26: predicate.row_tensor_add_zeros_like 0.36% : 0.000002s : 13: predicate.row_tensor_eliminate 1.06% : 0.000006s : 26: predicate.same_eliminate 0.57% : 0.000003s : 26: predicate.set_cell_output_no_recompute 0.80% : 0.000004s : 26: predicate.shard_identity_eliminate 0.77% : 0.000004s : 26: predicate.special_op_eliminate 0.69% : 0.000004s : 26: predicate.specialize_transform 1.06% : 0.000006s : 26: predicate.split_environ_get_set_with_tuple_value 0.79% : 0.000004s : 26: predicate.stack_unstack_eliminate 0.40% : 0.000002s : 13: predicate.switch_call_monad_eliminater 0.69% : 0.000004s : 28: predicate.switch_defer_inline 1.35% : 0.000008s : 54: predicate.switch_layer_defer_inline 2.83% : 0.000016s : 98: predicate.switch_simplify 0.72% : 0.000004s : 27: predicate.tile_eliminate 0.63% : 0.000004s : 27: predicate.transpose_eliminate 1.37% : 0.000008s : 53: predicate.tuple_list_convert_item_index_to_positive 1.52% : 0.000009s : 53: predicate.tuple_list_get_item_const_eliminator 1.44% : 0.000008s : 53: predicate.tuple_list_get_item_depend_reorder 2.52% : 0.000014s : 79: predicate.tuple_list_get_item_eliminator 1.41% : 0.000008s : 53: predicate.tuple_list_get_set_item_eliminator 2.19% : 0.000012s : 79: predicate.tuple_list_set_item_eliminator 1.32% : 0.000007s : 53: predicate.tuple_to_list_eliminator_ 1.90% : 0.000011s : 80: predicate.updatestate_pure_node_eliminater 2.77% : 0.000016s : 106: predicate.updatestate_useless_node_eliminater 0.39% : 0.000002s : 13: predicate.value_based_eliminate 0.74% : 0.000004s : 26: predicate.virtual_dataset_eliminate 0.66% : 0.000004s : 26: predicate.virtual_output_eliminate 0.32% : 0.000002s : 13: predicate.virtual_view_grad_eliminate 0.45% : 0.000003s : 13: predicate.zero_like_fill_zero ------[func_graph_cloner_run.] 0.016178 58 85.28% : 0.013797s : 55: func_graph_cloner_run.FuncGraphClonerGraph 14.72% : 0.002381s : 3: func_graph_cloner_run.FuncGraphSpecializer ------[meta_graph.] 0.000000 0 ------[manager.] 0.000000 0 ------[pynative] 0.000000 0 ------[others.] 3.937491 192 0.00% : 0.000004s : 1: ForceFp32Comm 5.96% : 0.234858s : 1: add_attr 5.96% : 0.234837s : 1: add_attr_with_inline 0.00% : 0.000004s : 1: add_comm_op_reuse_tag 0.00% : 0.000141s : 1: add_recomputation 0.00% : 0.000005s : 1: assign_add_opt 0.01% : 0.000375s : 1: auto_monad 0.00% : 0.000062s : 1: auto_monad_reorder 0.00% : 0.000004s : 1: begin_end_overlap_inline 0.00% : 0.000007s : 1: bias_add_comm_swap 0.02% : 0.000882s : 1: bootstrap 0.00% : 0.000055s : 1: cconv 0.00% : 0.000006s : 1: comm_op_add_attrs 0.00% : 0.000037s : 1: control_data_broadcast_order 0.00% : 0.000043s : 1: convert_after_rewriter 0.00% : 0.000072s : 1: cse_after_recomputation 0.00% : 0.000005s : 1: dataset_repeat_opt 0.00% : 0.000010s : 1: detach_backward 0.00% : 0.000030s : 1: environ_conv 0.00% : 0.000029s : 1: event_method 0.00% : 0.000005s : 1: full_micro_interleaved_order_control 0.00% : 0.000007s : 1: get_jit_bprop_graph 0.00% : 0.000011s : 1: graph_reusing 0.00% : 0.000005s : 1: grouped_pairwise_exchange_alltoall 0.00% : 0.000004s : 1: handle_group_info 0.00% : 0.000007s : 1: inline 0.00% : 0.000008s : 1: insert-virtual-dataset 0.00% : 0.000005s : 1: interleave_parallel_branches 0.00% : 0.000004s : 1: interleave_split_concat_branches 0.00% : 0.000006s : 1: label_fine_grained_interleaved_index 0.00% : 0.000011s : 1: label_micro_interleaved_index 0.02% : 0.000721s : 1: loop_unroll 0.00% : 0.000005s : 1: merge_cast_opt 0.00% : 0.000007s : 1: micro_interleaved_order_control 0.02% : 0.000919s : 1: mutable_eliminate 0.00% : 0.000012s : 1: offloading_packed_experts 0.00% : 0.000039s : 1: opt.transform.loop_unroll_optimizer 0.00% : 0.000047s : 1: opt.transform.mutable_eliminate 0.07% : 0.002638s : 78: opt.transform.opt_a 0.01% : 0.000208s : 1: opt.transform.opt_after_cconv 0.00% : 0.000074s : 1: opt.transform.opt_after_jit_grad 0.01% : 0.000365s : 28: opt.transform.opt_b 0.00% : 0.000148s : 2: opt.transform.opt_trans_graph 0.00% : 0.000108s : 4: opt.transform.symbol_engine_opt 0.42% : 0.016527s : 1: opt_a 0.01% : 0.000387s : 1: opt_after_cconv 0.02% : 0.000852s : 1: opt_after_jit_grad 0.02% : 0.000620s : 1: opt_b 4.27% : 0.168200s : 1: optimize 0.00% : 0.000043s : 1: optimize_parallel_all_gather_comm 0.00% : 0.000018s : 1: order_py_execute_after_rewriter 0.00% : 0.000046s : 1: overlap_grad_flash_sp 0.00% : 0.000004s : 1: overlap_grad_matmul_and_grad_allreduce 0.00% : 0.000014s : 1: overlap_grad_ring_attention 0.00% : 0.000004s : 1: overlap_opt_shard_grad_in_pipeline 0.00% : 0.000004s : 1: overlap_opt_shard_in_pipeline 0.00% : 0.000006s : 1: overlap_param_gather 0.00% : 0.000006s : 1: overlap_recompute_allgather_and_fa_grad 0.00% : 0.000011s : 1: overlap_recompute_and_grad_model_parallel 0.00% : 0.000005s : 1: overlap_recompute_comm 0.00% : 0.000010s : 1: parallel-infer-symbol 0.00% : 0.000004s : 1: parallel-infer-symbol-second 0.00% : 0.000006s : 1: partial_unused_args_eliminate 0.00% : 0.000008s : 1: pipeline_parallel_scheduler 0.00% : 0.000006s : 1: pipeline_split 0.00% : 0.000041s : 1: pre_auto_parallel 0.00% : 0.000038s : 1: py_interpret_to_execute 0.00% : 0.000036s : 1: py_interpret_to_execute_after_opt_a 0.00% : 0.000004s : 1: remove_cast_before_assign_add 0.00% : 0.000090s : 1: remove_dup_value 0.23% : 0.008871s : 1: renormalize.infer 0.07% : 0.002797s : 1: renormalize.specialize 0.00% : 0.000006s : 1: reorder_send_recv_between_fp_bp 0.00% : 0.000011s : 1: rewriter_after_jit_bprop_graph 3.75% : 0.147601s : 1: rewriter_after_opt_a 0.00% : 0.000086s : 1: rewriter_before_opt_a 0.00% : 0.000006s : 1: slice_cell_reuse_recomputed_activation 0.00% : 0.000005s : 1: slice_recompute_activation 0.00% : 0.000007s : 1: split_layernorm_comm 0.00% : 0.000006s : 1: split_matmul_comm_elemetwise 0.00% : 0.000017s : 1: swap_dp_allreduce_reducescatter 0.00% : 0.000191s : 1: symbol_engine_optimizer 0.00% : 0.000192s : 1: tuple_transform 79.08% : 3.113873s : 1: type_inference . [hook] pytest_runtest_teardown:test_matmul_qkv_bf16_1[False-mstype0-k_n_shape1-1024] tests/st/infer/ops/test_internal_ops/test_matmul_split.py::test_matmul_qkv_bf16_1[False-mstype0-k_n_shape1-1024],max_mem:170.0M [WARNING] ME(168319:281472866918192,MainProcess):2026-01-29-17:41:32.882.039 [mindspore/context.py:1334] For 'context.set_context', the parameter 'device_target' will be deprecated and removed in a future version. Please use the api mindspore.set_device() instead. TotalTime = 2.37966, [21] [bootstrap]: 0.00081557 [type_inference]: 2.14334 [event_method]: 1.999e-05 [auto_monad]: 0.00039057 [graph_reusing]: 7.03998e-06 [inline]: 3.19001e-06 [add_attr]: 0.00700176, [1] [add_attr_with_inline]: 0.0069841, [1] [Cycle 1]: 7.436e-05, [2] [tag_attr]: 2.328e-05 [meta_addattr_fg_expand]: 4.58999e-06 [parallel-infer-symbol]: 5.79999e-06 [pre_auto_parallel]: 3.755e-05 [insert-virtual-dataset]: 2.98998e-06 [parallel-infer-symbol-second]: 9.89996e-07 [dataset_repeat_opt]: 2.58e-06 [pipeline_split]: 1.92001e-06 [optimize]: 0.226853, [53] [py_interpret_to_execute]: 3.285e-05 [rewriter_before_opt_a]: 8.348e-05 [opt_a]: 0.142263, [2] [Cycle 1]: 0.140358, [45] [expand_dump_flag]: 2.88e-06 [switch_simplify]: 3.554e-05 [loop_unroll]: 2.049e-05 [a_1]: 0.125213 [with_stream_mark]: 5.029e-05 [recompute_prepare]: 3.236e-05 [updatestate_depend_eliminate]: 1.426e-05 [updatestate_assign_eliminate]: 8.76002e-06 [updatestate_loads_eliminate]: 1.208e-05 [parameter_eliminate]: 3.15998e-06 [a_2]: 0.00027911 [accelerated_algorithm]: 5.928e-05 [shard]: 3.2e-06 [meta_shard_fg_expand]: 8.27e-06 [shard_inline]: 2.041e-05 [merge_send_recv]: 1.726e-05 [auto_parallel]: 1.697e-05 [parallel]: 3.831e-05 [flash_sp]: 1.796e-05 [merge_comm]: 1.077e-05 [allreduce_fusion]: 8.40001e-06 [matmul_add_comm_reduction]: 2.156e-05 [allreduce_slice_to_reducescatter]: 6.59988e-07 [virtual_shard_identity]: 2.731e-05 [virtual_dataset]: 2.05e-05 [get_grad_eliminate_]: 1.948e-05 [virtual_output]: 1.963e-05 [merge_forward]: 1.157e-05 [cell_reuse_recompute_pass]: 3.72998e-06 [offload_activation]: 2.191e-05 [cell_reuse_handle_not_recompute_node_pass]: 4.024e-05 [merge_recompute_call_nodes]: 1.84e-06 [before_grad]: 3.318e-05 [set_forward_comm_id_for_comm_node_pass]: 1.154e-05 [meta_fg_expand]: 6.66e-06 [flash_sp_send_recv_attached]: 7.84002e-06 [receive_attached]: 2.49001e-06 [after_resolve]: 3.136e-05 [a_after_grad]: 3.614e-05 [renormalize]: 0.012126 [add_forward_monad_depend]: 2.24e-05 [auto_monad_grad]: 3.31001e-06 [auto_monad_eliminator]: 6.982e-05 [cse]: 0.00133532 [a_3]: 0.00014833 [Cycle 2]: 0.00188601, [45] [expand_dump_flag]: 2.346e-05 [switch_simplify]: 1.976e-05 [loop_unroll]: 1.809e-05 [a_1]: 0.00048341 [with_stream_mark]: 4.129e-05 [recompute_prepare]: 2.237e-05 [updatestate_depend_eliminate]: 1.11e-05 [updatestate_assign_eliminate]: 8.04002e-06 [updatestate_loads_eliminate]: 1.125e-05 [parameter_eliminate]: 2.64999e-06 [a_2]: 0.00023261 [accelerated_algorithm]: 2.411e-05 [shard]: 3.36999e-06 [meta_shard_fg_expand]: 6.88998e-06 [shard_inline]: 1.687e-05 [merge_send_recv]: 1.802e-05 [auto_parallel]: 1.674e-05 [parallel]: 1.124e-05 [flash_sp]: 5.47001e-06 [merge_comm]: 9.52001e-06 [allreduce_fusion]: 9.89001e-06 [matmul_add_comm_reduction]: 2.04e-05 [allreduce_slice_to_reducescatter]: 1.12999e-06 [virtual_shard_identity]: 2.384e-05 [virtual_dataset]: 1.557e-05 [get_grad_eliminate_]: 1.552e-05 [virtual_output]: 1.58e-05 [merge_forward]: 1.082e-05 [cell_reuse_recompute_pass]: 4.08999e-06 [offload_activation]: 1.894e-05 [cell_reuse_handle_not_recompute_node_pass]: 3.448e-05 [merge_recompute_call_nodes]: 2.14e-06 [before_grad]: 2.679e-05 [set_forward_comm_id_for_comm_node_pass]: 1.064e-05 [meta_fg_expand]: 6.98e-06 [flash_sp_send_recv_attached]: 2.48e-06 [receive_attached]: 2.39001e-06 [after_resolve]: 2.416e-05 [a_after_grad]: 2.777e-05 [renormalize]: 1.19995e-07 [add_forward_monad_depend]: 4.67e-06 [auto_monad_grad]: 3.35e-06 [auto_monad_eliminator]: 6.245e-05 [cse]: 0.00014724 [a_3]: 0.00011043 [py_interpret_to_execute_after_opt_a]: 3.171e-05 [slice_cell_reuse_recomputed_activation]: 2.32999e-06 [rewriter_after_opt_a]: 0.00021238 [convert_after_rewriter]: 1.841e-05 [order_py_execute_after_rewriter]: 1.232e-05 [mutable_eliminate]: 0.00087312 [opt_b]: 0.0006067, [1] [Cycle 1]: 0.00059384, [7] [b_1]: 0.00038635 [b_2]: 2.149e-05 [updatestate_depend_eliminate]: 1.93e-05 [updatestate_assign_eliminate]: 8.05e-06 [updatestate_loads_eliminate]: 1.288e-05 [renormalize]: 8.09989e-07 [cse]: 8.853e-05 [optimize_parallel_all_gather_comm]: 3.74e-05 [overlap_param_gather]: 1.94e-06 [cconv]: 4.75e-05 [loop_unroll]: 0.0812315 [opt_after_cconv]: 0.00031399, [1] [Cycle 1]: 0.00030199, [7] [c_1]: 0.00012228 [parameter_eliminate]: 7.85e-06 [updatestate_depend_eliminate]: 2.268e-05 [updatestate_assign_eliminate]: 7.98001e-06 [updatestate_loads_eliminate]: 1.192e-05 [cse]: 8.589e-05 [renormalize]: 1.40999e-06 [remove_dup_value]: 7.417e-05 [tuple_transform]: 0.00016497, [1] [Cycle 1]: 0.00015913, [4] [d_1]: 0.00011833 [none_parameter_eliminate]: 1.86998e-06 [renormalize]: 2.30008e-07 [switch_simplify]: 1.694e-05 [partial_unused_args_eliminate]: 2.21e-06 [add_recomputation]: 0.00012168 [cse_after_recomputation]: 6.197e-05, [1] [Cycle 1]: 5.658e-05, [1] [cse]: 4.93e-05 [environ_conv]: 2.204e-05 [swap_dp_allreduce_reducescatter]: 1.36e-05 [bias_add_comm_swap]: 3.23998e-06 [label_micro_interleaved_index]: 7.7e-06 [label_fine_grained_interleaved_index]: 2.61999e-06 [merge_cast_opt]: 1.79998e-06 [slice_recompute_activation]: 2.62001e-06 [micro_interleaved_order_control]: 2.48e-06 [assign_add_opt]: 1.54e-06 [ForceFp32Comm]: 8.00006e-07 [remove_cast_before_assign_add]: 1.12e-06 [full_micro_interleaved_order_control]: 2.17999e-06 [reorder_send_recv_between_fp_bp]: 2.88e-06 [comm_op_add_attrs]: 1.24998e-06 [add_comm_op_reuse_tag]: 1.09998e-06 [interleave_split_concat_branches]: 1.74e-06 [interleave_parallel_branches]: 1.12999e-06 [overlap_opt_shard_in_pipeline]: 1.37999e-06 [overlap_opt_shard_grad_in_pipeline]: 1.72999e-06 [control_data_broadcast_order]: 2.917e-05 [grouped_pairwise_exchange_alltoall]: 1.76e-06 [offloading_packed_experts]: 7.19001e-06 [overlap_recompute_and_grad_model_parallel]: 8.23999e-06 [overlap_grad_matmul_and_grad_allreduce]: 1.27e-06 [overlap_recompute_allgather_and_fa_grad]: 1.45001e-06 [overlap_recompute_comm]: 2.46e-06 [overlap_grad_ring_attention]: 8.25999e-06 [overlap_grad_flash_sp]: 3.814e-05 [begin_end_overlap_inline]: 6.30011e-07 [split_matmul_comm_elemetwise]: 2.49999e-06 [split_layernorm_comm]: 2.22999e-06 [handle_group_info]: 1.14998e-06 [symbol_engine_optimizer]: 0.00016193, [1] [Cycle 1]: 0.00015553, [6] [build]: 1.599e-05 [elim_shapecalc]: 2.351e-05 [elim_not_effective]: 3.218e-05 [opt_reshape]: 2.377e-05 [fold_const_symbol]: 2.482e-05 [renormalize]: 2.69996e-07 [detach_backward]: 2.51e-06 [pipeline_parallel_scheduler]: 1.50001e-06 [auto_monad_reorder]: 4.726e-05 [get_jit_bprop_graph]: 1.99999e-06 [rewriter_after_jit_bprop_graph]: 7.28e-06 [opt_after_jit_grad]: 0.00074875 [validate]: 9.838e-05 Sums bootstrap : 0.000816s : 0.03% type_inference : 2.143339s : 90.39% event_method : 0.000020s : 0.00% auto_monad : 0.000391s : 0.02% graph_reusing : 0.000007s : 0.00% inline : 0.000003s : 0.00% add_attr.add_attr_with_inline.tag_attr : 0.000023s : 0.00% add_attr.add_attr_with_inline.meta_addattr_fg_expand : 0.000005s : 0.00% parallel-infer-symbol : 0.000006s : 0.00% pre_auto_parallel : 0.000038s : 0.00% insert-virtual-dataset : 0.000003s : 0.00% parallel-infer-symbol-second : 0.000001s : 0.00% dataset_repeat_opt : 0.000003s : 0.00% pipeline_split : 0.000002s : 0.00% optimize.py_interpret_to_execute : 0.000033s : 0.00% optimize.rewriter_before_opt_a : 0.000083s : 0.00% optimize.opt_a.expand_dump_flag : 0.000026s : 0.00% optimize.opt_a.switch_simplify : 0.000055s : 0.00% optimize.opt_a.loop_unroll : 0.000039s : 0.00% optimize.opt_a.a_1 : 0.125697s : 5.30% optimize.opt_a.with_stream_mark : 0.000092s : 0.00% optimize.opt_a.recompute_prepare : 0.000055s : 0.00% optimize.opt_a.updatestate_depend_eliminate : 0.000025s : 0.00% optimize.opt_a.updatestate_assign_eliminate : 0.000017s : 0.00% optimize.opt_a.updatestate_loads_eliminate : 0.000023s : 0.00% optimize.opt_a.parameter_eliminate : 0.000006s : 0.00% optimize.opt_a.a_2 : 0.000512s : 0.02% optimize.opt_a.accelerated_algorithm : 0.000083s : 0.00% optimize.opt_a.shard : 0.000007s : 0.00% optimize.opt_a.meta_shard_fg_expand : 0.000015s : 0.00% optimize.opt_a.shard_inline : 0.000037s : 0.00% optimize.opt_a.merge_send_recv : 0.000035s : 0.00% optimize.opt_a.auto_parallel : 0.000034s : 0.00% optimize.opt_a.parallel : 0.000050s : 0.00% optimize.opt_a.flash_sp : 0.000023s : 0.00% optimize.opt_a.merge_comm : 0.000020s : 0.00% optimize.opt_a.allreduce_fusion : 0.000018s : 0.00% optimize.opt_a.matmul_add_comm_reduction : 0.000042s : 0.00% optimize.opt_a.allreduce_slice_to_reducescatter : 0.000002s : 0.00% optimize.opt_a.virtual_shard_identity : 0.000051s : 0.00% optimize.opt_a.virtual_dataset : 0.000036s : 0.00% optimize.opt_a.get_grad_eliminate_ : 0.000035s : 0.00% optimize.opt_a.virtual_output : 0.000035s : 0.00% optimize.opt_a.merge_forward : 0.000022s : 0.00% optimize.opt_a.cell_reuse_recompute_pass : 0.000008s : 0.00% optimize.opt_a.offload_activation : 0.000041s : 0.00% optimize.opt_a.cell_reuse_handle_not_recompute_node_pass : 0.000075s : 0.00% optimize.opt_a.merge_recompute_call_nodes : 0.000004s : 0.00% optimize.opt_a.before_grad : 0.000060s : 0.00% optimize.opt_a.set_forward_comm_id_for_comm_node_pass : 0.000022s : 0.00% optimize.opt_a.meta_fg_expand : 0.000014s : 0.00% optimize.opt_a.flash_sp_send_recv_attached : 0.000010s : 0.00% optimize.opt_a.receive_attached : 0.000005s : 0.00% optimize.opt_a.after_resolve : 0.000056s : 0.00% optimize.opt_a.a_after_grad : 0.000064s : 0.00% optimize.opt_a.renormalize : 0.012126s : 0.51% optimize.opt_a.add_forward_monad_depend : 0.000027s : 0.00% optimize.opt_a.auto_monad_grad : 0.000007s : 0.00% optimize.opt_a.auto_monad_eliminator : 0.000132s : 0.01% optimize.opt_a.cse : 0.001483s : 0.06% optimize.opt_a.a_3 : 0.000259s : 0.01% optimize.py_interpret_to_execute_after_opt_a : 0.000032s : 0.00% optimize.slice_cell_reuse_recomputed_activation : 0.000002s : 0.00% optimize.rewriter_after_opt_a : 0.000212s : 0.01% optimize.convert_after_rewriter : 0.000018s : 0.00% optimize.order_py_execute_after_rewriter : 0.000012s : 0.00% optimize.mutable_eliminate : 0.000873s : 0.04% optimize.opt_b.b_1 : 0.000386s : 0.02% optimize.opt_b.b_2 : 0.000021s : 0.00% optimize.opt_b.updatestate_depend_eliminate : 0.000019s : 0.00% optimize.opt_b.updatestate_assign_eliminate : 0.000008s : 0.00% optimize.opt_b.updatestate_loads_eliminate : 0.000013s : 0.00% optimize.opt_b.renormalize : 0.000001s : 0.00% optimize.opt_b.cse : 0.000089s : 0.00% optimize.optimize_parallel_all_gather_comm : 0.000037s : 0.00% optimize.overlap_param_gather : 0.000002s : 0.00% optimize.cconv : 0.000047s : 0.00% optimize.loop_unroll : 0.081232s : 3.43% optimize.opt_after_cconv.c_1 : 0.000122s : 0.01% optimize.opt_after_cconv.parameter_eliminate : 0.000008s : 0.00% optimize.opt_after_cconv.updatestate_depend_eliminate : 0.000023s : 0.00% optimize.opt_after_cconv.updatestate_assign_eliminate : 0.000008s : 0.00% optimize.opt_after_cconv.updatestate_loads_eliminate : 0.000012s : 0.00% optimize.opt_after_cconv.cse : 0.000086s : 0.00% optimize.opt_after_cconv.renormalize : 0.000001s : 0.00% optimize.remove_dup_value : 0.000074s : 0.00% optimize.tuple_transform.d_1 : 0.000118s : 0.00% optimize.tuple_transform.none_parameter_eliminate : 0.000002s : 0.00% optimize.tuple_transform.renormalize : 0.000000s : 0.00% optimize.tuple_transform.switch_simplify : 0.000017s : 0.00% optimize.partial_unused_args_eliminate : 0.000002s : 0.00% optimize.add_recomputation : 0.000122s : 0.01% optimize.cse_after_recomputation.cse : 0.000049s : 0.00% optimize.environ_conv : 0.000022s : 0.00% optimize.swap_dp_allreduce_reducescatter : 0.000014s : 0.00% optimize.bias_add_comm_swap : 0.000003s : 0.00% optimize.label_micro_interleaved_index : 0.000008s : 0.00% optimize.label_fine_grained_interleaved_index : 0.000003s : 0.00% optimize.merge_cast_opt : 0.000002s : 0.00% optimize.slice_recompute_activation : 0.000003s : 0.00% optimize.micro_interleaved_order_control : 0.000002s : 0.00% optimize.assign_add_opt : 0.000002s : 0.00% optimize.ForceFp32Comm : 0.000001s : 0.00% optimize.remove_cast_before_assign_add : 0.000001s : 0.00% optimize.full_micro_interleaved_order_control : 0.000002s : 0.00% optimize.reorder_send_recv_between_fp_bp : 0.000003s : 0.00% optimize.comm_op_add_attrs : 0.000001s : 0.00% optimize.add_comm_op_reuse_tag : 0.000001s : 0.00% optimize.interleave_split_concat_branches : 0.000002s : 0.00% optimize.interleave_parallel_branches : 0.000001s : 0.00% optimize.overlap_opt_shard_in_pipeline : 0.000001s : 0.00% optimize.overlap_opt_shard_grad_in_pipeline : 0.000002s : 0.00% optimize.control_data_broadcast_order : 0.000029s : 0.00% optimize.grouped_pairwise_exchange_alltoall : 0.000002s : 0.00% optimize.offloading_packed_experts : 0.000007s : 0.00% optimize.overlap_recompute_and_grad_model_parallel : 0.000008s : 0.00% optimize.overlap_grad_matmul_and_grad_allreduce : 0.000001s : 0.00% optimize.overlap_recompute_allgather_and_fa_grad : 0.000001s : 0.00% optimize.overlap_recompute_comm : 0.000002s : 0.00% optimize.overlap_grad_ring_attention : 0.000008s : 0.00% optimize.overlap_grad_flash_sp : 0.000038s : 0.00% optimize.begin_end_overlap_inline : 0.000001s : 0.00% optimize.split_matmul_comm_elemetwise : 0.000002s : 0.00% optimize.split_layernorm_comm : 0.000002s : 0.00% optimize.handle_group_info : 0.000001s : 0.00% optimize.symbol_engine_optimizer.build : 0.000016s : 0.00% optimize.symbol_engine_optimizer.elim_shapecalc : 0.000024s : 0.00% optimize.symbol_engine_optimizer.elim_not_effective : 0.000032s : 0.00% optimize.symbol_engine_optimizer.opt_reshape : 0.000024s : 0.00% optimize.symbol_engine_optimizer.fold_const_symbol : 0.000025s : 0.00% optimize.symbol_engine_optimizer.renormalize : 0.000000s : 0.00% detach_backward : 0.000003s : 0.00% pipeline_parallel_scheduler : 0.000002s : 0.00% auto_monad_reorder : 0.000047s : 0.00% get_jit_bprop_graph : 0.000002s : 0.00% rewriter_after_jit_bprop_graph : 0.000007s : 0.00% opt_after_jit_grad : 0.000749s : 0.03% validate : 0.000098s : 0.00% Time group info: ------[substitution.] 0.000326 96 3.44% : 0.000011s : 2: substitution.depend_value_elim 1.23% : 0.000004s : 8: substitution.elim_not_effective 1.14% : 0.000004s : 8: substitution.fold_const_symbol 3.62% : 0.000012s : 13: substitution.graph_param_transform 39.12% : 0.000127s : 1: substitution.inline 3.10% : 0.000010s : 16: substitution.j_node_and_user_rematch 9.21% : 0.000030s : 2: substitution.less_batch_normalization 2.46% : 0.000008s : 6: substitution.load_eliminater 0.99% : 0.000003s : 2: substitution.opt_reshape 3.82% : 0.000012s : 16: substitution.remove_not_recompute_node 2.69% : 0.000009s : 4: substitution.replace_old_param 11.21% : 0.000037s : 4: substitution.reshape_eliminate 10.33% : 0.000034s : 6: substitution.updatestate_pure_node_eliminater 7.64% : 0.000025s : 8: substitution.updatestate_useless_node_eliminater ------[type_inference.] 2.143232 2 99.75% : 2.137770s : 1: type_inference.infer 0.25% : 0.005462s : 1: type_inference.specialize ------[replace.] 0.000019 1 100.00% : 0.000019s : 1: replace.inline ------[match.] 0.000126 1 100.00% : 0.000126s : 1: match.inline ------[predicate.] 0.000493 3002 0.86% : 0.000004s : 27: predicate.accumulaten_eliminater 0.81% : 0.000004s : 13: predicate.ad_related_special_op_eliminate 0.69% : 0.000003s : 26: predicate.addn_check_dump 0.82% : 0.000004s : 27: predicate.addn_zero_filter 0.72% : 0.000004s : 27: predicate.adjust_all_reduce_mul_add 2.36% : 0.000012s : 53: predicate.arithmetic_simplify 0.92% : 0.000005s : 27: predicate.cast_eliminate 0.81% : 0.000004s : 26: predicate.check_bprop_eliminate 0.71% : 0.000003s : 26: predicate.compare_switch_simplify 0.22% : 0.000001s : 13: predicate.const_output_eliminate 0.84% : 0.000004s : 26: predicate.depend_value_elim 0.80% : 0.000004s : 27: predicate.dict_get_item_const_eliminator 0.91% : 0.000004s : 27: predicate.dict_get_item_eliminator 0.81% : 0.000004s : 27: predicate.dict_set_item_eliminator 0.94% : 0.000005s : 26: predicate.dumpgradient_eliminate 0.23% : 0.000001s : 13: predicate.elim_not_effective 0.53% : 0.000003s : 13: predicate.elim_shapecalc_of_broadcastargs 1.12% : 0.000006s : 40: predicate.environ_add_const_eliminate 1.07% : 0.000005s : 40: predicate.environ_get_add_eliminate 1.08% : 0.000005s : 40: predicate.environ_get_depend_swap 1.89% : 0.000009s : 66: predicate.environ_get_eliminate 1.06% : 0.000005s : 40: predicate.environ_get_set_eliminate 0.76% : 0.000004s : 28: predicate.exchange_switch_depend_value 1.66% : 0.000008s : 28: predicate.float_depend_g_call 0.67% : 0.000003s : 26: predicate.float_environ_get_switch 1.03% : 0.000005s : 39: predicate.float_tuple_getitem_switch 0.20% : 0.000001s : 13: predicate.fold_const_symbol 0.86% : 0.000004s : 26: predicate.get_grad_eliminate 0.27% : 0.000001s : 13: predicate.graph_param_transform 0.68% : 0.000003s : 26: predicate.incorporate_call 0.65% : 0.000003s : 26: predicate.incorporate_call_switch 5.42% : 0.000027s : 133: predicate.inline 1.10% : 0.000005s : 26: predicate.inline_without_move 0.39% : 0.000002s : 26: predicate.j_node_and_user_rematch 1.16% : 0.000006s : 27: predicate.less_batch_normalization 1.77% : 0.000009s : 53: predicate.list_to_tuple_eliminator_ 2.29% : 0.000011s : 80: predicate.load_eliminater 1.99% : 0.000010s : 13: predicate.loop_unroll_after_grad 0.95% : 0.000005s : 31: predicate.loop_unroll_before_grad 1.72% : 0.000008s : 53: predicate.make_slice_get_slice_eliminator 0.73% : 0.000004s : 26: predicate.merge_addn 0.72% : 0.000004s : 26: predicate.micro_step_allgather_replace 0.73% : 0.000004s : 26: predicate.mini_step_allgather_replace 0.70% : 0.000003s : 27: predicate.minmaximum_grad 1.23% : 0.000006s : 13: predicate.mutable_eliminate 0.51% : 0.000003s : 13: predicate.opt_reshape 0.44% : 0.000002s : 13: predicate.parallel_virtual_node 1.48% : 0.000007s : 28: predicate.partial_defer_inline 1.17% : 0.000006s : 40: predicate.partial_eliminate 0.81% : 0.000004s : 27: predicate.print_const_string_wrapper 0.78% : 0.000004s : 26: predicate.reduce_all_const_elim 1.16% : 0.000006s : 27: predicate.reduce_eliminate 2.22% : 0.000011s : 80: predicate.redundant_stop_gradient_eliminater 0.58% : 0.000003s : 26: predicate.remove_not_recompute_node 1.27% : 0.000006s : 53: predicate.replace_applicator 0.55% : 0.000003s : 26: predicate.replace_old_param 0.37% : 0.000002s : 13: predicate.reset_defer_inline 1.19% : 0.000006s : 27: predicate.reshape_eliminate 0.78% : 0.000004s : 26: predicate.row_tensor_add_zeros_like 0.44% : 0.000002s : 13: predicate.row_tensor_eliminate 1.27% : 0.000006s : 26: predicate.same_eliminate 0.52% : 0.000003s : 26: predicate.set_cell_output_no_recompute 1.23% : 0.000006s : 26: predicate.shard_identity_eliminate 0.87% : 0.000004s : 26: predicate.special_op_eliminate 0.99% : 0.000005s : 26: predicate.specialize_transform 1.34% : 0.000007s : 26: predicate.split_environ_get_set_with_tuple_value 0.90% : 0.000004s : 26: predicate.stack_unstack_eliminate 0.43% : 0.000002s : 13: predicate.switch_call_monad_eliminater 0.89% : 0.000004s : 28: predicate.switch_defer_inline 2.92% : 0.000014s : 54: predicate.switch_layer_defer_inline 3.27% : 0.000016s : 98: predicate.switch_simplify 0.85% : 0.000004s : 27: predicate.tile_eliminate 0.78% : 0.000004s : 27: predicate.transpose_eliminate 1.64% : 0.000008s : 53: predicate.tuple_list_convert_item_index_to_positive 1.60% : 0.000008s : 53: predicate.tuple_list_get_item_const_eliminator 1.43% : 0.000007s : 53: predicate.tuple_list_get_item_depend_reorder 3.26% : 0.000016s : 79: predicate.tuple_list_get_item_eliminator 1.65% : 0.000008s : 53: predicate.tuple_list_get_set_item_eliminator 2.46% : 0.000012s : 79: predicate.tuple_list_set_item_eliminator 1.68% : 0.000008s : 53: predicate.tuple_to_list_eliminator_ 2.22% : 0.000011s : 80: predicate.updatestate_pure_node_eliminater 3.09% : 0.000015s : 106: predicate.updatestate_useless_node_eliminater 0.46% : 0.000002s : 13: predicate.value_based_eliminate 0.81% : 0.000004s : 26: predicate.virtual_dataset_eliminate 0.86% : 0.000004s : 26: predicate.virtual_output_eliminate 0.44% : 0.000002s : 13: predicate.virtual_view_grad_eliminate 0.51% : 0.000003s : 13: predicate.zero_like_fill_zero ------[func_graph_cloner_run.] 0.156783 58 97.45% : 0.152787s : 55: func_graph_cloner_run.FuncGraphClonerGraph 2.55% : 0.003996s : 3: func_graph_cloner_run.FuncGraphSpecializer ------[meta_graph.] 0.000000 0 ------[manager.] 0.000000 0 ------[pynative] 0.000000 0 ------[others.] 2.753266 192 0.00% : 0.000003s : 1: ForceFp32Comm 0.25% : 0.007009s : 1: add_attr 0.25% : 0.006989s : 1: add_attr_with_inline 0.00% : 0.000004s : 1: add_comm_op_reuse_tag 0.00% : 0.000127s : 1: add_recomputation 0.00% : 0.000004s : 1: assign_add_opt 0.01% : 0.000405s : 1: auto_monad 0.00% : 0.000052s : 1: auto_monad_reorder 0.00% : 0.000003s : 1: begin_end_overlap_inline 0.00% : 0.000006s : 1: bias_add_comm_swap 0.03% : 0.000908s : 1: bootstrap 0.00% : 0.000054s : 1: cconv 0.00% : 0.000005s : 1: comm_op_add_attrs 0.00% : 0.000033s : 1: control_data_broadcast_order 0.00% : 0.000023s : 1: convert_after_rewriter 0.00% : 0.000065s : 1: cse_after_recomputation 0.00% : 0.000006s : 1: dataset_repeat_opt 0.00% : 0.000006s : 1: detach_backward 0.00% : 0.000027s : 1: environ_conv 0.00% : 0.000027s : 1: event_method 0.00% : 0.000005s : 1: full_micro_interleaved_order_control 0.00% : 0.000006s : 1: get_jit_bprop_graph 0.00% : 0.000012s : 1: graph_reusing 0.00% : 0.000005s : 1: grouped_pairwise_exchange_alltoall 0.00% : 0.000004s : 1: handle_group_info 0.00% : 0.000006s : 1: inline 0.00% : 0.000008s : 1: insert-virtual-dataset 0.00% : 0.000005s : 1: interleave_parallel_branches 0.00% : 0.000004s : 1: interleave_split_concat_branches 0.00% : 0.000006s : 1: label_fine_grained_interleaved_index 0.00% : 0.000011s : 1: label_micro_interleaved_index 2.95% : 0.081254s : 1: loop_unroll 0.00% : 0.000005s : 1: merge_cast_opt 0.00% : 0.000005s : 1: micro_interleaved_order_control 0.03% : 0.000894s : 1: mutable_eliminate 0.00% : 0.000010s : 1: offloading_packed_experts 0.00% : 0.000055s : 1: opt.transform.loop_unroll_optimizer 0.00% : 0.000043s : 1: opt.transform.mutable_eliminate 4.61% : 0.127000s : 78: opt.transform.opt_a 0.00% : 0.000120s : 1: opt.transform.opt_after_cconv 0.00% : 0.000063s : 1: opt.transform.opt_after_jit_grad 0.01% : 0.000368s : 28: opt.transform.opt_b 0.00% : 0.000133s : 2: opt.transform.opt_trans_graph 0.00% : 0.000099s : 4: opt.transform.symbol_engine_opt 5.17% : 0.142267s : 1: opt_a 0.01% : 0.000319s : 1: opt_after_cconv 0.03% : 0.000764s : 1: opt_after_jit_grad 0.02% : 0.000611s : 1: opt_b 8.24% : 0.226860s : 1: optimize 0.00% : 0.000044s : 1: optimize_parallel_all_gather_comm 0.00% : 0.000017s : 1: order_py_execute_after_rewriter 0.00% : 0.000042s : 1: overlap_grad_flash_sp 0.00% : 0.000004s : 1: overlap_grad_matmul_and_grad_allreduce 0.00% : 0.000013s : 1: overlap_grad_ring_attention 0.00% : 0.000004s : 1: overlap_opt_shard_grad_in_pipeline 0.00% : 0.000004s : 1: overlap_opt_shard_in_pipeline 0.00% : 0.000007s : 1: overlap_param_gather 0.00% : 0.000006s : 1: overlap_recompute_allgather_and_fa_grad 0.00% : 0.000011s : 1: overlap_recompute_and_grad_model_parallel 0.00% : 0.000005s : 1: overlap_recompute_comm 0.00% : 0.000010s : 1: parallel-infer-symbol 0.00% : 0.000004s : 1: parallel-infer-symbol-second 0.00% : 0.000006s : 1: partial_unused_args_eliminate 0.00% : 0.000005s : 1: pipeline_parallel_scheduler 0.00% : 0.000006s : 1: pipeline_split 0.00% : 0.000042s : 1: pre_auto_parallel 0.00% : 0.000037s : 1: py_interpret_to_execute 0.00% : 0.000036s : 1: py_interpret_to_execute_after_opt_a 0.00% : 0.000004s : 1: remove_cast_before_assign_add 0.00% : 0.000080s : 1: remove_dup_value 0.28% : 0.007699s : 1: renormalize.infer 0.16% : 0.004407s : 1: renormalize.specialize 0.00% : 0.000006s : 1: reorder_send_recv_between_fp_bp 0.00% : 0.000010s : 1: rewriter_after_jit_bprop_graph 0.01% : 0.000223s : 1: rewriter_after_opt_a 0.00% : 0.000088s : 1: rewriter_before_opt_a 0.00% : 0.000005s : 1: slice_cell_reuse_recomputed_activation 0.00% : 0.000005s : 1: slice_recompute_activation 0.00% : 0.000006s : 1: split_layernorm_comm 0.00% : 0.000005s : 1: split_matmul_comm_elemetwise 0.00% : 0.000017s : 1: swap_dp_allreduce_reducescatter 0.01% : 0.000165s : 1: symbol_engine_optimizer 0.01% : 0.000168s : 1: tuple_transform 77.85% : 2.143377s : 1: type_inference . [hook] pytest_runtest_teardown:test_matmul_qkv_bf16_1[False-mstype0-k_n_shape2-256] tests/st/infer/ops/test_internal_ops/test_matmul_split.py::test_matmul_qkv_bf16_1[False-mstype0-k_n_shape2-256],max_mem:212.0M [WARNING] ME(168319:281472866918192,MainProcess):2026-01-29-17:42:28.895.711 [mindspore/context.py:1334] For 'context.set_context', the parameter 'device_target' will be deprecated and removed in a future version. Please use the api mindspore.set_device() instead. TotalTime = 1.52682, [21] [bootstrap]: 0.00070809 [type_inference]: 1.41962 [event_method]: 1.907e-05 [auto_monad]: 0.00034026 [graph_reusing]: 8.03001e-06 [inline]: 3.15002e-06 [add_attr]: 0.0712117, [1] [add_attr_with_inline]: 0.0711949, [1] [Cycle 1]: 7.465e-05, [2] [tag_attr]: 2.468e-05 [meta_addattr_fg_expand]: 4.42e-06 [parallel-infer-symbol]: 5.09003e-06 [pre_auto_parallel]: 4.074e-05 [insert-virtual-dataset]: 2.96999e-06 [parallel-infer-symbol-second]: 8.70001e-07 [dataset_repeat_opt]: 2.07999e-06 [pipeline_split]: 1.57001e-06 [optimize]: 0.0201211, [53] [py_interpret_to_execute]: 3.402e-05 [rewriter_before_opt_a]: 8.724e-05 [opt_a]: 0.0165129, [2] [Cycle 1]: 0.0147837, [45] [expand_dump_flag]: 3.63999e-06 [switch_simplify]: 3.37e-05 [loop_unroll]: 1.949e-05 [a_1]: 0.00086937 [with_stream_mark]: 3.242e-05 [recompute_prepare]: 2.585e-05 [updatestate_depend_eliminate]: 1.134e-05 [updatestate_assign_eliminate]: 1.526e-05 [updatestate_loads_eliminate]: 1.175e-05 [parameter_eliminate]: 2.91e-06 [a_2]: 0.00025505 [accelerated_algorithm]: 4.581e-05 [shard]: 3.23998e-06 [meta_shard_fg_expand]: 4.32998e-06 [shard_inline]: 1.796e-05 [merge_send_recv]: 1.504e-05 [auto_parallel]: 1.461e-05 [parallel]: 3.516e-05 [flash_sp]: 1.327e-05 [merge_comm]: 9.04998e-06 [allreduce_fusion]: 8.54e-06 [matmul_add_comm_reduction]: 1.796e-05 [allreduce_slice_to_reducescatter]: 9.20001e-07 [virtual_shard_identity]: 2.066e-05 [virtual_dataset]: 1.789e-05 [get_grad_eliminate_]: 1.891e-05 [virtual_output]: 1.784e-05 [merge_forward]: 9.89001e-06 [cell_reuse_recompute_pass]: 2.11e-06 [offload_activation]: 1.929e-05 [cell_reuse_handle_not_recompute_node_pass]: 3.585e-05 [merge_recompute_call_nodes]: 2.11e-06 [before_grad]: 2.945e-05 [set_forward_comm_id_for_comm_node_pass]: 9.25001e-06 [meta_fg_expand]: 6.08998e-06 [flash_sp_send_recv_attached]: 7.31001e-06 [receive_attached]: 2.66999e-06 [after_resolve]: 2.611e-05 [a_after_grad]: 3.103e-05 [renormalize]: 0.0112527 [add_forward_monad_depend]: 2.154e-05 [auto_monad_grad]: 2.79999e-06 [auto_monad_eliminator]: 6.62e-05 [cse]: 0.00121889 [a_3]: 0.00014395 [Cycle 2]: 0.00171426, [45] [expand_dump_flag]: 3.88999e-06 [switch_simplify]: 1.947e-05 [loop_unroll]: 1.762e-05 [a_1]: 0.00047574 [with_stream_mark]: 3.078e-05 [recompute_prepare]: 1.871e-05 [updatestate_depend_eliminate]: 1.001e-05 [updatestate_assign_eliminate]: 7.46999e-06 [updatestate_loads_eliminate]: 1.194e-05 [parameter_eliminate]: 2.93e-06 [a_2]: 0.00022271 [accelerated_algorithm]: 2.292e-05 [shard]: 2.22001e-06 [meta_shard_fg_expand]: 5.32001e-06 [shard_inline]: 1.553e-05 [merge_send_recv]: 1.5e-05 [auto_parallel]: 1.603e-05 [parallel]: 1.113e-05 [flash_sp]: 5.38002e-06 [merge_comm]: 8.64e-06 [allreduce_fusion]: 9.34e-06 [matmul_add_comm_reduction]: 1.923e-05 [allreduce_slice_to_reducescatter]: 8.70001e-07 [virtual_shard_identity]: 1.939e-05 [virtual_dataset]: 1.5e-05 [get_grad_eliminate_]: 1.487e-05 [virtual_output]: 1.579e-05 [merge_forward]: 8.45999e-06 [cell_reuse_recompute_pass]: 3.63999e-06 [offload_activation]: 1.774e-05 [cell_reuse_handle_not_recompute_node_pass]: 2.994e-05 [merge_recompute_call_nodes]: 1.72001e-06 [before_grad]: 2.535e-05 [set_forward_comm_id_for_comm_node_pass]: 9.62001e-06 [meta_fg_expand]: 6.91001e-06 [flash_sp_send_recv_attached]: 1.64998e-06 [receive_attached]: 2.88e-06 [after_resolve]: 2.512e-05 [a_after_grad]: 2.781e-05 [renormalize]: 8.00064e-08 [add_forward_monad_depend]: 4.03001e-06 [auto_monad_grad]: 3.11001e-06 [auto_monad_eliminator]: 5.57e-05 [cse]: 0.00010093 [a_3]: 0.00010644 [py_interpret_to_execute_after_opt_a]: 2.976e-05 [slice_cell_reuse_recomputed_activation]: 2.22999e-06 [rewriter_after_opt_a]: 0.00020689 [convert_after_rewriter]: 1.714e-05 [order_py_execute_after_rewriter]: 1.1e-05 [mutable_eliminate]: 0.00081936 [opt_b]: 0.00055615, [1] [Cycle 1]: 0.0005459, [7] [b_1]: 0.00037996 [b_2]: 2.01e-05 [updatestate_depend_eliminate]: 1.413e-05 [updatestate_assign_eliminate]: 7.2e-06 [updatestate_loads_eliminate]: 1.043e-05 [renormalize]: 5.69999e-07 [cse]: 6.672e-05 [optimize_parallel_all_gather_comm]: 3.048e-05 [overlap_param_gather]: 1.91e-06 [cconv]: 4.054e-05 [loop_unroll]: 0.00054999 [opt_after_cconv]: 0.00022302, [1] [Cycle 1]: 0.00021461, [7] [c_1]: 9.791e-05 [parameter_eliminate]: 5.20999e-06 [updatestate_depend_eliminate]: 1.349e-05 [updatestate_assign_eliminate]: 6.92002e-06 [updatestate_loads_eliminate]: 8.97999e-06 [cse]: 4.561e-05 [renormalize]: 5.19998e-07 [remove_dup_value]: 7.317e-05 [tuple_transform]: 0.00016173, [1] [Cycle 1]: 0.00015615, [4] [d_1]: 0.00011438 [none_parameter_eliminate]: 1.97001e-06 [renormalize]: 1.79978e-07 [switch_simplify]: 1.655e-05 [partial_unused_args_eliminate]: 2.53e-06 [add_recomputation]: 0.00011457 [cse_after_recomputation]: 5.012e-05, [1] [Cycle 1]: 4.505e-05, [1] [cse]: 3.911e-05 [environ_conv]: 1.805e-05 [swap_dp_allreduce_reducescatter]: 1.185e-05 [bias_add_comm_swap]: 4.22e-06 [label_micro_interleaved_index]: 4.75999e-06 [label_fine_grained_interleaved_index]: 2.93e-06 [merge_cast_opt]: 1.50999e-06 [slice_recompute_activation]: 2.49999e-06 [micro_interleaved_order_control]: 2.34999e-06 [assign_add_opt]: 1.59998e-06 [ForceFp32Comm]: 1.07e-06 [remove_cast_before_assign_add]: 1.32e-06 [full_micro_interleaved_order_control]: 2.20002e-06 [reorder_send_recv_between_fp_bp]: 2.85998e-06 [comm_op_add_attrs]: 1.14998e-06 [add_comm_op_reuse_tag]: 1.04e-06 [interleave_split_concat_branches]: 1.35999e-06 [interleave_parallel_branches]: 1.15001e-06 [overlap_opt_shard_in_pipeline]: 1.31002e-06 [overlap_opt_shard_grad_in_pipeline]: 2.02001e-06 [control_data_broadcast_order]: 2.542e-05 [grouped_pairwise_exchange_alltoall]: 1.78002e-06 [offloading_packed_experts]: 6.69999e-06 [overlap_recompute_and_grad_model_parallel]: 7.45e-06 [overlap_grad_matmul_and_grad_allreduce]: 1.71002e-06 [overlap_recompute_allgather_and_fa_grad]: 1.81e-06 [overlap_recompute_comm]: 2.51e-06 [overlap_grad_ring_attention]: 8.13001e-06 [overlap_grad_flash_sp]: 3.891e-05 [begin_end_overlap_inline]: 5.40022e-07 [split_matmul_comm_elemetwise]: 2.21e-06 [split_layernorm_comm]: 1.92999e-06 [handle_group_info]: 1.32999e-06 [symbol_engine_optimizer]: 0.00015221, [1] [Cycle 1]: 0.00014614, [6] [build]: 1.623e-05 [elim_shapecalc]: 2.141e-05 [elim_not_effective]: 2.902e-05 [opt_reshape]: 2.239e-05 [fold_const_symbol]: 2.452e-05 [renormalize]: 2.20025e-07 [detach_backward]: 2.41998e-06 [pipeline_parallel_scheduler]: 1.52001e-06 [auto_monad_reorder]: 4.396e-05 [get_jit_bprop_graph]: 1.89e-06 [rewriter_after_jit_bprop_graph]: 6.07001e-06 [opt_after_jit_grad]: 0.0142955 [validate]: 0.00012372 Sums bootstrap : 0.000708s : 0.05% type_inference : 1.419624s : 97.62% event_method : 0.000019s : 0.00% auto_monad : 0.000340s : 0.02% graph_reusing : 0.000008s : 0.00% inline : 0.000003s : 0.00% add_attr.add_attr_with_inline.tag_attr : 0.000025s : 0.00% add_attr.add_attr_with_inline.meta_addattr_fg_expand : 0.000004s : 0.00% parallel-infer-symbol : 0.000005s : 0.00% pre_auto_parallel : 0.000041s : 0.00% insert-virtual-dataset : 0.000003s : 0.00% parallel-infer-symbol-second : 0.000001s : 0.00% dataset_repeat_opt : 0.000002s : 0.00% pipeline_split : 0.000002s : 0.00% optimize.py_interpret_to_execute : 0.000034s : 0.00% optimize.rewriter_before_opt_a : 0.000087s : 0.01% optimize.opt_a.expand_dump_flag : 0.000008s : 0.00% optimize.opt_a.switch_simplify : 0.000053s : 0.00% optimize.opt_a.loop_unroll : 0.000037s : 0.00% optimize.opt_a.a_1 : 0.001345s : 0.09% optimize.opt_a.with_stream_mark : 0.000063s : 0.00% optimize.opt_a.recompute_prepare : 0.000045s : 0.00% optimize.opt_a.updatestate_depend_eliminate : 0.000021s : 0.00% optimize.opt_a.updatestate_assign_eliminate : 0.000023s : 0.00% optimize.opt_a.updatestate_loads_eliminate : 0.000024s : 0.00% optimize.opt_a.parameter_eliminate : 0.000006s : 0.00% optimize.opt_a.a_2 : 0.000478s : 0.03% optimize.opt_a.accelerated_algorithm : 0.000069s : 0.00% optimize.opt_a.shard : 0.000005s : 0.00% optimize.opt_a.meta_shard_fg_expand : 0.000010s : 0.00% optimize.opt_a.shard_inline : 0.000033s : 0.00% optimize.opt_a.merge_send_recv : 0.000030s : 0.00% optimize.opt_a.auto_parallel : 0.000031s : 0.00% optimize.opt_a.parallel : 0.000046s : 0.00% optimize.opt_a.flash_sp : 0.000019s : 0.00% optimize.opt_a.merge_comm : 0.000018s : 0.00% optimize.opt_a.allreduce_fusion : 0.000018s : 0.00% optimize.opt_a.matmul_add_comm_reduction : 0.000037s : 0.00% optimize.opt_a.allreduce_slice_to_reducescatter : 0.000002s : 0.00% optimize.opt_a.virtual_shard_identity : 0.000040s : 0.00% optimize.opt_a.virtual_dataset : 0.000033s : 0.00% optimize.opt_a.get_grad_eliminate_ : 0.000034s : 0.00% optimize.opt_a.virtual_output : 0.000034s : 0.00% optimize.opt_a.merge_forward : 0.000018s : 0.00% optimize.opt_a.cell_reuse_recompute_pass : 0.000006s : 0.00% optimize.opt_a.offload_activation : 0.000037s : 0.00% optimize.opt_a.cell_reuse_handle_not_recompute_node_pass : 0.000066s : 0.00% optimize.opt_a.merge_recompute_call_nodes : 0.000004s : 0.00% optimize.opt_a.before_grad : 0.000055s : 0.00% optimize.opt_a.set_forward_comm_id_for_comm_node_pass : 0.000019s : 0.00% optimize.opt_a.meta_fg_expand : 0.000013s : 0.00% optimize.opt_a.flash_sp_send_recv_attached : 0.000009s : 0.00% optimize.opt_a.receive_attached : 0.000006s : 0.00% optimize.opt_a.after_resolve : 0.000051s : 0.00% optimize.opt_a.a_after_grad : 0.000059s : 0.00% optimize.opt_a.renormalize : 0.011253s : 0.77% optimize.opt_a.add_forward_monad_depend : 0.000026s : 0.00% optimize.opt_a.auto_monad_grad : 0.000006s : 0.00% optimize.opt_a.auto_monad_eliminator : 0.000122s : 0.01% optimize.opt_a.cse : 0.001320s : 0.09% optimize.opt_a.a_3 : 0.000250s : 0.02% optimize.py_interpret_to_execute_after_opt_a : 0.000030s : 0.00% optimize.slice_cell_reuse_recomputed_activation : 0.000002s : 0.00% optimize.rewriter_after_opt_a : 0.000207s : 0.01% optimize.convert_after_rewriter : 0.000017s : 0.00% optimize.order_py_execute_after_rewriter : 0.000011s : 0.00% optimize.mutable_eliminate : 0.000819s : 0.06% optimize.opt_b.b_1 : 0.000380s : 0.03% optimize.opt_b.b_2 : 0.000020s : 0.00% optimize.opt_b.updatestate_depend_eliminate : 0.000014s : 0.00% optimize.opt_b.updatestate_assign_eliminate : 0.000007s : 0.00% optimize.opt_b.updatestate_loads_eliminate : 0.000010s : 0.00% optimize.opt_b.renormalize : 0.000001s : 0.00% optimize.opt_b.cse : 0.000067s : 0.00% optimize.optimize_parallel_all_gather_comm : 0.000030s : 0.00% optimize.overlap_param_gather : 0.000002s : 0.00% optimize.cconv : 0.000041s : 0.00% optimize.loop_unroll : 0.000550s : 0.04% optimize.opt_after_cconv.c_1 : 0.000098s : 0.01% optimize.opt_after_cconv.parameter_eliminate : 0.000005s : 0.00% optimize.opt_after_cconv.updatestate_depend_eliminate : 0.000013s : 0.00% optimize.opt_after_cconv.updatestate_assign_eliminate : 0.000007s : 0.00% optimize.opt_after_cconv.updatestate_loads_eliminate : 0.000009s : 0.00% optimize.opt_after_cconv.cse : 0.000046s : 0.00% optimize.opt_after_cconv.renormalize : 0.000001s : 0.00% optimize.remove_dup_value : 0.000073s : 0.01% optimize.tuple_transform.d_1 : 0.000114s : 0.01% optimize.tuple_transform.none_parameter_eliminate : 0.000002s : 0.00% optimize.tuple_transform.renormalize : 0.000000s : 0.00% optimize.tuple_transform.switch_simplify : 0.000017s : 0.00% optimize.partial_unused_args_eliminate : 0.000003s : 0.00% optimize.add_recomputation : 0.000115s : 0.01% optimize.cse_after_recomputation.cse : 0.000039s : 0.00% optimize.environ_conv : 0.000018s : 0.00% optimize.swap_dp_allreduce_reducescatter : 0.000012s : 0.00% optimize.bias_add_comm_swap : 0.000004s : 0.00% optimize.label_micro_interleaved_index : 0.000005s : 0.00% optimize.label_fine_grained_interleaved_index : 0.000003s : 0.00% optimize.merge_cast_opt : 0.000002s : 0.00% optimize.slice_recompute_activation : 0.000002s : 0.00% optimize.micro_interleaved_order_control : 0.000002s : 0.00% optimize.assign_add_opt : 0.000002s : 0.00% optimize.ForceFp32Comm : 0.000001s : 0.00% optimize.remove_cast_before_assign_add : 0.000001s : 0.00% optimize.full_micro_interleaved_order_control : 0.000002s : 0.00% optimize.reorder_send_recv_between_fp_bp : 0.000003s : 0.00% optimize.comm_op_add_attrs : 0.000001s : 0.00% optimize.add_comm_op_reuse_tag : 0.000001s : 0.00% optimize.interleave_split_concat_branches : 0.000001s : 0.00% optimize.interleave_parallel_branches : 0.000001s : 0.00% optimize.overlap_opt_shard_in_pipeline : 0.000001s : 0.00% optimize.overlap_opt_shard_grad_in_pipeline : 0.000002s : 0.00% optimize.control_data_broadcast_order : 0.000025s : 0.00% optimize.grouped_pairwise_exchange_alltoall : 0.000002s : 0.00% optimize.offloading_packed_experts : 0.000007s : 0.00% optimize.overlap_recompute_and_grad_model_parallel : 0.000007s : 0.00% optimize.overlap_grad_matmul_and_grad_allreduce : 0.000002s : 0.00% optimize.overlap_recompute_allgather_and_fa_grad : 0.000002s : 0.00% optimize.overlap_recompute_comm : 0.000003s : 0.00% optimize.overlap_grad_ring_attention : 0.000008s : 0.00% optimize.overlap_grad_flash_sp : 0.000039s : 0.00% optimize.begin_end_overlap_inline : 0.000001s : 0.00% optimize.split_matmul_comm_elemetwise : 0.000002s : 0.00% optimize.split_layernorm_comm : 0.000002s : 0.00% optimize.handle_group_info : 0.000001s : 0.00% optimize.symbol_engine_optimizer.build : 0.000016s : 0.00% optimize.symbol_engine_optimizer.elim_shapecalc : 0.000021s : 0.00% optimize.symbol_engine_optimizer.elim_not_effective : 0.000029s : 0.00% optimize.symbol_engine_optimizer.opt_reshape : 0.000022s : 0.00% optimize.symbol_engine_optimizer.fold_const_symbol : 0.000025s : 0.00% optimize.symbol_engine_optimizer.renormalize : 0.000000s : 0.00% detach_backward : 0.000002s : 0.00% pipeline_parallel_scheduler : 0.000002s : 0.00% auto_monad_reorder : 0.000044s : 0.00% get_jit_bprop_graph : 0.000002s : 0.00% rewriter_after_jit_bprop_graph : 0.000006s : 0.00% opt_after_jit_grad : 0.014296s : 0.98% validate : 0.000124s : 0.01% Time group info: ------[substitution.] 0.000303 96 3.90% : 0.000012s : 2: substitution.depend_value_elim 1.26% : 0.000004s : 8: substitution.elim_not_effective 1.12% : 0.000003s : 8: substitution.fold_const_symbol 4.03% : 0.000012s : 13: substitution.graph_param_transform 44.22% : 0.000134s : 1: substitution.inline 3.08% : 0.000009s : 16: substitution.j_node_and_user_rematch 7.92% : 0.000024s : 2: substitution.less_batch_normalization 1.98% : 0.000006s : 6: substitution.load_eliminater 1.02% : 0.000003s : 2: substitution.opt_reshape 4.01% : 0.000012s : 16: substitution.remove_not_recompute_node 2.53% : 0.000008s : 4: substitution.replace_old_param 10.73% : 0.000032s : 4: substitution.reshape_eliminate 8.19% : 0.000025s : 6: substitution.updatestate_pure_node_eliminater 6.01% : 0.000018s : 8: substitution.updatestate_useless_node_eliminater ------[type_inference.] 1.419520 2 99.59% : 1.413651s : 1: type_inference.infer 0.41% : 0.005869s : 1: type_inference.specialize ------[replace.] 0.000022 1 100.00% : 0.000022s : 1: replace.inline ------[match.] 0.000132 1 100.00% : 0.000132s : 1: match.inline ------[predicate.] 0.000448 3002 0.86% : 0.000004s : 27: predicate.accumulaten_eliminater 2.35% : 0.000011s : 13: predicate.ad_related_special_op_eliminate 0.72% : 0.000003s : 26: predicate.addn_check_dump 0.79% : 0.000004s : 27: predicate.addn_zero_filter 0.74% : 0.000003s : 27: predicate.adjust_all_reduce_mul_add 2.00% : 0.000009s : 53: predicate.arithmetic_simplify 0.89% : 0.000004s : 27: predicate.cast_eliminate 0.83% : 0.000004s : 26: predicate.check_bprop_eliminate 0.73% : 0.000003s : 26: predicate.compare_switch_simplify 0.25% : 0.000001s : 13: predicate.const_output_eliminate 0.82% : 0.000004s : 26: predicate.depend_value_elim 0.83% : 0.000004s : 27: predicate.dict_get_item_const_eliminator 0.97% : 0.000004s : 27: predicate.dict_get_item_eliminator 0.79% : 0.000004s : 27: predicate.dict_set_item_eliminator 1.78% : 0.000008s : 26: predicate.dumpgradient_eliminate 0.30% : 0.000001s : 13: predicate.elim_not_effective 0.47% : 0.000002s : 13: predicate.elim_shapecalc_of_broadcastargs 1.14% : 0.000005s : 40: predicate.environ_add_const_eliminate 1.10% : 0.000005s : 40: predicate.environ_get_add_eliminate 1.14% : 0.000005s : 40: predicate.environ_get_depend_swap 1.89% : 0.000008s : 66: predicate.environ_get_eliminate 1.11% : 0.000005s : 40: predicate.environ_get_set_eliminate 0.77% : 0.000003s : 28: predicate.exchange_switch_depend_value 1.42% : 0.000006s : 28: predicate.float_depend_g_call 0.74% : 0.000003s : 26: predicate.float_environ_get_switch 1.06% : 0.000005s : 39: predicate.float_tuple_getitem_switch 0.22% : 0.000001s : 13: predicate.fold_const_symbol 0.86% : 0.000004s : 26: predicate.get_grad_eliminate 0.28% : 0.000001s : 13: predicate.graph_param_transform 0.80% : 0.000004s : 26: predicate.incorporate_call 0.71% : 0.000003s : 26: predicate.incorporate_call_switch 5.02% : 0.000023s : 133: predicate.inline 1.14% : 0.000005s : 26: predicate.inline_without_move 0.43% : 0.000002s : 26: predicate.j_node_and_user_rematch 1.26% : 0.000006s : 27: predicate.less_batch_normalization 1.62% : 0.000007s : 53: predicate.list_to_tuple_eliminator_ 2.32% : 0.000010s : 80: predicate.load_eliminater 0.84% : 0.000004s : 13: predicate.loop_unroll_after_grad 0.97% : 0.000004s : 31: predicate.loop_unroll_before_grad 1.71% : 0.000008s : 53: predicate.make_slice_get_slice_eliminator 0.77% : 0.000003s : 26: predicate.merge_addn 0.83% : 0.000004s : 26: predicate.micro_step_allgather_replace 0.74% : 0.000003s : 26: predicate.mini_step_allgather_replace 0.82% : 0.000004s : 27: predicate.minmaximum_grad 1.09% : 0.000005s : 13: predicate.mutable_eliminate 0.56% : 0.000003s : 13: predicate.opt_reshape 0.60% : 0.000003s : 13: predicate.parallel_virtual_node 1.33% : 0.000006s : 28: predicate.partial_defer_inline 1.31% : 0.000006s : 40: predicate.partial_eliminate 0.79% : 0.000004s : 27: predicate.print_const_string_wrapper 0.77% : 0.000003s : 26: predicate.reduce_all_const_elim 1.01% : 0.000005s : 27: predicate.reduce_eliminate 2.27% : 0.000010s : 80: predicate.redundant_stop_gradient_eliminater 0.54% : 0.000002s : 26: predicate.remove_not_recompute_node 1.32% : 0.000006s : 53: predicate.replace_applicator 0.57% : 0.000003s : 26: predicate.replace_old_param 0.30% : 0.000001s : 13: predicate.reset_defer_inline 1.13% : 0.000005s : 27: predicate.reshape_eliminate 0.81% : 0.000004s : 26: predicate.row_tensor_add_zeros_like 0.55% : 0.000002s : 13: predicate.row_tensor_eliminate 1.23% : 0.000006s : 26: predicate.same_eliminate 0.59% : 0.000003s : 26: predicate.set_cell_output_no_recompute 0.89% : 0.000004s : 26: predicate.shard_identity_eliminate 0.99% : 0.000004s : 26: predicate.special_op_eliminate 0.86% : 0.000004s : 26: predicate.specialize_transform 1.22% : 0.000005s : 26: predicate.split_environ_get_set_with_tuple_value 1.09% : 0.000005s : 26: predicate.stack_unstack_eliminate 0.43% : 0.000002s : 13: predicate.switch_call_monad_eliminater 0.88% : 0.000004s : 28: predicate.switch_defer_inline 1.64% : 0.000007s : 54: predicate.switch_layer_defer_inline 3.31% : 0.000015s : 98: predicate.switch_simplify 0.80% : 0.000004s : 27: predicate.tile_eliminate 0.82% : 0.000004s : 27: predicate.transpose_eliminate 1.63% : 0.000007s : 53: predicate.tuple_list_convert_item_index_to_positive 1.74% : 0.000008s : 53: predicate.tuple_list_get_item_const_eliminator 1.58% : 0.000007s : 53: predicate.tuple_list_get_item_depend_reorder 2.95% : 0.000013s : 79: predicate.tuple_list_get_item_eliminator 1.66% : 0.000007s : 53: predicate.tuple_list_get_set_item_eliminator 2.41% : 0.000011s : 79: predicate.tuple_list_set_item_eliminator 1.64% : 0.000007s : 53: predicate.tuple_to_list_eliminator_ 2.20% : 0.000010s : 80: predicate.updatestate_pure_node_eliminater 3.33% : 0.000015s : 106: predicate.updatestate_useless_node_eliminater 0.47% : 0.000002s : 13: predicate.value_based_eliminate 0.83% : 0.000004s : 26: predicate.virtual_dataset_eliminate 0.97% : 0.000004s : 26: predicate.virtual_output_eliminate 0.55% : 0.000002s : 13: predicate.virtual_view_grad_eliminate 0.44% : 0.000002s : 13: predicate.zero_like_fill_zero ------[func_graph_cloner_run.] 0.014070 58 71.49% : 0.010059s : 55: func_graph_cloner_run.FuncGraphClonerGraph 28.51% : 0.004011s : 3: func_graph_cloner_run.FuncGraphSpecializer ------[meta_graph.] 0.000000 0 ------[manager.] 0.000000 0 ------[pynative] 0.000000 0 ------[others.] 1.632532 192 0.00% : 0.000004s : 1: ForceFp32Comm 4.36% : 0.071219s : 1: add_attr 4.36% : 0.071200s : 1: add_attr_with_inline 0.00% : 0.000004s : 1: add_comm_op_reuse_tag 0.01% : 0.000120s : 1: add_recomputation 0.00% : 0.000004s : 1: assign_add_opt 0.02% : 0.000353s : 1: auto_monad 0.00% : 0.000048s : 1: auto_monad_reorder 0.00% : 0.000003s : 1: begin_end_overlap_inline 0.00% : 0.000007s : 1: bias_add_comm_swap 0.05% : 0.000772s : 1: bootstrap 0.00% : 0.000044s : 1: cconv 0.00% : 0.000005s : 1: comm_op_add_attrs 0.00% : 0.000029s : 1: control_data_broadcast_order 0.00% : 0.000023s : 1: convert_after_rewriter 0.00% : 0.000053s : 1: cse_after_recomputation 0.00% : 0.000005s : 1: dataset_repeat_opt 0.00% : 0.000006s : 1: detach_backward 0.00% : 0.000022s : 1: environ_conv 0.00% : 0.000027s : 1: event_method 0.00% : 0.000005s : 1: full_micro_interleaved_order_control 0.00% : 0.000006s : 1: get_jit_bprop_graph 0.00% : 0.000013s : 1: graph_reusing 0.00% : 0.000005s : 1: grouped_pairwise_exchange_alltoall 0.00% : 0.000004s : 1: handle_group_info 0.00% : 0.000007s : 1: inline 0.00% : 0.000007s : 1: insert-virtual-dataset 0.00% : 0.000005s : 1: interleave_parallel_branches 0.00% : 0.000004s : 1: interleave_split_concat_branches 0.00% : 0.000006s : 1: label_fine_grained_interleaved_index 0.00% : 0.000008s : 1: label_micro_interleaved_index 0.03% : 0.000561s : 1: loop_unroll 0.00% : 0.000004s : 1: merge_cast_opt 0.00% : 0.000005s : 1: micro_interleaved_order_control 0.05% : 0.000834s : 1: mutable_eliminate 0.00% : 0.000010s : 1: offloading_packed_experts 0.00% : 0.000028s : 1: opt.transform.loop_unroll_optimizer 0.00% : 0.000036s : 1: opt.transform.mutable_eliminate 0.16% : 0.002553s : 78: opt.transform.opt_a 0.01% : 0.000097s : 1: opt.transform.opt_after_cconv 0.01% : 0.000108s : 1: opt.transform.opt_after_jit_grad 0.02% : 0.000364s : 28: opt.transform.opt_b 0.01% : 0.000128s : 2: opt.transform.opt_trans_graph 0.01% : 0.000094s : 4: opt.transform.symbol_engine_opt 1.01% : 0.016517s : 1: opt_a 0.01% : 0.000226s : 1: opt_after_cconv 0.88% : 0.014331s : 1: opt_after_jit_grad 0.03% : 0.000560s : 1: opt_b 1.23% : 0.020127s : 1: optimize 0.00% : 0.000035s : 1: optimize_parallel_all_gather_comm 0.00% : 0.000015s : 1: order_py_execute_after_rewriter 0.00% : 0.000042s : 1: overlap_grad_flash_sp 0.00% : 0.000004s : 1: overlap_grad_matmul_and_grad_allreduce 0.00% : 0.000012s : 1: overlap_grad_ring_attention 0.00% : 0.000005s : 1: overlap_opt_shard_grad_in_pipeline 0.00% : 0.000004s : 1: overlap_opt_shard_in_pipeline 0.00% : 0.000006s : 1: overlap_param_gather 0.00% : 0.000005s : 1: overlap_recompute_allgather_and_fa_grad 0.00% : 0.000010s : 1: overlap_recompute_and_grad_model_parallel 0.00% : 0.000005s : 1: overlap_recompute_comm 0.00% : 0.000009s : 1: parallel-infer-symbol 0.00% : 0.000004s : 1: parallel-infer-symbol-second 0.00% : 0.000006s : 1: partial_unused_args_eliminate 0.00% : 0.000005s : 1: pipeline_parallel_scheduler 0.00% : 0.000006s : 1: pipeline_split 0.00% : 0.000045s : 1: pre_auto_parallel 0.00% : 0.000039s : 1: py_interpret_to_execute 0.00% : 0.000034s : 1: py_interpret_to_execute_after_opt_a 0.00% : 0.000004s : 1: remove_cast_before_assign_add 0.00% : 0.000078s : 1: remove_dup_value 0.45% : 0.007279s : 1: renormalize.infer 0.24% : 0.003942s : 1: renormalize.specialize 0.00% : 0.000006s : 1: reorder_send_recv_between_fp_bp 0.00% : 0.000010s : 1: rewriter_after_jit_bprop_graph 0.01% : 0.000218s : 1: rewriter_after_opt_a 0.01% : 0.000092s : 1: rewriter_before_opt_a 0.00% : 0.000005s : 1: slice_cell_reuse_recomputed_activation 0.00% : 0.000005s : 1: slice_recompute_activation 0.00% : 0.000006s : 1: split_layernorm_comm 0.00% : 0.000005s : 1: split_matmul_comm_elemetwise 0.00% : 0.000015s : 1: swap_dp_allreduce_reducescatter 0.01% : 0.000155s : 1: symbol_engine_optimizer 0.01% : 0.000165s : 1: tuple_transform 86.96% : 1.419655s : 1: type_inference . [hook] pytest_runtest_teardown:test_matmul_qkv_bf16_1[False-mstype0-k_n_shape2-1024] tests/st/infer/ops/test_internal_ops/test_matmul_split.py::test_matmul_qkv_bf16_1[False-mstype0-k_n_shape2-1024],max_mem:260.0M [WARNING] ME(168319:281472866918192,MainProcess):2026-01-29-17:43:07.231.523 [mindspore/context.py:1334] For 'context.set_context', the parameter 'device_target' will be deprecated and removed in a future version. Please use the api mindspore.set_device() instead. TotalTime = 1.5951, [21] [bootstrap]: 0.00074224 [type_inference]: 1.35262 [event_method]: 1.992e-05 [auto_monad]: 0.00034586 [graph_reusing]: 5.72999e-06 [inline]: 3.10002e-06 [add_attr]: 0.0725202, [1] [add_attr_with_inline]: 0.0724999, [1] [Cycle 1]: 8.276e-05, [2] [tag_attr]: 2.508e-05 [meta_addattr_fg_expand]: 4.14002e-06 [parallel-infer-symbol]: 5.55001e-06 [pre_auto_parallel]: 4.102e-05 [insert-virtual-dataset]: 2.96001e-06 [parallel-infer-symbol-second]: 7.10017e-07 [dataset_repeat_opt]: 2.09e-06 [pipeline_split]: 1.92001e-06 [optimize]: 0.10055, [53] [py_interpret_to_execute]: 3.891e-05 [rewriter_before_opt_a]: 9.636e-05 [opt_a]: 0.096639, [2] [Cycle 1]: 0.0949387, [45] [expand_dump_flag]: 3.46999e-06 [switch_simplify]: 3.832e-05 [loop_unroll]: 2.015e-05 [a_1]: 0.0009294 [with_stream_mark]: 3.981e-05 [recompute_prepare]: 2.732e-05 [updatestate_depend_eliminate]: 1.289e-05 [updatestate_assign_eliminate]: 1.045e-05 [updatestate_loads_eliminate]: 1.179e-05 [parameter_eliminate]: 3.09001e-06 [a_2]: 0.0002584 [accelerated_algorithm]: 4.873e-05 [shard]: 2.34999e-06 [meta_shard_fg_expand]: 4.52998e-06 [shard_inline]: 1.783e-05 [merge_send_recv]: 1.501e-05 [auto_parallel]: 1.566e-05 [parallel]: 4.477e-05 [flash_sp]: 1.516e-05 [merge_comm]: 8.62998e-06 [allreduce_fusion]: 8.40001e-06 [matmul_add_comm_reduction]: 1.8e-05 [allreduce_slice_to_reducescatter]: 1.22999e-06 [virtual_shard_identity]: 2.535e-05 [virtual_dataset]: 1.836e-05 [get_grad_eliminate_]: 1.947e-05 [virtual_output]: 1.77e-05 [merge_forward]: 8.38999e-06 [cell_reuse_recompute_pass]: 2.84999e-06 [offload_activation]: 1.955e-05 [cell_reuse_handle_not_recompute_node_pass]: 3.915e-05 [merge_recompute_call_nodes]: 1.79998e-06 [before_grad]: 3.535e-05 [set_forward_comm_id_for_comm_node_pass]: 9.07001e-06 [meta_fg_expand]: 6.69001e-06 [flash_sp_send_recv_attached]: 7.49002e-06 [receive_attached]: 2.41e-06 [after_resolve]: 2.764e-05 [a_after_grad]: 3.097e-05 [renormalize]: 0.0910482 [add_forward_monad_depend]: 1.86e-05 [auto_monad_grad]: 3.3e-06 [auto_monad_eliminator]: 6.546e-05 [cse]: 0.00137636 [a_3]: 0.00022454 [Cycle 2]: 0.00168191, [45] [expand_dump_flag]: 3.53e-06 [switch_simplify]: 2.022e-05 [loop_unroll]: 1.745e-05 [a_1]: 0.00046792 [with_stream_mark]: 3.501e-05 [recompute_prepare]: 1.722e-05 [updatestate_depend_eliminate]: 1.09e-05 [updatestate_assign_eliminate]: 8.07e-06 [updatestate_loads_eliminate]: 1.063e-05 [parameter_eliminate]: 2.74999e-06 [a_2]: 0.00022064 [accelerated_algorithm]: 2.407e-05 [shard]: 2.78003e-06 [meta_shard_fg_expand]: 5.29e-06 [shard_inline]: 1.557e-05 [merge_send_recv]: 1.502e-05 [auto_parallel]: 1.56e-05 [parallel]: 1.049e-05 [flash_sp]: 4.82e-06 [merge_comm]: 8.86002e-06 [allreduce_fusion]: 9.22001e-06 [matmul_add_comm_reduction]: 1.78e-05 [allreduce_slice_to_reducescatter]: 9.39996e-07 [virtual_shard_identity]: 1.893e-05 [virtual_dataset]: 1.65e-05 [get_grad_eliminate_]: 1.567e-05 [virtual_output]: 1.515e-05 [merge_forward]: 9.24998e-06 [cell_reuse_recompute_pass]: 3.48999e-06 [offload_activation]: 1.897e-05 [cell_reuse_handle_not_recompute_node_pass]: 2.987e-05 [merge_recompute_call_nodes]: 1.77999e-06 [before_grad]: 2.557e-05 [set_forward_comm_id_for_comm_node_pass]: 8.80001e-06 [meta_fg_expand]: 6.34999e-06 [flash_sp_send_recv_attached]: 1.81003e-06 [receive_attached]: 2.90998e-06 [after_resolve]: 2.362e-05 [a_after_grad]: 2.617e-05 [renormalize]: 1.00001e-07 [add_forward_monad_depend]: 3.69002e-06 [auto_monad_grad]: 3.2e-06 [auto_monad_eliminator]: 5.572e-05 [cse]: 6.207e-05 [a_3]: 0.00014088 [py_interpret_to_execute_after_opt_a]: 3.184e-05 [slice_cell_reuse_recomputed_activation]: 2.46e-06 [rewriter_after_opt_a]: 0.00022216 [convert_after_rewriter]: 1.607e-05 [order_py_execute_after_rewriter]: 1.139e-05 [mutable_eliminate]: 0.00086812 [opt_b]: 0.00056987, [1] [Cycle 1]: 0.00056026, [7] [b_1]: 0.00037158 [b_2]: 2e-05 [updatestate_depend_eliminate]: 2.089e-05 [updatestate_assign_eliminate]: 8.10999e-06 [updatestate_loads_eliminate]: 1.189e-05 [renormalize]: 7.40023e-07 [cse]: 7.648e-05 [optimize_parallel_all_gather_comm]: 3.828e-05 [overlap_param_gather]: 3.20998e-06 [cconv]: 4.119e-05 [loop_unroll]: 0.00060733 [opt_after_cconv]: 0.00025013, [1] [Cycle 1]: 0.00024138, [7] [c_1]: 0.00010389 [parameter_eliminate]: 5.07999e-06 [updatestate_depend_eliminate]: 1.616e-05 [updatestate_assign_eliminate]: 7.61999e-06 [updatestate_loads_eliminate]: 9.84999e-06 [cse]: 5.72e-05 [renormalize]: 6.80011e-07 [remove_dup_value]: 7.479e-05 [tuple_transform]: 0.00017243, [1] [Cycle 1]: 0.00016537, [4] [d_1]: 0.0001178 [none_parameter_eliminate]: 2.56e-06 [renormalize]: 2.30008e-07 [switch_simplify]: 1.823e-05 [partial_unused_args_eliminate]: 2.33002e-06 [add_recomputation]: 0.00012522 [cse_after_recomputation]: 6.615e-05, [1] [Cycle 1]: 5.99e-05, [1] [cse]: 5.096e-05 [environ_conv]: 2.157e-05 [swap_dp_allreduce_reducescatter]: 1.342e-05 [bias_add_comm_swap]: 3.76001e-06 [label_micro_interleaved_index]: 7.36999e-06 [label_fine_grained_interleaved_index]: 2.67001e-06 [merge_cast_opt]: 1.52999e-06 [slice_recompute_activation]: 2.11003e-06 [micro_interleaved_order_control]: 2.36e-06 [assign_add_opt]: 1.30001e-06 [ForceFp32Comm]: 1.14998e-06 [remove_cast_before_assign_add]: 1.20999e-06 [full_micro_interleaved_order_control]: 2.56998e-06 [reorder_send_recv_between_fp_bp]: 2.57001e-06 [comm_op_add_attrs]: 1.12e-06 [add_comm_op_reuse_tag]: 1.04003e-06 [interleave_split_concat_branches]: 1.24e-06 [interleave_parallel_branches]: 1.18001e-06 [overlap_opt_shard_in_pipeline]: 2.98998e-06 [overlap_opt_shard_grad_in_pipeline]: 2.05002e-06 [control_data_broadcast_order]: 3.011e-05 [grouped_pairwise_exchange_alltoall]: 1.96003e-06 [offloading_packed_experts]: 7.43e-06 [overlap_recompute_and_grad_model_parallel]: 7.82e-06 [overlap_grad_matmul_and_grad_allreduce]: 1.25999e-06 [overlap_recompute_allgather_and_fa_grad]: 1.84998e-06 [overlap_recompute_comm]: 2.37999e-06 [overlap_grad_ring_attention]: 8.62e-06 [overlap_grad_flash_sp]: 4.236e-05 [begin_end_overlap_inline]: 5.50004e-07 [split_matmul_comm_elemetwise]: 2.71e-06 [split_layernorm_comm]: 2.14999e-06 [handle_group_info]: 1.05001e-06 [symbol_engine_optimizer]: 0.00017763, [1] [Cycle 1]: 0.00016919, [6] [build]: 1.865e-05 [elim_shapecalc]: 2.69e-05 [elim_not_effective]: 3.243e-05 [opt_reshape]: 2.283e-05 [fold_const_symbol]: 2.628e-05 [renormalize]: 2.00002e-07 [detach_backward]: 2.98e-06 [pipeline_parallel_scheduler]: 1.77999e-06 [auto_monad_reorder]: 5.456e-05 [get_jit_bprop_graph]: 2.22001e-06 [rewriter_after_jit_bprop_graph]: 6.64999e-06 [opt_after_jit_grad]: 0.0677617 [validate]: 0.00014078 Sums bootstrap : 0.000742s : 0.05% type_inference : 1.352625s : 88.92% event_method : 0.000020s : 0.00% auto_monad : 0.000346s : 0.02% graph_reusing : 0.000006s : 0.00% inline : 0.000003s : 0.00% add_attr.add_attr_with_inline.tag_attr : 0.000025s : 0.00% add_attr.add_attr_with_inline.meta_addattr_fg_expand : 0.000004s : 0.00% parallel-infer-symbol : 0.000006s : 0.00% pre_auto_parallel : 0.000041s : 0.00% insert-virtual-dataset : 0.000003s : 0.00% parallel-infer-symbol-second : 0.000001s : 0.00% dataset_repeat_opt : 0.000002s : 0.00% pipeline_split : 0.000002s : 0.00% optimize.py_interpret_to_execute : 0.000039s : 0.00% optimize.rewriter_before_opt_a : 0.000096s : 0.01% optimize.opt_a.expand_dump_flag : 0.000007s : 0.00% optimize.opt_a.switch_simplify : 0.000059s : 0.00% optimize.opt_a.loop_unroll : 0.000038s : 0.00% optimize.opt_a.a_1 : 0.001397s : 0.09% optimize.opt_a.with_stream_mark : 0.000075s : 0.00% optimize.opt_a.recompute_prepare : 0.000045s : 0.00% optimize.opt_a.updatestate_depend_eliminate : 0.000024s : 0.00% optimize.opt_a.updatestate_assign_eliminate : 0.000019s : 0.00% optimize.opt_a.updatestate_loads_eliminate : 0.000022s : 0.00% optimize.opt_a.parameter_eliminate : 0.000006s : 0.00% optimize.opt_a.a_2 : 0.000479s : 0.03% optimize.opt_a.accelerated_algorithm : 0.000073s : 0.00% optimize.opt_a.shard : 0.000005s : 0.00% optimize.opt_a.meta_shard_fg_expand : 0.000010s : 0.00% optimize.opt_a.shard_inline : 0.000033s : 0.00% optimize.opt_a.merge_send_recv : 0.000030s : 0.00% optimize.opt_a.auto_parallel : 0.000031s : 0.00% optimize.opt_a.parallel : 0.000055s : 0.00% optimize.opt_a.flash_sp : 0.000020s : 0.00% optimize.opt_a.merge_comm : 0.000017s : 0.00% optimize.opt_a.allreduce_fusion : 0.000018s : 0.00% optimize.opt_a.matmul_add_comm_reduction : 0.000036s : 0.00% optimize.opt_a.allreduce_slice_to_reducescatter : 0.000002s : 0.00% optimize.opt_a.virtual_shard_identity : 0.000044s : 0.00% optimize.opt_a.virtual_dataset : 0.000035s : 0.00% optimize.opt_a.get_grad_eliminate_ : 0.000035s : 0.00% optimize.opt_a.virtual_output : 0.000033s : 0.00% optimize.opt_a.merge_forward : 0.000018s : 0.00% optimize.opt_a.cell_reuse_recompute_pass : 0.000006s : 0.00% optimize.opt_a.offload_activation : 0.000039s : 0.00% optimize.opt_a.cell_reuse_handle_not_recompute_node_pass : 0.000069s : 0.00% optimize.opt_a.merge_recompute_call_nodes : 0.000004s : 0.00% optimize.opt_a.before_grad : 0.000061s : 0.00% optimize.opt_a.set_forward_comm_id_for_comm_node_pass : 0.000018s : 0.00% optimize.opt_a.meta_fg_expand : 0.000013s : 0.00% optimize.opt_a.flash_sp_send_recv_attached : 0.000009s : 0.00% optimize.opt_a.receive_attached : 0.000005s : 0.00% optimize.opt_a.after_resolve : 0.000051s : 0.00% optimize.opt_a.a_after_grad : 0.000057s : 0.00% optimize.opt_a.renormalize : 0.091048s : 5.99% optimize.opt_a.add_forward_monad_depend : 0.000022s : 0.00% optimize.opt_a.auto_monad_grad : 0.000006s : 0.00% optimize.opt_a.auto_monad_eliminator : 0.000121s : 0.01% optimize.opt_a.cse : 0.001438s : 0.09% optimize.opt_a.a_3 : 0.000365s : 0.02% optimize.py_interpret_to_execute_after_opt_a : 0.000032s : 0.00% optimize.slice_cell_reuse_recomputed_activation : 0.000002s : 0.00% optimize.rewriter_after_opt_a : 0.000222s : 0.01% optimize.convert_after_rewriter : 0.000016s : 0.00% optimize.order_py_execute_after_rewriter : 0.000011s : 0.00% optimize.mutable_eliminate : 0.000868s : 0.06% optimize.opt_b.b_1 : 0.000372s : 0.02% optimize.opt_b.b_2 : 0.000020s : 0.00% optimize.opt_b.updatestate_depend_eliminate : 0.000021s : 0.00% optimize.opt_b.updatestate_assign_eliminate : 0.000008s : 0.00% optimize.opt_b.updatestate_loads_eliminate : 0.000012s : 0.00% optimize.opt_b.renormalize : 0.000001s : 0.00% optimize.opt_b.cse : 0.000076s : 0.01% optimize.optimize_parallel_all_gather_comm : 0.000038s : 0.00% optimize.overlap_param_gather : 0.000003s : 0.00% optimize.cconv : 0.000041s : 0.00% optimize.loop_unroll : 0.000607s : 0.04% optimize.opt_after_cconv.c_1 : 0.000104s : 0.01% optimize.opt_after_cconv.parameter_eliminate : 0.000005s : 0.00% optimize.opt_after_cconv.updatestate_depend_eliminate : 0.000016s : 0.00% optimize.opt_after_cconv.updatestate_assign_eliminate : 0.000008s : 0.00% optimize.opt_after_cconv.updatestate_loads_eliminate : 0.000010s : 0.00% optimize.opt_after_cconv.cse : 0.000057s : 0.00% optimize.opt_after_cconv.renormalize : 0.000001s : 0.00% optimize.remove_dup_value : 0.000075s : 0.00% optimize.tuple_transform.d_1 : 0.000118s : 0.01% optimize.tuple_transform.none_parameter_eliminate : 0.000003s : 0.00% optimize.tuple_transform.renormalize : 0.000000s : 0.00% optimize.tuple_transform.switch_simplify : 0.000018s : 0.00% optimize.partial_unused_args_eliminate : 0.000002s : 0.00% optimize.add_recomputation : 0.000125s : 0.01% optimize.cse_after_recomputation.cse : 0.000051s : 0.00% optimize.environ_conv : 0.000022s : 0.00% optimize.swap_dp_allreduce_reducescatter : 0.000013s : 0.00% optimize.bias_add_comm_swap : 0.000004s : 0.00% optimize.label_micro_interleaved_index : 0.000007s : 0.00% optimize.label_fine_grained_interleaved_index : 0.000003s : 0.00% optimize.merge_cast_opt : 0.000002s : 0.00% optimize.slice_recompute_activation : 0.000002s : 0.00% optimize.micro_interleaved_order_control : 0.000002s : 0.00% optimize.assign_add_opt : 0.000001s : 0.00% optimize.ForceFp32Comm : 0.000001s : 0.00% optimize.remove_cast_before_assign_add : 0.000001s : 0.00% optimize.full_micro_interleaved_order_control : 0.000003s : 0.00% optimize.reorder_send_recv_between_fp_bp : 0.000003s : 0.00% optimize.comm_op_add_attrs : 0.000001s : 0.00% optimize.add_comm_op_reuse_tag : 0.000001s : 0.00% optimize.interleave_split_concat_branches : 0.000001s : 0.00% optimize.interleave_parallel_branches : 0.000001s : 0.00% optimize.overlap_opt_shard_in_pipeline : 0.000003s : 0.00% optimize.overlap_opt_shard_grad_in_pipeline : 0.000002s : 0.00% optimize.control_data_broadcast_order : 0.000030s : 0.00% optimize.grouped_pairwise_exchange_alltoall : 0.000002s : 0.00% optimize.offloading_packed_experts : 0.000007s : 0.00% optimize.overlap_recompute_and_grad_model_parallel : 0.000008s : 0.00% optimize.overlap_grad_matmul_and_grad_allreduce : 0.000001s : 0.00% optimize.overlap_recompute_allgather_and_fa_grad : 0.000002s : 0.00% optimize.overlap_recompute_comm : 0.000002s : 0.00% optimize.overlap_grad_ring_attention : 0.000009s : 0.00% optimize.overlap_grad_flash_sp : 0.000042s : 0.00% optimize.begin_end_overlap_inline : 0.000001s : 0.00% optimize.split_matmul_comm_elemetwise : 0.000003s : 0.00% optimize.split_layernorm_comm : 0.000002s : 0.00% optimize.handle_group_info : 0.000001s : 0.00% optimize.symbol_engine_optimizer.build : 0.000019s : 0.00% optimize.symbol_engine_optimizer.elim_shapecalc : 0.000027s : 0.00% optimize.symbol_engine_optimizer.elim_not_effective : 0.000032s : 0.00% optimize.symbol_engine_optimizer.opt_reshape : 0.000023s : 0.00% optimize.symbol_engine_optimizer.fold_const_symbol : 0.000026s : 0.00% optimize.symbol_engine_optimizer.renormalize : 0.000000s : 0.00% detach_backward : 0.000003s : 0.00% pipeline_parallel_scheduler : 0.000002s : 0.00% auto_monad_reorder : 0.000055s : 0.00% get_jit_bprop_graph : 0.000002s : 0.00% rewriter_after_jit_bprop_graph : 0.000007s : 0.00% opt_after_jit_grad : 0.067762s : 4.45% validate : 0.000141s : 0.01% Time group info: ------[substitution.] 0.000313 96 3.68% : 0.000012s : 2: substitution.depend_value_elim 1.32% : 0.000004s : 8: substitution.elim_not_effective 1.13% : 0.000004s : 8: substitution.fold_const_symbol 3.89% : 0.000012s : 13: substitution.graph_param_transform 44.69% : 0.000140s : 1: substitution.inline 3.01% : 0.000009s : 16: substitution.j_node_and_user_rematch 8.26% : 0.000026s : 2: substitution.less_batch_normalization 2.04% : 0.000006s : 6: substitution.load_eliminater 0.86% : 0.000003s : 2: substitution.opt_reshape 4.00% : 0.000013s : 16: substitution.remove_not_recompute_node 2.63% : 0.000008s : 4: substitution.replace_old_param 9.79% : 0.000031s : 4: substitution.reshape_eliminate 8.32% : 0.000026s : 6: substitution.updatestate_pure_node_eliminater 6.39% : 0.000020s : 8: substitution.updatestate_useless_node_eliminater ------[type_inference.] 1.352517 2 99.32% : 1.343293s : 1: type_inference.infer 0.68% : 0.009223s : 1: type_inference.specialize ------[replace.] 0.000021 1 100.00% : 0.000021s : 1: replace.inline ------[match.] 0.000139 1 100.00% : 0.000139s : 1: match.inline ------[predicate.] 0.000463 3002 0.79% : 0.000004s : 27: predicate.accumulaten_eliminater 2.34% : 0.000011s : 13: predicate.ad_related_special_op_eliminate 0.72% : 0.000003s : 26: predicate.addn_check_dump 0.81% : 0.000004s : 27: predicate.addn_zero_filter 0.77% : 0.000004s : 27: predicate.adjust_all_reduce_mul_add 2.10% : 0.000010s : 53: predicate.arithmetic_simplify 0.83% : 0.000004s : 27: predicate.cast_eliminate 0.81% : 0.000004s : 26: predicate.check_bprop_eliminate 0.78% : 0.000004s : 26: predicate.compare_switch_simplify 0.25% : 0.000001s : 13: predicate.const_output_eliminate 0.82% : 0.000004s : 26: predicate.depend_value_elim 0.82% : 0.000004s : 27: predicate.dict_get_item_const_eliminator 0.98% : 0.000005s : 27: predicate.dict_get_item_eliminator 0.79% : 0.000004s : 27: predicate.dict_set_item_eliminator 1.65% : 0.000008s : 26: predicate.dumpgradient_eliminate 0.28% : 0.000001s : 13: predicate.elim_not_effective 0.58% : 0.000003s : 13: predicate.elim_shapecalc_of_broadcastargs 1.19% : 0.000006s : 40: predicate.environ_add_const_eliminate 1.06% : 0.000005s : 40: predicate.environ_get_add_eliminate 1.06% : 0.000005s : 40: predicate.environ_get_depend_swap 1.84% : 0.000009s : 66: predicate.environ_get_eliminate 1.07% : 0.000005s : 40: predicate.environ_get_set_eliminate 0.76% : 0.000004s : 28: predicate.exchange_switch_depend_value 1.33% : 0.000006s : 28: predicate.float_depend_g_call 0.75% : 0.000003s : 26: predicate.float_environ_get_switch 1.09% : 0.000005s : 39: predicate.float_tuple_getitem_switch 0.22% : 0.000001s : 13: predicate.fold_const_symbol 0.80% : 0.000004s : 26: predicate.get_grad_eliminate 0.27% : 0.000001s : 13: predicate.graph_param_transform 0.73% : 0.000003s : 26: predicate.incorporate_call 0.69% : 0.000003s : 26: predicate.incorporate_call_switch 5.33% : 0.000025s : 133: predicate.inline 1.06% : 0.000005s : 26: predicate.inline_without_move 0.42% : 0.000002s : 26: predicate.j_node_and_user_rematch 1.44% : 0.000007s : 27: predicate.less_batch_normalization 1.74% : 0.000008s : 53: predicate.list_to_tuple_eliminator_ 2.34% : 0.000011s : 80: predicate.load_eliminater 1.13% : 0.000005s : 13: predicate.loop_unroll_after_grad 0.94% : 0.000004s : 31: predicate.loop_unroll_before_grad 1.64% : 0.000008s : 53: predicate.make_slice_get_slice_eliminator 0.79% : 0.000004s : 26: predicate.merge_addn 0.82% : 0.000004s : 26: predicate.micro_step_allgather_replace 0.78% : 0.000004s : 26: predicate.mini_step_allgather_replace 0.71% : 0.000003s : 27: predicate.minmaximum_grad 1.34% : 0.000006s : 13: predicate.mutable_eliminate 0.54% : 0.000002s : 13: predicate.opt_reshape 0.45% : 0.000002s : 13: predicate.parallel_virtual_node 1.36% : 0.000006s : 28: predicate.partial_defer_inline 1.22% : 0.000006s : 40: predicate.partial_eliminate 0.80% : 0.000004s : 27: predicate.print_const_string_wrapper 0.84% : 0.000004s : 26: predicate.reduce_all_const_elim 1.04% : 0.000005s : 27: predicate.reduce_eliminate 2.16% : 0.000010s : 80: predicate.redundant_stop_gradient_eliminater 0.53% : 0.000002s : 26: predicate.remove_not_recompute_node 1.36% : 0.000006s : 53: predicate.replace_applicator 0.55% : 0.000003s : 26: predicate.replace_old_param 0.33% : 0.000002s : 13: predicate.reset_defer_inline 0.90% : 0.000004s : 27: predicate.reshape_eliminate 1.11% : 0.000005s : 26: predicate.row_tensor_add_zeros_like 0.52% : 0.000002s : 13: predicate.row_tensor_eliminate 1.09% : 0.000005s : 26: predicate.same_eliminate 0.55% : 0.000003s : 26: predicate.set_cell_output_no_recompute 0.99% : 0.000005s : 26: predicate.shard_identity_eliminate 0.95% : 0.000004s : 26: predicate.special_op_eliminate 0.90% : 0.000004s : 26: predicate.specialize_transform 1.33% : 0.000006s : 26: predicate.split_environ_get_set_with_tuple_value 1.01% : 0.000005s : 26: predicate.stack_unstack_eliminate 0.40% : 0.000002s : 13: predicate.switch_call_monad_eliminater 0.83% : 0.000004s : 28: predicate.switch_defer_inline 1.55% : 0.000007s : 54: predicate.switch_layer_defer_inline 3.36% : 0.000016s : 98: predicate.switch_simplify 0.90% : 0.000004s : 27: predicate.tile_eliminate 0.87% : 0.000004s : 27: predicate.transpose_eliminate 1.64% : 0.000008s : 53: predicate.tuple_list_convert_item_index_to_positive 1.68% : 0.000008s : 53: predicate.tuple_list_get_item_const_eliminator 1.52% : 0.000007s : 53: predicate.tuple_list_get_item_depend_reorder 3.08% : 0.000014s : 79: predicate.tuple_list_get_item_eliminator 1.54% : 0.000007s : 53: predicate.tuple_list_get_set_item_eliminator 2.52% : 0.000012s : 79: predicate.tuple_list_set_item_eliminator 1.57% : 0.000007s : 53: predicate.tuple_to_list_eliminator_ 2.13% : 0.000010s : 80: predicate.updatestate_pure_node_eliminater 3.40% : 0.000016s : 106: predicate.updatestate_useless_node_eliminater 0.41% : 0.000002s : 13: predicate.value_based_eliminate 0.84% : 0.000004s : 26: predicate.virtual_dataset_eliminate 0.85% : 0.000004s : 26: predicate.virtual_output_eliminate 0.41% : 0.000002s : 13: predicate.virtual_view_grad_eliminate 0.51% : 0.000002s : 13: predicate.zero_like_fill_zero ------[func_graph_cloner_run.] 0.100068 58 96.23% : 0.096290s : 55: func_graph_cloner_run.FuncGraphClonerGraph 3.77% : 0.003777s : 3: func_graph_cloner_run.FuncGraphSpecializer ------[meta_graph.] 0.000000 0 ------[manager.] 0.000000 0 ------[pynative] 0.000000 0 ------[others.] 1.862525 192 0.00% : 0.000004s : 1: ForceFp32Comm 3.89% : 0.072528s : 1: add_attr 3.89% : 0.072504s : 1: add_attr_with_inline 0.00% : 0.000004s : 1: add_comm_op_reuse_tag 0.01% : 0.000133s : 1: add_recomputation 0.00% : 0.000004s : 1: assign_add_opt 0.02% : 0.000357s : 1: auto_monad 0.00% : 0.000064s : 1: auto_monad_reorder 0.00% : 0.000004s : 1: begin_end_overlap_inline 0.00% : 0.000007s : 1: bias_add_comm_swap 0.04% : 0.000811s : 1: bootstrap 0.00% : 0.000045s : 1: cconv 0.00% : 0.000005s : 1: comm_op_add_attrs 0.00% : 0.000034s : 1: control_data_broadcast_order 0.00% : 0.000021s : 1: convert_after_rewriter 0.00% : 0.000070s : 1: cse_after_recomputation 0.00% : 0.000005s : 1: dataset_repeat_opt 0.00% : 0.000006s : 1: detach_backward 0.00% : 0.000027s : 1: environ_conv 0.00% : 0.000027s : 1: event_method 0.00% : 0.000006s : 1: full_micro_interleaved_order_control 0.00% : 0.000008s : 1: get_jit_bprop_graph 0.00% : 0.000010s : 1: graph_reusing 0.00% : 0.000005s : 1: grouped_pairwise_exchange_alltoall 0.00% : 0.000004s : 1: handle_group_info 0.00% : 0.000006s : 1: inline 0.00% : 0.000008s : 1: insert-virtual-dataset 0.00% : 0.000005s : 1: interleave_parallel_branches 0.00% : 0.000004s : 1: interleave_split_concat_branches 0.00% : 0.000006s : 1: label_fine_grained_interleaved_index 0.00% : 0.000010s : 1: label_micro_interleaved_index 0.03% : 0.000620s : 1: loop_unroll 0.00% : 0.000004s : 1: merge_cast_opt 0.00% : 0.000005s : 1: micro_interleaved_order_control 0.05% : 0.000887s : 1: mutable_eliminate 0.00% : 0.000011s : 1: offloading_packed_experts 0.00% : 0.000034s : 1: opt.transform.loop_unroll_optimizer 0.00% : 0.000043s : 1: opt.transform.mutable_eliminate 0.15% : 0.002744s : 78: opt.transform.opt_a 0.01% : 0.000102s : 1: opt.transform.opt_after_cconv 0.01% : 0.000104s : 1: opt.transform.opt_after_jit_grad 0.02% : 0.000355s : 28: opt.transform.opt_b 0.01% : 0.000133s : 2: opt.transform.opt_trans_graph 0.01% : 0.000103s : 4: opt.transform.symbol_engine_opt 5.19% : 0.096643s : 1: opt_a 0.01% : 0.000254s : 1: opt_after_cconv 3.64% : 0.067786s : 1: opt_after_jit_grad 0.03% : 0.000574s : 1: opt_b 5.40% : 0.100557s : 1: optimize 0.00% : 0.000043s : 1: optimize_parallel_all_gather_comm 0.00% : 0.000015s : 1: order_py_execute_after_rewriter 0.00% : 0.000047s : 1: overlap_grad_flash_sp 0.00% : 0.000004s : 1: overlap_grad_matmul_and_grad_allreduce 0.00% : 0.000013s : 1: overlap_grad_ring_attention 0.00% : 0.000005s : 1: overlap_opt_shard_grad_in_pipeline 0.00% : 0.000007s : 1: overlap_opt_shard_in_pipeline 0.00% : 0.000008s : 1: overlap_param_gather 0.00% : 0.000006s : 1: overlap_recompute_allgather_and_fa_grad 0.00% : 0.000011s : 1: overlap_recompute_and_grad_model_parallel 0.00% : 0.000005s : 1: overlap_recompute_comm 0.00% : 0.000010s : 1: parallel-infer-symbol 0.00% : 0.000004s : 1: parallel-infer-symbol-second 0.00% : 0.000006s : 1: partial_unused_args_eliminate 0.00% : 0.000005s : 1: pipeline_parallel_scheduler 0.00% : 0.000006s : 1: pipeline_split 0.00% : 0.000047s : 1: pre_auto_parallel 0.00% : 0.000044s : 1: py_interpret_to_execute 0.00% : 0.000037s : 1: py_interpret_to_execute_after_opt_a 0.00% : 0.000004s : 1: remove_cast_before_assign_add 0.00% : 0.000080s : 1: remove_dup_value 4.66% : 0.086777s : 1: renormalize.infer 0.23% : 0.004250s : 1: renormalize.specialize 0.00% : 0.000006s : 1: reorder_send_recv_between_fp_bp 0.00% : 0.000010s : 1: rewriter_after_jit_bprop_graph 0.01% : 0.000232s : 1: rewriter_after_opt_a 0.01% : 0.000103s : 1: rewriter_before_opt_a 0.00% : 0.000005s : 1: slice_cell_reuse_recomputed_activation 0.00% : 0.000005s : 1: slice_recompute_activation 0.00% : 0.000006s : 1: split_layernorm_comm 0.00% : 0.000005s : 1: split_matmul_comm_elemetwise 0.00% : 0.000017s : 1: swap_dp_allreduce_reducescatter 0.01% : 0.000181s : 1: symbol_engine_optimizer 0.01% : 0.000176s : 1: tuple_transform 72.62% : 1.352655s : 1: type_inference . [hook] pytest_runtest_teardown:test_matmul_qkv_bf16_1[False-mstype0-k_n_shape3-256] tests/st/infer/ops/test_internal_ops/test_matmul_split.py::test_matmul_qkv_bf16_1[False-mstype0-k_n_shape3-256],max_mem:260.0M [WARNING] ME(168319:281472866918192,MainProcess):2026-01-29-17:43:42.180.13 [mindspore/context.py:1334] For 'context.set_context', the parameter 'device_target' will be deprecated and removed in a future version. Please use the api mindspore.set_device() instead. TotalTime = 1.18291, [21] [bootstrap]: 0.00080196 [type_inference]: 1.04612 [event_method]: 1.943e-05 [auto_monad]: 0.00037094 [graph_reusing]: 6.65998e-06 [inline]: 3.43e-06 [add_attr]: 0.00736696, [1] [add_attr_with_inline]: 0.00734683, [1] [Cycle 1]: 7.654e-05, [2] [tag_attr]: 2.311e-05 [meta_addattr_fg_expand]: 4.27e-06 [parallel-infer-symbol]: 4.89e-06 [pre_auto_parallel]: 3.918e-05 [insert-virtual-dataset]: 2.68e-06 [parallel-infer-symbol-second]: 1.02e-06 [dataset_repeat_opt]: 2.03002e-06 [pipeline_split]: 1.70001e-06 [optimize]: 0.0751337, [53] [py_interpret_to_execute]: 3.473e-05 [rewriter_before_opt_a]: 9.435e-05 [opt_a]: 0.0716715, [2] [Cycle 1]: 0.0701435, [45] [expand_dump_flag]: 3.25998e-06 [switch_simplify]: 3.595e-05 [loop_unroll]: 1.96e-05 [a_1]: 0.057684 [with_stream_mark]: 4.303e-05 [recompute_prepare]: 2.877e-05 [updatestate_depend_eliminate]: 1.313e-05 [updatestate_assign_eliminate]: 7.82e-06 [updatestate_loads_eliminate]: 1.161e-05 [parameter_eliminate]: 2.83e-06 [a_2]: 0.00026187 [accelerated_algorithm]: 4.867e-05 [shard]: 3.21999e-06 [meta_shard_fg_expand]: 5.81e-06 [shard_inline]: 1.892e-05 [merge_send_recv]: 1.603e-05 [auto_parallel]: 1.527e-05 [parallel]: 3.801e-05 [flash_sp]: 1.489e-05 [merge_comm]: 8.30999e-06 [allreduce_fusion]: 8.95001e-06 [matmul_add_comm_reduction]: 1.765e-05 [allreduce_slice_to_reducescatter]: 1.27999e-06 [virtual_shard_identity]: 2.318e-05 [virtual_dataset]: 1.865e-05 [get_grad_eliminate_]: 2.094e-05 [virtual_output]: 1.937e-05 [merge_forward]: 8.82e-06 [cell_reuse_recompute_pass]: 3.18998e-06 [offload_activation]: 1.987e-05 [cell_reuse_handle_not_recompute_node_pass]: 3.558e-05 [merge_recompute_call_nodes]: 1.66002e-06 [before_grad]: 3.223e-05 [set_forward_comm_id_for_comm_node_pass]: 9.94001e-06 [meta_fg_expand]: 7.88999e-06 [flash_sp_send_recv_attached]: 6.85002e-06 [receive_attached]: 2.71e-06 [after_resolve]: 3.24e-05 [a_after_grad]: 3.076e-05 [renormalize]: 0.0102551 [add_forward_monad_depend]: 1.91e-05 [auto_monad_grad]: 2.95002e-06 [auto_monad_eliminator]: 6.172e-05 [cse]: 0.00072098 [a_3]: 0.00013642 [Cycle 2]: 0.00151254, [45] [expand_dump_flag]: 2.68e-06 [switch_simplify]: 1.887e-05 [loop_unroll]: 1.617e-05 [a_1]: 0.00045537 [with_stream_mark]: 2.415e-05 [recompute_prepare]: 1.553e-05 [updatestate_depend_eliminate]: 1.08e-05 [updatestate_assign_eliminate]: 8.10999e-06 [updatestate_loads_eliminate]: 1.074e-05 [parameter_eliminate]: 2.26e-06 [a_2]: 0.00021534 [accelerated_algorithm]: 1.999e-05 [shard]: 2.87002e-06 [meta_shard_fg_expand]: 3.98999e-06 [shard_inline]: 1.542e-05 [merge_send_recv]: 1.4e-05 [auto_parallel]: 1.521e-05 [parallel]: 1.015e-05 [flash_sp]: 4.43001e-06 [merge_comm]: 8.31002e-06 [allreduce_fusion]: 8.70999e-06 [matmul_add_comm_reduction]: 1.624e-05 [allreduce_slice_to_reducescatter]: 8.59989e-07 [virtual_shard_identity]: 1.697e-05 [virtual_dataset]: 1.504e-05 [get_grad_eliminate_]: 1.485e-05 [virtual_output]: 1.449e-05 [merge_forward]: 8.87e-06 [cell_reuse_recompute_pass]: 3.38e-06 [offload_activation]: 1.699e-05 [cell_reuse_handle_not_recompute_node_pass]: 2.796e-05 [merge_recompute_call_nodes]: 1.60001e-06 [before_grad]: 2.512e-05 [set_forward_comm_id_for_comm_node_pass]: 8.57e-06 [meta_fg_expand]: 6.61e-06 [flash_sp_send_recv_attached]: 1.95001e-06 [receive_attached]: 2.53998e-06 [after_resolve]: 2.12e-05 [a_after_grad]: 2.478e-05 [renormalize]: 8.9989e-08 [add_forward_monad_depend]: 1.86e-06 [auto_monad_grad]: 1.55001e-06 [auto_monad_eliminator]: 3.362e-05 [cse]: 4.507e-05 [a_3]: 9.944e-05 [py_interpret_to_execute_after_opt_a]: 2.66e-05 [slice_cell_reuse_recomputed_activation]: 2.01e-06 [rewriter_after_opt_a]: 0.00018801 [convert_after_rewriter]: 1.344e-05 [order_py_execute_after_rewriter]: 1.069e-05 [mutable_eliminate]: 0.00081347 [opt_b]: 0.00051255, [1] [Cycle 1]: 0.00050397, [7] [b_1]: 0.00036053 [b_2]: 1.772e-05 [updatestate_depend_eliminate]: 1.237e-05 [updatestate_assign_eliminate]: 7.25e-06 [updatestate_loads_eliminate]: 8.96002e-06 [renormalize]: 2.60014e-07 [cse]: 5.632e-05 [optimize_parallel_all_gather_comm]: 2.654e-05 [overlap_param_gather]: 1.86e-06 [cconv]: 3.561e-05 [loop_unroll]: 0.00051323 [opt_after_cconv]: 0.00021701, [1] [Cycle 1]: 0.00020969, [7] [c_1]: 9.788e-05 [parameter_eliminate]: 3.95e-06 [updatestate_depend_eliminate]: 1.202e-05 [updatestate_assign_eliminate]: 6.79999e-06 [updatestate_loads_eliminate]: 8.22e-06 [cse]: 4.398e-05 [renormalize]: 6.10016e-07 [remove_dup_value]: 6.771e-05 [tuple_transform]: 0.0001561, [1] [Cycle 1]: 0.0001511, [4] [d_1]: 0.00011166 [none_parameter_eliminate]: 2.10002e-06 [renormalize]: 2.19996e-07 [switch_simplify]: 1.62e-05 [partial_unused_args_eliminate]: 1.86e-06 [add_recomputation]: 0.00011065 [cse_after_recomputation]: 5.485e-05, [1] [Cycle 1]: 5.009e-05, [1] [cse]: 4.343e-05 [environ_conv]: 1.749e-05 [swap_dp_allreduce_reducescatter]: 1.141e-05 [bias_add_comm_swap]: 2.91e-06 [label_micro_interleaved_index]: 5.56002e-06 [label_fine_grained_interleaved_index]: 2.68998e-06 [merge_cast_opt]: 1.82001e-06 [slice_recompute_activation]: 2.63e-06 [micro_interleaved_order_control]: 2.17999e-06 [assign_add_opt]: 1.52999e-06 [ForceFp32Comm]: 8.50006e-07 [remove_cast_before_assign_add]: 1.15001e-06 [full_micro_interleaved_order_control]: 2.55002e-06 [reorder_send_recv_between_fp_bp]: 2.86e-06 [comm_op_add_attrs]: 1.39998e-06 [add_comm_op_reuse_tag]: 1.44e-06 [interleave_split_concat_branches]: 1.22e-06 [interleave_parallel_branches]: 1.09003e-06 [overlap_opt_shard_in_pipeline]: 1.27e-06 [overlap_opt_shard_grad_in_pipeline]: 1.72999e-06 [control_data_broadcast_order]: 2.513e-05 [grouped_pairwise_exchange_alltoall]: 2.01e-06 [offloading_packed_experts]: 6.79999e-06 [overlap_recompute_and_grad_model_parallel]: 7.78999e-06 [overlap_grad_matmul_and_grad_allreduce]: 1.50001e-06 [overlap_recompute_allgather_and_fa_grad]: 1.43002e-06 [overlap_recompute_comm]: 2.36e-06 [overlap_grad_ring_attention]: 7.88001e-06 [overlap_grad_flash_sp]: 3.752e-05 [begin_end_overlap_inline]: 4.89992e-07 [split_matmul_comm_elemetwise]: 2.51e-06 [split_layernorm_comm]: 1.71e-06 [handle_group_info]: 1.08001e-06 [symbol_engine_optimizer]: 0.00015789, [1] [Cycle 1]: 0.00015163, [6] [build]: 1.807e-05 [elim_shapecalc]: 2.203e-05 [elim_not_effective]: 2.976e-05 [opt_reshape]: 2.238e-05 [fold_const_symbol]: 2.508e-05 [renormalize]: 1.79978e-07 [detach_backward]: 2.38998e-06 [pipeline_parallel_scheduler]: 1.43002e-06 [auto_monad_reorder]: 4.613e-05 [get_jit_bprop_graph]: 2.23998e-06 [rewriter_after_jit_bprop_graph]: 5.87001e-06 [opt_after_jit_grad]: 0.0525978 [validate]: 0.00013109 Sums bootstrap : 0.000802s : 0.07% type_inference : 1.046124s : 89.09% event_method : 0.000019s : 0.00% auto_monad : 0.000371s : 0.03% graph_reusing : 0.000007s : 0.00% inline : 0.000003s : 0.00% add_attr.add_attr_with_inline.tag_attr : 0.000023s : 0.00% add_attr.add_attr_with_inline.meta_addattr_fg_expand : 0.000004s : 0.00% parallel-infer-symbol : 0.000005s : 0.00% pre_auto_parallel : 0.000039s : 0.00% insert-virtual-dataset : 0.000003s : 0.00% parallel-infer-symbol-second : 0.000001s : 0.00% dataset_repeat_opt : 0.000002s : 0.00% pipeline_split : 0.000002s : 0.00% optimize.py_interpret_to_execute : 0.000035s : 0.00% optimize.rewriter_before_opt_a : 0.000094s : 0.01% optimize.opt_a.expand_dump_flag : 0.000006s : 0.00% optimize.opt_a.switch_simplify : 0.000055s : 0.00% optimize.opt_a.loop_unroll : 0.000036s : 0.00% optimize.opt_a.a_1 : 0.058139s : 4.95% optimize.opt_a.with_stream_mark : 0.000067s : 0.01% optimize.opt_a.recompute_prepare : 0.000044s : 0.00% optimize.opt_a.updatestate_depend_eliminate : 0.000024s : 0.00% optimize.opt_a.updatestate_assign_eliminate : 0.000016s : 0.00% optimize.opt_a.updatestate_loads_eliminate : 0.000022s : 0.00% optimize.opt_a.parameter_eliminate : 0.000005s : 0.00% optimize.opt_a.a_2 : 0.000477s : 0.04% optimize.opt_a.accelerated_algorithm : 0.000069s : 0.01% optimize.opt_a.shard : 0.000006s : 0.00% optimize.opt_a.meta_shard_fg_expand : 0.000010s : 0.00% optimize.opt_a.shard_inline : 0.000034s : 0.00% optimize.opt_a.merge_send_recv : 0.000030s : 0.00% optimize.opt_a.auto_parallel : 0.000030s : 0.00% optimize.opt_a.parallel : 0.000048s : 0.00% optimize.opt_a.flash_sp : 0.000019s : 0.00% optimize.opt_a.merge_comm : 0.000017s : 0.00% optimize.opt_a.allreduce_fusion : 0.000018s : 0.00% optimize.opt_a.matmul_add_comm_reduction : 0.000034s : 0.00% optimize.opt_a.allreduce_slice_to_reducescatter : 0.000002s : 0.00% optimize.opt_a.virtual_shard_identity : 0.000040s : 0.00% optimize.opt_a.virtual_dataset : 0.000034s : 0.00% optimize.opt_a.get_grad_eliminate_ : 0.000036s : 0.00% optimize.opt_a.virtual_output : 0.000034s : 0.00% optimize.opt_a.merge_forward : 0.000018s : 0.00% optimize.opt_a.cell_reuse_recompute_pass : 0.000007s : 0.00% optimize.opt_a.offload_activation : 0.000037s : 0.00% optimize.opt_a.cell_reuse_handle_not_recompute_node_pass : 0.000064s : 0.01% optimize.opt_a.merge_recompute_call_nodes : 0.000003s : 0.00% optimize.opt_a.before_grad : 0.000057s : 0.00% optimize.opt_a.set_forward_comm_id_for_comm_node_pass : 0.000019s : 0.00% optimize.opt_a.meta_fg_expand : 0.000014s : 0.00% optimize.opt_a.flash_sp_send_recv_attached : 0.000009s : 0.00% optimize.opt_a.receive_attached : 0.000005s : 0.00% optimize.opt_a.after_resolve : 0.000054s : 0.00% optimize.opt_a.a_after_grad : 0.000056s : 0.00% optimize.opt_a.renormalize : 0.010255s : 0.87% optimize.opt_a.add_forward_monad_depend : 0.000021s : 0.00% optimize.opt_a.auto_monad_grad : 0.000005s : 0.00% optimize.opt_a.auto_monad_eliminator : 0.000095s : 0.01% optimize.opt_a.cse : 0.000766s : 0.07% optimize.opt_a.a_3 : 0.000236s : 0.02% optimize.py_interpret_to_execute_after_opt_a : 0.000027s : 0.00% optimize.slice_cell_reuse_recomputed_activation : 0.000002s : 0.00% optimize.rewriter_after_opt_a : 0.000188s : 0.02% optimize.convert_after_rewriter : 0.000013s : 0.00% optimize.order_py_execute_after_rewriter : 0.000011s : 0.00% optimize.mutable_eliminate : 0.000813s : 0.07% optimize.opt_b.b_1 : 0.000361s : 0.03% optimize.opt_b.b_2 : 0.000018s : 0.00% optimize.opt_b.updatestate_depend_eliminate : 0.000012s : 0.00% optimize.opt_b.updatestate_assign_eliminate : 0.000007s : 0.00% optimize.opt_b.updatestate_loads_eliminate : 0.000009s : 0.00% optimize.opt_b.renormalize : 0.000000s : 0.00% optimize.opt_b.cse : 0.000056s : 0.00% optimize.optimize_parallel_all_gather_comm : 0.000027s : 0.00% optimize.overlap_param_gather : 0.000002s : 0.00% optimize.cconv : 0.000036s : 0.00% optimize.loop_unroll : 0.000513s : 0.04% optimize.opt_after_cconv.c_1 : 0.000098s : 0.01% optimize.opt_after_cconv.parameter_eliminate : 0.000004s : 0.00% optimize.opt_after_cconv.updatestate_depend_eliminate : 0.000012s : 0.00% optimize.opt_after_cconv.updatestate_assign_eliminate : 0.000007s : 0.00% optimize.opt_after_cconv.updatestate_loads_eliminate : 0.000008s : 0.00% optimize.opt_after_cconv.cse : 0.000044s : 0.00% optimize.opt_after_cconv.renormalize : 0.000001s : 0.00% optimize.remove_dup_value : 0.000068s : 0.01% optimize.tuple_transform.d_1 : 0.000112s : 0.01% optimize.tuple_transform.none_parameter_eliminate : 0.000002s : 0.00% optimize.tuple_transform.renormalize : 0.000000s : 0.00% optimize.tuple_transform.switch_simplify : 0.000016s : 0.00% optimize.partial_unused_args_eliminate : 0.000002s : 0.00% optimize.add_recomputation : 0.000111s : 0.01% optimize.cse_after_recomputation.cse : 0.000043s : 0.00% optimize.environ_conv : 0.000017s : 0.00% optimize.swap_dp_allreduce_reducescatter : 0.000011s : 0.00% optimize.bias_add_comm_swap : 0.000003s : 0.00% optimize.label_micro_interleaved_index : 0.000006s : 0.00% optimize.label_fine_grained_interleaved_index : 0.000003s : 0.00% optimize.merge_cast_opt : 0.000002s : 0.00% optimize.slice_recompute_activation : 0.000003s : 0.00% optimize.micro_interleaved_order_control : 0.000002s : 0.00% optimize.assign_add_opt : 0.000002s : 0.00% optimize.ForceFp32Comm : 0.000001s : 0.00% optimize.remove_cast_before_assign_add : 0.000001s : 0.00% optimize.full_micro_interleaved_order_control : 0.000003s : 0.00% optimize.reorder_send_recv_between_fp_bp : 0.000003s : 0.00% optimize.comm_op_add_attrs : 0.000001s : 0.00% optimize.add_comm_op_reuse_tag : 0.000001s : 0.00% optimize.interleave_split_concat_branches : 0.000001s : 0.00% optimize.interleave_parallel_branches : 0.000001s : 0.00% optimize.overlap_opt_shard_in_pipeline : 0.000001s : 0.00% optimize.overlap_opt_shard_grad_in_pipeline : 0.000002s : 0.00% optimize.control_data_broadcast_order : 0.000025s : 0.00% optimize.grouped_pairwise_exchange_alltoall : 0.000002s : 0.00% optimize.offloading_packed_experts : 0.000007s : 0.00% optimize.overlap_recompute_and_grad_model_parallel : 0.000008s : 0.00% optimize.overlap_grad_matmul_and_grad_allreduce : 0.000002s : 0.00% optimize.overlap_recompute_allgather_and_fa_grad : 0.000001s : 0.00% optimize.overlap_recompute_comm : 0.000002s : 0.00% optimize.overlap_grad_ring_attention : 0.000008s : 0.00% optimize.overlap_grad_flash_sp : 0.000038s : 0.00% optimize.begin_end_overlap_inline : 0.000000s : 0.00% optimize.split_matmul_comm_elemetwise : 0.000003s : 0.00% optimize.split_layernorm_comm : 0.000002s : 0.00% optimize.handle_group_info : 0.000001s : 0.00% optimize.symbol_engine_optimizer.build : 0.000018s : 0.00% optimize.symbol_engine_optimizer.elim_shapecalc : 0.000022s : 0.00% optimize.symbol_engine_optimizer.elim_not_effective : 0.000030s : 0.00% optimize.symbol_engine_optimizer.opt_reshape : 0.000022s : 0.00% optimize.symbol_engine_optimizer.fold_const_symbol : 0.000025s : 0.00% optimize.symbol_engine_optimizer.renormalize : 0.000000s : 0.00% detach_backward : 0.000002s : 0.00% pipeline_parallel_scheduler : 0.000001s : 0.00% auto_monad_reorder : 0.000046s : 0.00% get_jit_bprop_graph : 0.000002s : 0.00% rewriter_after_jit_bprop_graph : 0.000006s : 0.00% opt_after_jit_grad : 0.052598s : 4.48% validate : 0.000131s : 0.01% Time group info: ------[substitution.] 0.000329 96 3.49% : 0.000011s : 2: substitution.depend_value_elim 1.23% : 0.000004s : 8: substitution.elim_not_effective 1.02% : 0.000003s : 8: substitution.fold_const_symbol 3.58% : 0.000012s : 13: substitution.graph_param_transform 43.85% : 0.000144s : 1: substitution.inline 2.67% : 0.000009s : 16: substitution.j_node_and_user_rematch 7.28% : 0.000024s : 2: substitution.less_batch_normalization 1.80% : 0.000006s : 6: substitution.load_eliminater 0.84% : 0.000003s : 2: substitution.opt_reshape 3.76% : 0.000012s : 16: substitution.remove_not_recompute_node 2.50% : 0.000008s : 4: substitution.replace_old_param 11.78% : 0.000039s : 4: substitution.reshape_eliminate 9.35% : 0.000031s : 6: substitution.updatestate_pure_node_eliminater 6.87% : 0.000023s : 8: substitution.updatestate_useless_node_eliminater ------[type_inference.] 1.046010 2 99.47% : 1.040442s : 1: type_inference.infer 0.53% : 0.005568s : 1: type_inference.specialize ------[replace.] 0.000020 1 100.00% : 0.000020s : 1: replace.inline ------[match.] 0.000143 1 100.00% : 0.000143s : 1: match.inline ------[predicate.] 0.000466 3002 0.88% : 0.000004s : 27: predicate.accumulaten_eliminater 2.05% : 0.000010s : 13: predicate.ad_related_special_op_eliminate 0.72% : 0.000003s : 26: predicate.addn_check_dump 0.98% : 0.000005s : 27: predicate.addn_zero_filter 0.75% : 0.000003s : 27: predicate.adjust_all_reduce_mul_add 2.49% : 0.000012s : 53: predicate.arithmetic_simplify 0.81% : 0.000004s : 27: predicate.cast_eliminate 0.77% : 0.000004s : 26: predicate.check_bprop_eliminate 0.70% : 0.000003s : 26: predicate.compare_switch_simplify 0.24% : 0.000001s : 13: predicate.const_output_eliminate 0.85% : 0.000004s : 26: predicate.depend_value_elim 0.86% : 0.000004s : 27: predicate.dict_get_item_const_eliminator 0.96% : 0.000004s : 27: predicate.dict_get_item_eliminator 0.82% : 0.000004s : 27: predicate.dict_set_item_eliminator 1.38% : 0.000006s : 26: predicate.dumpgradient_eliminate 0.25% : 0.000001s : 13: predicate.elim_not_effective 0.44% : 0.000002s : 13: predicate.elim_shapecalc_of_broadcastargs 1.23% : 0.000006s : 40: predicate.environ_add_const_eliminate 1.10% : 0.000005s : 40: predicate.environ_get_add_eliminate 1.09% : 0.000005s : 40: predicate.environ_get_depend_swap 1.93% : 0.000009s : 66: predicate.environ_get_eliminate 1.16% : 0.000005s : 40: predicate.environ_get_set_eliminate 0.78% : 0.000004s : 28: predicate.exchange_switch_depend_value 1.33% : 0.000006s : 28: predicate.float_depend_g_call 0.77% : 0.000004s : 26: predicate.float_environ_get_switch 1.07% : 0.000005s : 39: predicate.float_tuple_getitem_switch 0.21% : 0.000001s : 13: predicate.fold_const_symbol 0.82% : 0.000004s : 26: predicate.get_grad_eliminate 0.26% : 0.000001s : 13: predicate.graph_param_transform 0.72% : 0.000003s : 26: predicate.incorporate_call 0.66% : 0.000003s : 26: predicate.incorporate_call_switch 5.49% : 0.000026s : 133: predicate.inline 0.99% : 0.000005s : 26: predicate.inline_without_move 0.42% : 0.000002s : 26: predicate.j_node_and_user_rematch 1.16% : 0.000005s : 27: predicate.less_batch_normalization 1.75% : 0.000008s : 53: predicate.list_to_tuple_eliminator_ 2.31% : 0.000011s : 80: predicate.load_eliminater 0.87% : 0.000004s : 13: predicate.loop_unroll_after_grad 0.94% : 0.000004s : 31: predicate.loop_unroll_before_grad 2.24% : 0.000010s : 53: predicate.make_slice_get_slice_eliminator 0.74% : 0.000003s : 26: predicate.merge_addn 0.70% : 0.000003s : 26: predicate.micro_step_allgather_replace 0.72% : 0.000003s : 26: predicate.mini_step_allgather_replace 0.72% : 0.000003s : 27: predicate.minmaximum_grad 1.08% : 0.000005s : 13: predicate.mutable_eliminate 0.48% : 0.000002s : 13: predicate.opt_reshape 0.46% : 0.000002s : 13: predicate.parallel_virtual_node 1.42% : 0.000007s : 28: predicate.partial_defer_inline 1.21% : 0.000006s : 40: predicate.partial_eliminate 0.85% : 0.000004s : 27: predicate.print_const_string_wrapper 0.77% : 0.000004s : 26: predicate.reduce_all_const_elim 1.33% : 0.000006s : 27: predicate.reduce_eliminate 2.32% : 0.000011s : 80: predicate.redundant_stop_gradient_eliminater 0.44% : 0.000002s : 26: predicate.remove_not_recompute_node 1.24% : 0.000006s : 53: predicate.replace_applicator 0.53% : 0.000002s : 26: predicate.replace_old_param 0.32% : 0.000002s : 13: predicate.reset_defer_inline 1.06% : 0.000005s : 27: predicate.reshape_eliminate 0.76% : 0.000004s : 26: predicate.row_tensor_add_zeros_like 0.47% : 0.000002s : 13: predicate.row_tensor_eliminate 0.96% : 0.000004s : 26: predicate.same_eliminate 0.51% : 0.000002s : 26: predicate.set_cell_output_no_recompute 0.98% : 0.000005s : 26: predicate.shard_identity_eliminate 0.94% : 0.000004s : 26: predicate.special_op_eliminate 0.84% : 0.000004s : 26: predicate.specialize_transform 1.07% : 0.000005s : 26: predicate.split_environ_get_set_with_tuple_value 0.89% : 0.000004s : 26: predicate.stack_unstack_eliminate 0.47% : 0.000002s : 13: predicate.switch_call_monad_eliminater 0.93% : 0.000004s : 28: predicate.switch_defer_inline 1.62% : 0.000008s : 54: predicate.switch_layer_defer_inline 3.14% : 0.000015s : 98: predicate.switch_simplify 1.01% : 0.000005s : 27: predicate.tile_eliminate 0.96% : 0.000004s : 27: predicate.transpose_eliminate 1.62% : 0.000008s : 53: predicate.tuple_list_convert_item_index_to_positive 1.62% : 0.000008s : 53: predicate.tuple_list_get_item_const_eliminator 1.60% : 0.000007s : 53: predicate.tuple_list_get_item_depend_reorder 3.68% : 0.000017s : 79: predicate.tuple_list_get_item_eliminator 1.70% : 0.000008s : 53: predicate.tuple_list_get_set_item_eliminator 2.55% : 0.000012s : 79: predicate.tuple_list_set_item_eliminator 1.60% : 0.000007s : 53: predicate.tuple_to_list_eliminator_ 2.19% : 0.000010s : 80: predicate.updatestate_pure_node_eliminater 3.13% : 0.000015s : 106: predicate.updatestate_useless_node_eliminater 0.43% : 0.000002s : 13: predicate.value_based_eliminate 0.89% : 0.000004s : 26: predicate.virtual_dataset_eliminate 0.85% : 0.000004s : 26: predicate.virtual_output_eliminate 0.40% : 0.000002s : 13: predicate.virtual_view_grad_eliminate 0.52% : 0.000002s : 13: predicate.zero_like_fill_zero ------[func_graph_cloner_run.] 0.056982 58 94.75% : 0.053992s : 55: func_graph_cloner_run.FuncGraphClonerGraph 5.25% : 0.002990s : 3: func_graph_cloner_run.FuncGraphSpecializer ------[meta_graph.] 0.000000 0 ------[manager.] 0.000000 0 ------[pynative] 0.000000 0 ------[others.] 1.335548 192 0.00% : 0.000003s : 1: ForceFp32Comm 0.55% : 0.007374s : 1: add_attr 0.55% : 0.007351s : 1: add_attr_with_inline 0.00% : 0.000004s : 1: add_comm_op_reuse_tag 0.01% : 0.000116s : 1: add_recomputation 0.00% : 0.000004s : 1: assign_add_opt 0.03% : 0.000383s : 1: auto_monad 0.00% : 0.000051s : 1: auto_monad_reorder 0.00% : 0.000003s : 1: begin_end_overlap_inline 0.00% : 0.000006s : 1: bias_add_comm_swap 0.06% : 0.000860s : 1: bootstrap 0.00% : 0.000039s : 1: cconv 0.00% : 0.000005s : 1: comm_op_add_attrs 0.00% : 0.000029s : 1: control_data_broadcast_order 0.00% : 0.000017s : 1: convert_after_rewriter 0.00% : 0.000058s : 1: cse_after_recomputation 0.00% : 0.000007s : 1: dataset_repeat_opt 0.00% : 0.000005s : 1: detach_backward 0.00% : 0.000022s : 1: environ_conv 0.00% : 0.000028s : 1: event_method 0.00% : 0.000005s : 1: full_micro_interleaved_order_control 0.00% : 0.000006s : 1: get_jit_bprop_graph 0.00% : 0.000011s : 1: graph_reusing 0.00% : 0.000005s : 1: grouped_pairwise_exchange_alltoall 0.00% : 0.000004s : 1: handle_group_info 0.00% : 0.000007s : 1: inline 0.00% : 0.000007s : 1: insert-virtual-dataset 0.00% : 0.000005s : 1: interleave_parallel_branches 0.00% : 0.000004s : 1: interleave_split_concat_branches 0.00% : 0.000006s : 1: label_fine_grained_interleaved_index 0.00% : 0.000008s : 1: label_micro_interleaved_index 0.04% : 0.000523s : 1: loop_unroll 0.00% : 0.000005s : 1: merge_cast_opt 0.00% : 0.000005s : 1: micro_interleaved_order_control 0.06% : 0.000827s : 1: mutable_eliminate 0.00% : 0.000010s : 1: offloading_packed_experts 0.00% : 0.000026s : 1: opt.transform.loop_unroll_optimizer 0.00% : 0.000033s : 1: opt.transform.mutable_eliminate 4.44% : 0.059343s : 78: opt.transform.opt_a 0.01% : 0.000096s : 1: opt.transform.opt_after_cconv 0.01% : 0.000096s : 1: opt.transform.opt_after_jit_grad 0.03% : 0.000343s : 28: opt.transform.opt_b 0.01% : 0.000125s : 2: opt.transform.opt_trans_graph 0.01% : 0.000095s : 4: opt.transform.symbol_engine_opt 5.37% : 0.071676s : 1: opt_a 0.02% : 0.000221s : 1: opt_after_cconv 3.94% : 0.052622s : 1: opt_after_jit_grad 0.04% : 0.000516s : 1: opt_b 5.63% : 0.075140s : 1: optimize 0.00% : 0.000030s : 1: optimize_parallel_all_gather_comm 0.00% : 0.000014s : 1: order_py_execute_after_rewriter 0.00% : 0.000041s : 1: overlap_grad_flash_sp 0.00% : 0.000004s : 1: overlap_grad_matmul_and_grad_allreduce 0.00% : 0.000012s : 1: overlap_grad_ring_attention 0.00% : 0.000004s : 1: overlap_opt_shard_grad_in_pipeline 0.00% : 0.000004s : 1: overlap_opt_shard_in_pipeline 0.00% : 0.000006s : 1: overlap_param_gather 0.00% : 0.000005s : 1: overlap_recompute_allgather_and_fa_grad 0.00% : 0.000011s : 1: overlap_recompute_and_grad_model_parallel 0.00% : 0.000005s : 1: overlap_recompute_comm 0.00% : 0.000009s : 1: parallel-infer-symbol 0.00% : 0.000004s : 1: parallel-infer-symbol-second 0.00% : 0.000005s : 1: partial_unused_args_eliminate 0.00% : 0.000005s : 1: pipeline_parallel_scheduler 0.00% : 0.000006s : 1: pipeline_split 0.00% : 0.000045s : 1: pre_auto_parallel 0.00% : 0.000039s : 1: py_interpret_to_execute 0.00% : 0.000030s : 1: py_interpret_to_execute_after_opt_a 0.00% : 0.000004s : 1: remove_cast_before_assign_add 0.01% : 0.000072s : 1: remove_dup_value 0.51% : 0.006858s : 1: renormalize.infer 0.25% : 0.003380s : 1: renormalize.specialize 0.00% : 0.000006s : 1: reorder_send_recv_between_fp_bp 0.00% : 0.000010s : 1: rewriter_after_jit_bprop_graph 0.01% : 0.000193s : 1: rewriter_after_opt_a 0.01% : 0.000099s : 1: rewriter_before_opt_a 0.00% : 0.000005s : 1: slice_cell_reuse_recomputed_activation 0.00% : 0.000005s : 1: slice_recompute_activation 0.00% : 0.000006s : 1: split_layernorm_comm 0.00% : 0.000005s : 1: split_matmul_comm_elemetwise 0.00% : 0.000015s : 1: swap_dp_allreduce_reducescatter 0.01% : 0.000161s : 1: symbol_engine_optimizer 0.01% : 0.000159s : 1: tuple_transform 78.33% : 1.046157s : 1: type_inference . [hook] pytest_runtest_teardown:test_matmul_qkv_bf16_1[False-mstype0-k_n_shape3-1024] tests/st/infer/ops/test_internal_ops/test_matmul_split.py::test_matmul_qkv_bf16_1[False-mstype0-k_n_shape3-1024],max_mem:252.0M [WARNING] ME(168319:281472866918192,MainProcess):2026-01-29-17:44:16.857.371 [mindspore/context.py:1334] For 'context.set_context', the parameter 'device_target' will be deprecated and removed in a future version. Please use the api mindspore.set_device() instead. TotalTime = 0.722874, [21] [bootstrap]: 0.00194256 [type_inference]: 0.548735 [event_method]: 0.00021983 [auto_monad]: 0.00078255 [graph_reusing]: 5.056e-05 [inline]: 3.8e-06 [add_attr]: 0.0169724, [1] [add_attr_with_inline]: 0.0169534, [1] [Cycle 1]: 0.00030874, [2] [tag_attr]: 0.00020006 [meta_addattr_fg_expand]: 5.216e-05 [parallel-infer-symbol]: 4.92e-06 [pre_auto_parallel]: 0.00025583 [insert-virtual-dataset]: 3.08e-06 [parallel-infer-symbol-second]: 1.47001e-06 [dataset_repeat_opt]: 2.58998e-06 [pipeline_split]: 1.67001e-06 [optimize]: 0.152512, [53] [py_interpret_to_execute]: 0.0002437 [rewriter_before_opt_a]: 0.00080151 [opt_a]: 0.146103, [2] [Cycle 1]: 0.131345, [45] [expand_dump_flag]: 1.822e-05 [switch_simplify]: 0.00091053 [loop_unroll]: 0.00034135 [a_1]: 0.0732728 [with_stream_mark]: 6.428e-05 [recompute_prepare]: 4.351e-05 [updatestate_depend_eliminate]: 2.182e-05 [updatestate_assign_eliminate]: 1.641e-05 [updatestate_loads_eliminate]: 1.922e-05 [parameter_eliminate]: 3.71999e-06 [a_2]: 0.00079685 [accelerated_algorithm]: 7.414e-05 [shard]: 2.98998e-06 [meta_shard_fg_expand]: 2.716e-05 [shard_inline]: 2.325e-05 [merge_send_recv]: 2.129e-05 [auto_parallel]: 2.303e-05 [parallel]: 3.536e-05 [flash_sp]: 1.713e-05 [merge_comm]: 1.478e-05 [allreduce_fusion]: 1.317e-05 [matmul_add_comm_reduction]: 2.276e-05 [allreduce_slice_to_reducescatter]: 8.39995e-07 [virtual_shard_identity]: 2.892e-05 [virtual_dataset]: 2.345e-05 [get_grad_eliminate_]: 2.157e-05 [virtual_output]: 2.146e-05 [merge_forward]: 1.399e-05 [cell_reuse_recompute_pass]: 2.34001e-06 [offload_activation]: 2.408e-05 [cell_reuse_handle_not_recompute_node_pass]: 5.106e-05 [merge_recompute_call_nodes]: 2.27999e-06 [before_grad]: 4.163e-05 [set_forward_comm_id_for_comm_node_pass]: 1.642e-05 [meta_fg_expand]: 1.911e-05 [flash_sp_send_recv_attached]: 6.91001e-06 [receive_attached]: 2.59999e-06 [after_resolve]: 3.006e-05 [a_after_grad]: 3.606e-05 [renormalize]: 0.0522292 [add_forward_monad_depend]: 2.776e-05 [auto_monad_grad]: 4.05e-06 [auto_monad_eliminator]: 8.926e-05 [cse]: 0.00221551 [a_3]: 0.00018464 [Cycle 2]: 0.014739, [45] [expand_dump_flag]: 4.13999e-06 [switch_simplify]: 2.564e-05 [loop_unroll]: 2.107e-05 [a_1]: 0.00068385 [with_stream_mark]: 4.201e-05 [recompute_prepare]: 3.13e-05 [updatestate_depend_eliminate]: 1.69e-05 [updatestate_assign_eliminate]: 1.228e-05 [updatestate_loads_eliminate]: 1.415e-05 [parameter_eliminate]: 2.64999e-06 [a_2]: 0.00032158 [accelerated_algorithm]: 3.312e-05 [shard]: 2.17001e-06 [meta_shard_fg_expand]: 8.67998e-06 [shard_inline]: 2.103e-05 [merge_send_recv]: 2.119e-05 [auto_parallel]: 2.03e-05 [parallel]: 1.152e-05 [flash_sp]: 5.09003e-06 [merge_comm]: 1.398e-05 [allreduce_fusion]: 1.211e-05 [matmul_add_comm_reduction]: 2.258e-05 [allreduce_slice_to_reducescatter]: 1.04998e-06 [virtual_shard_identity]: 2.839e-05 [virtual_dataset]: 2.152e-05 [get_grad_eliminate_]: 2.285e-05 [virtual_output]: 1.987e-05 [merge_forward]: 1.348e-05 [cell_reuse_recompute_pass]: 3.65e-06 [offload_activation]: 2.612e-05 [cell_reuse_handle_not_recompute_node_pass]: 4.469e-05 [merge_recompute_call_nodes]: 1.64998e-06 [before_grad]: 3.735e-05 [set_forward_comm_id_for_comm_node_pass]: 1.447e-05 [meta_fg_expand]: 9.34998e-06 [flash_sp_send_recv_attached]: 1.67999e-06 [receive_attached]: 2.45997e-06 [after_resolve]: 2.863e-05 [a_after_grad]: 3.244e-05 [renormalize]: 1.50001e-07 [add_forward_monad_depend]: 4.02e-06 [auto_monad_grad]: 4.07998e-06 [auto_monad_eliminator]: 7.018e-05 [cse]: 0.0001562 [a_3]: 0.00017919 [py_interpret_to_execute_after_opt_a]: 4.013e-05 [slice_cell_reuse_recomputed_activation]: 3.55998e-06 [rewriter_after_opt_a]: 0.00088188 [convert_after_rewriter]: 2.236e-05 [order_py_execute_after_rewriter]: 1.522e-05 [mutable_eliminate]: 0.00084267 [opt_b]: 0.00078498, [1] [Cycle 1]: 0.00077483, [7] [b_1]: 0.0005592 [b_2]: 2.635e-05 [updatestate_depend_eliminate]: 2.173e-05 [updatestate_assign_eliminate]: 1.209e-05 [updatestate_loads_eliminate]: 1.467e-05 [renormalize]: 1.00001e-06 [cse]: 9.047e-05 [optimize_parallel_all_gather_comm]: 4.163e-05 [overlap_param_gather]: 1.97999e-06 [cconv]: 4.329e-05 [loop_unroll]: 0.00062893 [opt_after_cconv]: 0.00031169, [1] [Cycle 1]: 0.00030107, [7] [c_1]: 0.00012913 [parameter_eliminate]: 5.92001e-06 [updatestate_depend_eliminate]: 1.943e-05 [updatestate_assign_eliminate]: 1.097e-05 [updatestate_loads_eliminate]: 1.326e-05 [cse]: 8.077e-05 [renormalize]: 8.39995e-07 [remove_dup_value]: 0.00010759 [tuple_transform]: 0.00026705, [1] [Cycle 1]: 0.00026035, [4] [d_1]: 0.00020785 [none_parameter_eliminate]: 3.13998e-06 [renormalize]: 2.30008e-07 [switch_simplify]: 2.31e-05 [partial_unused_args_eliminate]: 2.31e-06 [add_recomputation]: 0.00015448 [cse_after_recomputation]: 8.109e-05, [1] [Cycle 1]: 7.344e-05, [1] [cse]: 6.469e-05 [environ_conv]: 2.519e-05 [swap_dp_allreduce_reducescatter]: 1.96e-05 [bias_add_comm_swap]: 3.28998e-06 [label_micro_interleaved_index]: 6.93e-06 [label_fine_grained_interleaved_index]: 3.13e-06 [merge_cast_opt]: 2.38002e-06 [slice_recompute_activation]: 2.39999e-06 [micro_interleaved_order_control]: 2.59001e-06 [assign_add_opt]: 1.68997e-06 [ForceFp32Comm]: 1.09e-06 [remove_cast_before_assign_add]: 1.17999e-06 [full_micro_interleaved_order_control]: 2.57001e-06 [reorder_send_recv_between_fp_bp]: 3.2e-06 [comm_op_add_attrs]: 1.82999e-06 [add_comm_op_reuse_tag]: 1.28002e-06 [interleave_split_concat_branches]: 1.39e-06 [interleave_parallel_branches]: 1.34e-06 [overlap_opt_shard_in_pipeline]: 1.57001e-06 [overlap_opt_shard_grad_in_pipeline]: 2.17001e-06 [control_data_broadcast_order]: 4.004e-05 [grouped_pairwise_exchange_alltoall]: 2.16e-06 [offloading_packed_experts]: 1.073e-05 [overlap_recompute_and_grad_model_parallel]: 1.06e-05 [overlap_grad_matmul_and_grad_allreduce]: 1.27999e-06 [overlap_recompute_allgather_and_fa_grad]: 1.47001e-06 [overlap_recompute_comm]: 2.43e-06 [overlap_grad_ring_attention]: 9.20001e-06 [overlap_grad_flash_sp]: 5.146e-05 [begin_end_overlap_inline]: 8.09989e-07 [split_matmul_comm_elemetwise]: 2.31998e-06 [split_layernorm_comm]: 2.21e-06 [handle_group_info]: 1.10001e-06 [symbol_engine_optimizer]: 0.00059734, [1] [Cycle 1]: 0.00058959, [6] [build]: 0.00033682 [elim_shapecalc]: 3.788e-05 [elim_not_effective]: 6.64e-05 [opt_reshape]: 3.425e-05 [fold_const_symbol]: 6.11e-05 [renormalize]: 5.39992e-07 [detach_backward]: 3.37002e-06 [pipeline_parallel_scheduler]: 1.65001e-06 [auto_monad_reorder]: 6.921e-05 [get_jit_bprop_graph]: 2.11e-06 [rewriter_after_jit_bprop_graph]: 6.73e-06 [opt_after_jit_grad]: 0.0008043 [validate]: 0.00011635 Sums bootstrap : 0.001943s : 0.28% type_inference : 0.548735s : 79.29% event_method : 0.000220s : 0.03% auto_monad : 0.000783s : 0.11% graph_reusing : 0.000051s : 0.01% inline : 0.000004s : 0.00% add_attr.add_attr_with_inline.tag_attr : 0.000200s : 0.03% add_attr.add_attr_with_inline.meta_addattr_fg_expand : 0.000052s : 0.01% parallel-infer-symbol : 0.000005s : 0.00% pre_auto_parallel : 0.000256s : 0.04% insert-virtual-dataset : 0.000003s : 0.00% parallel-infer-symbol-second : 0.000001s : 0.00% dataset_repeat_opt : 0.000003s : 0.00% pipeline_split : 0.000002s : 0.00% optimize.py_interpret_to_execute : 0.000244s : 0.04% optimize.rewriter_before_opt_a : 0.000802s : 0.12% optimize.opt_a.expand_dump_flag : 0.000022s : 0.00% optimize.opt_a.switch_simplify : 0.000936s : 0.14% optimize.opt_a.loop_unroll : 0.000362s : 0.05% optimize.opt_a.a_1 : 0.073957s : 10.69% optimize.opt_a.with_stream_mark : 0.000106s : 0.02% optimize.opt_a.recompute_prepare : 0.000075s : 0.01% optimize.opt_a.updatestate_depend_eliminate : 0.000039s : 0.01% optimize.opt_a.updatestate_assign_eliminate : 0.000029s : 0.00% optimize.opt_a.updatestate_loads_eliminate : 0.000033s : 0.00% optimize.opt_a.parameter_eliminate : 0.000006s : 0.00% optimize.opt_a.a_2 : 0.001118s : 0.16% optimize.opt_a.accelerated_algorithm : 0.000107s : 0.02% optimize.opt_a.shard : 0.000005s : 0.00% optimize.opt_a.meta_shard_fg_expand : 0.000036s : 0.01% optimize.opt_a.shard_inline : 0.000044s : 0.01% optimize.opt_a.merge_send_recv : 0.000042s : 0.01% optimize.opt_a.auto_parallel : 0.000043s : 0.01% optimize.opt_a.parallel : 0.000047s : 0.01% optimize.opt_a.flash_sp : 0.000022s : 0.00% optimize.opt_a.merge_comm : 0.000029s : 0.00% optimize.opt_a.allreduce_fusion : 0.000025s : 0.00% optimize.opt_a.matmul_add_comm_reduction : 0.000045s : 0.01% optimize.opt_a.allreduce_slice_to_reducescatter : 0.000002s : 0.00% optimize.opt_a.virtual_shard_identity : 0.000057s : 0.01% optimize.opt_a.virtual_dataset : 0.000045s : 0.01% optimize.opt_a.get_grad_eliminate_ : 0.000044s : 0.01% optimize.opt_a.virtual_output : 0.000041s : 0.01% optimize.opt_a.merge_forward : 0.000027s : 0.00% optimize.opt_a.cell_reuse_recompute_pass : 0.000006s : 0.00% optimize.opt_a.offload_activation : 0.000050s : 0.01% optimize.opt_a.cell_reuse_handle_not_recompute_node_pass : 0.000096s : 0.01% optimize.opt_a.merge_recompute_call_nodes : 0.000004s : 0.00% optimize.opt_a.before_grad : 0.000079s : 0.01% optimize.opt_a.set_forward_comm_id_for_comm_node_pass : 0.000031s : 0.00% optimize.opt_a.meta_fg_expand : 0.000028s : 0.00% optimize.opt_a.flash_sp_send_recv_attached : 0.000009s : 0.00% optimize.opt_a.receive_attached : 0.000005s : 0.00% optimize.opt_a.after_resolve : 0.000059s : 0.01% optimize.opt_a.a_after_grad : 0.000068s : 0.01% optimize.opt_a.renormalize : 0.052229s : 7.55% optimize.opt_a.add_forward_monad_depend : 0.000032s : 0.00% optimize.opt_a.auto_monad_grad : 0.000008s : 0.00% optimize.opt_a.auto_monad_eliminator : 0.000159s : 0.02% optimize.opt_a.cse : 0.002372s : 0.34% optimize.opt_a.a_3 : 0.000364s : 0.05% optimize.py_interpret_to_execute_after_opt_a : 0.000040s : 0.01% optimize.slice_cell_reuse_recomputed_activation : 0.000004s : 0.00% optimize.rewriter_after_opt_a : 0.000882s : 0.13% optimize.convert_after_rewriter : 0.000022s : 0.00% optimize.order_py_execute_after_rewriter : 0.000015s : 0.00% optimize.mutable_eliminate : 0.000843s : 0.12% optimize.opt_b.b_1 : 0.000559s : 0.08% optimize.opt_b.b_2 : 0.000026s : 0.00% optimize.opt_b.updatestate_depend_eliminate : 0.000022s : 0.00% optimize.opt_b.updatestate_assign_eliminate : 0.000012s : 0.00% optimize.opt_b.updatestate_loads_eliminate : 0.000015s : 0.00% optimize.opt_b.renormalize : 0.000001s : 0.00% optimize.opt_b.cse : 0.000090s : 0.01% optimize.optimize_parallel_all_gather_comm : 0.000042s : 0.01% optimize.overlap_param_gather : 0.000002s : 0.00% optimize.cconv : 0.000043s : 0.01% optimize.loop_unroll : 0.000629s : 0.09% optimize.opt_after_cconv.c_1 : 0.000129s : 0.02% optimize.opt_after_cconv.parameter_eliminate : 0.000006s : 0.00% optimize.opt_after_cconv.updatestate_depend_eliminate : 0.000019s : 0.00% optimize.opt_after_cconv.updatestate_assign_eliminate : 0.000011s : 0.00% optimize.opt_after_cconv.updatestate_loads_eliminate : 0.000013s : 0.00% optimize.opt_after_cconv.cse : 0.000081s : 0.01% optimize.opt_after_cconv.renormalize : 0.000001s : 0.00% optimize.remove_dup_value : 0.000108s : 0.02% optimize.tuple_transform.d_1 : 0.000208s : 0.03% optimize.tuple_transform.none_parameter_eliminate : 0.000003s : 0.00% optimize.tuple_transform.renormalize : 0.000000s : 0.00% optimize.tuple_transform.switch_simplify : 0.000023s : 0.00% optimize.partial_unused_args_eliminate : 0.000002s : 0.00% optimize.add_recomputation : 0.000154s : 0.02% optimize.cse_after_recomputation.cse : 0.000065s : 0.01% optimize.environ_conv : 0.000025s : 0.00% optimize.swap_dp_allreduce_reducescatter : 0.000020s : 0.00% optimize.bias_add_comm_swap : 0.000003s : 0.00% optimize.label_micro_interleaved_index : 0.000007s : 0.00% optimize.label_fine_grained_interleaved_index : 0.000003s : 0.00% optimize.merge_cast_opt : 0.000002s : 0.00% optimize.slice_recompute_activation : 0.000002s : 0.00% optimize.micro_interleaved_order_control : 0.000003s : 0.00% optimize.assign_add_opt : 0.000002s : 0.00% optimize.ForceFp32Comm : 0.000001s : 0.00% optimize.remove_cast_before_assign_add : 0.000001s : 0.00% optimize.full_micro_interleaved_order_control : 0.000003s : 0.00% optimize.reorder_send_recv_between_fp_bp : 0.000003s : 0.00% optimize.comm_op_add_attrs : 0.000002s : 0.00% optimize.add_comm_op_reuse_tag : 0.000001s : 0.00% optimize.interleave_split_concat_branches : 0.000001s : 0.00% optimize.interleave_parallel_branches : 0.000001s : 0.00% optimize.overlap_opt_shard_in_pipeline : 0.000002s : 0.00% optimize.overlap_opt_shard_grad_in_pipeline : 0.000002s : 0.00% optimize.control_data_broadcast_order : 0.000040s : 0.01% optimize.grouped_pairwise_exchange_alltoall : 0.000002s : 0.00% optimize.offloading_packed_experts : 0.000011s : 0.00% optimize.overlap_recompute_and_grad_model_parallel : 0.000011s : 0.00% optimize.overlap_grad_matmul_and_grad_allreduce : 0.000001s : 0.00% optimize.overlap_recompute_allgather_and_fa_grad : 0.000001s : 0.00% optimize.overlap_recompute_comm : 0.000002s : 0.00% optimize.overlap_grad_ring_attention : 0.000009s : 0.00% optimize.overlap_grad_flash_sp : 0.000051s : 0.01% optimize.begin_end_overlap_inline : 0.000001s : 0.00% optimize.split_matmul_comm_elemetwise : 0.000002s : 0.00% optimize.split_layernorm_comm : 0.000002s : 0.00% optimize.handle_group_info : 0.000001s : 0.00% optimize.symbol_engine_optimizer.build : 0.000337s : 0.05% optimize.symbol_engine_optimizer.elim_shapecalc : 0.000038s : 0.01% optimize.symbol_engine_optimizer.elim_not_effective : 0.000066s : 0.01% optimize.symbol_engine_optimizer.opt_reshape : 0.000034s : 0.00% optimize.symbol_engine_optimizer.fold_const_symbol : 0.000061s : 0.01% optimize.symbol_engine_optimizer.renormalize : 0.000001s : 0.00% detach_backward : 0.000003s : 0.00% pipeline_parallel_scheduler : 0.000002s : 0.00% auto_monad_reorder : 0.000069s : 0.01% get_jit_bprop_graph : 0.000002s : 0.00% rewriter_after_jit_bprop_graph : 0.000007s : 0.00% opt_after_jit_grad : 0.000804s : 0.12% validate : 0.000116s : 0.02% Time group info: ------[substitution.] 0.004180 569 0.40% : 0.000017s : 8: substitution.depend_value_elim 0.63% : 0.000027s : 14: substitution.elim_not_effective 0.48% : 0.000020s : 12: substitution.float_tuple_getitem_switch 0.67% : 0.000028s : 14: substitution.fold_const_symbol 0.37% : 0.000016s : 17: substitution.graph_param_transform 69.60% : 0.002909s : 71: substitution.inline 0.34% : 0.000014s : 29: substitution.j_node_and_user_rematch 0.96% : 0.000040s : 6: substitution.less_batch_normalization 1.54% : 0.000064s : 2: substitution.list_to_tuple_eliminator_ 0.16% : 0.000007s : 6: substitution.load_eliminater 1.05% : 0.000044s : 21: substitution.minmaximum_grad 0.20% : 0.000008s : 2: substitution.opt_reshape 0.47% : 0.000020s : 29: substitution.remove_not_recompute_node 0.22% : 0.000009s : 4: substitution.replace_old_param 1.59% : 0.000066s : 18: substitution.reshape_eliminate 1.35% : 0.000056s : 24: substitution.switch_simplify 4.19% : 0.000175s : 51: substitution.tuple_list_convert_item_index_to_positive 2.24% : 0.000093s : 51: substitution.tuple_list_get_item_const_eliminator 3.13% : 0.000131s : 51: substitution.tuple_list_get_item_depend_reorder 5.90% : 0.000246s : 73: substitution.tuple_list_get_item_eliminator 3.38% : 0.000141s : 51: substitution.tuple_list_get_set_item_eliminator 0.63% : 0.000026s : 6: substitution.updatestate_pure_node_eliminater 0.51% : 0.000021s : 9: substitution.updatestate_useless_node_eliminater ------[type_inference.] 0.548272 2 88.58% : 0.485674s : 1: type_inference.infer 11.42% : 0.062599s : 1: type_inference.specialize ------[replace.] 0.001309 115 3.83% : 0.000050s : 5: replace.depend_value_elim 55.35% : 0.000724s : 71: replace.inline 3.64% : 0.000048s : 2: replace.list_to_tuple_eliminator_ 27.57% : 0.000361s : 24: replace.switch_simplify 9.62% : 0.000126s : 13: replace.tuple_list_get_item_eliminator ------[match.] 0.003005 115 0.07% : 0.000002s : 5: match.depend_value_elim 95.23% : 0.002861s : 71: match.inline 2.01% : 0.000060s : 2: match.list_to_tuple_eliminator_ 1.28% : 0.000038s : 24: match.switch_simplify 1.41% : 0.000042s : 13: match.tuple_list_get_item_eliminator ------[predicate.] 0.001905 9701 1.07% : 0.000020s : 123: predicate.accumulaten_eliminater 0.31% : 0.000006s : 17: predicate.ad_related_special_op_eliminate 0.44% : 0.000008s : 58: predicate.addn_check_dump 1.09% : 0.000021s : 123: predicate.addn_zero_filter 0.96% : 0.000018s : 123: predicate.adjust_all_reduce_mul_add 2.11% : 0.000040s : 181: predicate.arithmetic_simplify 1.00% : 0.000019s : 123: predicate.cast_eliminate 0.26% : 0.000005s : 34: predicate.check_bprop_eliminate 0.43% : 0.000008s : 58: predicate.compare_switch_simplify 0.07% : 0.000001s : 17: predicate.const_output_eliminate 0.52% : 0.000010s : 61: predicate.depend_value_elim 1.05% : 0.000020s : 123: predicate.dict_get_item_const_eliminator 1.12% : 0.000021s : 123: predicate.dict_get_item_eliminator 0.97% : 0.000018s : 123: predicate.dict_set_item_eliminator 0.35% : 0.000007s : 34: predicate.dumpgradient_eliminate 0.09% : 0.000002s : 17: predicate.elim_not_effective 0.22% : 0.000004s : 17: predicate.elim_shapecalc_of_broadcastargs 1.07% : 0.000020s : 140: predicate.environ_add_const_eliminate 1.08% : 0.000020s : 140: predicate.environ_get_add_eliminate 1.05% : 0.000020s : 140: predicate.environ_get_depend_swap 1.54% : 0.000029s : 198: predicate.environ_get_eliminate 1.07% : 0.000020s : 140: predicate.environ_get_set_eliminate 1.65% : 0.000031s : 209: predicate.exchange_switch_depend_value 2.39% : 0.000045s : 209: predicate.float_depend_g_call 0.42% : 0.000008s : 58: predicate.float_environ_get_switch 0.62% : 0.000012s : 75: predicate.float_tuple_getitem_switch 0.06% : 0.000001s : 17: predicate.fold_const_symbol 0.30% : 0.000006s : 35: predicate.get_grad_eliminate 0.08% : 0.000002s : 17: predicate.graph_param_transform 0.44% : 0.000008s : 58: predicate.incorporate_call 0.39% : 0.000007s : 58: predicate.incorporate_call_switch 5.32% : 0.000101s : 459: predicate.inline 0.34% : 0.000007s : 35: predicate.inline_without_move 0.13% : 0.000002s : 35: predicate.j_node_and_user_rematch 0.43% : 0.000008s : 36: predicate.less_batch_normalization 1.61% : 0.000031s : 172: predicate.list_to_tuple_eliminator_ 2.40% : 0.000046s : 295: predicate.load_eliminater 0.34% : 0.000006s : 17: predicate.loop_unroll_after_grad 2.85% : 0.000054s : 314: predicate.loop_unroll_before_grad 1.36% : 0.000026s : 157: predicate.make_slice_get_slice_eliminator 0.44% : 0.000008s : 58: predicate.merge_addn 0.25% : 0.000005s : 34: predicate.micro_step_allgather_replace 0.26% : 0.000005s : 34: predicate.mini_step_allgather_replace 0.98% : 0.000019s : 123: predicate.minmaximum_grad 0.47% : 0.000009s : 17: predicate.mutable_eliminate 0.16% : 0.000003s : 17: predicate.opt_reshape 0.16% : 0.000003s : 17: predicate.parallel_virtual_node 2.90% : 0.000055s : 209: predicate.partial_defer_inline 1.30% : 0.000025s : 155: predicate.partial_eliminate 1.04% : 0.000020s : 123: predicate.print_const_string_wrapper 0.41% : 0.000008s : 53: predicate.reduce_all_const_elim 1.37% : 0.000026s : 123: predicate.reduce_eliminate 2.27% : 0.000043s : 295: predicate.redundant_stop_gradient_eliminater 0.14% : 0.000003s : 35: predicate.remove_not_recompute_node 14.20% : 0.000270s : 172: predicate.replace_applicator 0.14% : 0.000003s : 35: predicate.replace_old_param 0.09% : 0.000002s : 17: predicate.reset_defer_inline 1.07% : 0.000020s : 123: predicate.reshape_eliminate 0.28% : 0.000005s : 34: predicate.row_tensor_add_zeros_like 0.18% : 0.000003s : 17: predicate.row_tensor_eliminate 0.52% : 0.000010s : 34: predicate.same_eliminate 0.22% : 0.000004s : 40: predicate.set_cell_output_no_recompute 0.36% : 0.000007s : 35: predicate.shard_identity_eliminate 0.29% : 0.000005s : 34: predicate.special_op_eliminate 0.54% : 0.000010s : 58: predicate.specialize_transform 0.41% : 0.000008s : 34: predicate.split_environ_get_set_with_tuple_value 0.31% : 0.000006s : 35: predicate.stack_unstack_eliminate 0.15% : 0.000003s : 17: predicate.switch_call_monad_eliminater 1.82% : 0.000035s : 209: predicate.switch_defer_inline 2.00% : 0.000038s : 243: predicate.switch_layer_defer_inline 5.75% : 0.000110s : 646: predicate.switch_simplify 1.14% : 0.000022s : 123: predicate.tile_eliminate 1.00% : 0.000019s : 123: predicate.transpose_eliminate 1.52% : 0.000029s : 157: predicate.tuple_list_convert_item_index_to_positive 1.55% : 0.000030s : 157: predicate.tuple_list_get_item_const_eliminator 1.39% : 0.000027s : 157: predicate.tuple_list_get_item_depend_reorder 2.85% : 0.000054s : 228: predicate.tuple_list_get_item_eliminator 1.44% : 0.000028s : 157: predicate.tuple_list_get_set_item_eliminator 2.20% : 0.000042s : 215: predicate.tuple_list_set_item_eliminator 1.44% : 0.000027s : 170: predicate.tuple_to_list_eliminator_ 2.20% : 0.000042s : 295: predicate.updatestate_pure_node_eliminater 2.68% : 0.000051s : 353: predicate.updatestate_useless_node_eliminater 0.15% : 0.000003s : 17: predicate.value_based_eliminate 0.31% : 0.000006s : 35: predicate.virtual_dataset_eliminate 0.28% : 0.000005s : 35: predicate.virtual_output_eliminate 0.12% : 0.000002s : 17: predicate.virtual_view_grad_eliminate 0.21% : 0.000004s : 17: predicate.zero_like_fill_zero ------[func_graph_cloner_run.] 0.020625 135 64.62% : 0.013329s : 55: func_graph_cloner_run.FuncGraphClonerGraph 35.38% : 0.007296s : 80: func_graph_cloner_run.FuncGraphSpecializer ------[meta_graph.] 0.000000 0 ------[manager.] 0.000000 0 ------[pynative] 0.000000 0 ------[others.] 1.022842 209 0.00% : 0.000004s : 1: ForceFp32Comm 1.66% : 0.016980s : 1: add_attr 1.66% : 0.016960s : 1: add_attr_with_inline 0.00% : 0.000005s : 1: add_comm_op_reuse_tag 0.02% : 0.000160s : 1: add_recomputation 0.00% : 0.000006s : 1: assign_add_opt 0.08% : 0.000800s : 1: auto_monad 0.01% : 0.000075s : 1: auto_monad_reorder 0.00% : 0.000005s : 1: begin_end_overlap_inline 0.00% : 0.000006s : 1: bias_add_comm_swap 0.21% : 0.002107s : 1: bootstrap 0.00% : 0.000048s : 1: cconv 0.00% : 0.000005s : 1: comm_op_add_attrs 0.00% : 0.000045s : 1: control_data_broadcast_order 0.00% : 0.000027s : 1: convert_after_rewriter 0.01% : 0.000085s : 1: cse_after_recomputation 0.00% : 0.000006s : 1: dataset_repeat_opt 0.00% : 0.000007s : 1: detach_backward 0.00% : 0.000029s : 1: environ_conv 0.02% : 0.000234s : 1: event_method 0.00% : 0.000007s : 1: full_micro_interleaved_order_control 0.00% : 0.000007s : 1: get_jit_bprop_graph 0.01% : 0.000059s : 1: graph_reusing 0.00% : 0.000005s : 1: grouped_pairwise_exchange_alltoall 0.00% : 0.000005s : 1: handle_group_info 0.00% : 0.000008s : 1: inline 0.00% : 0.000008s : 1: insert-virtual-dataset 0.00% : 0.000004s : 1: interleave_parallel_branches 0.00% : 0.000004s : 1: interleave_split_concat_branches 0.00% : 0.000006s : 1: label_fine_grained_interleaved_index 0.00% : 0.000010s : 1: label_micro_interleaved_index 0.06% : 0.000642s : 1: loop_unroll 0.00% : 0.000006s : 1: merge_cast_opt 0.00% : 0.000005s : 1: micro_interleaved_order_control 0.08% : 0.000857s : 1: mutable_eliminate 0.00% : 0.000014s : 1: offloading_packed_experts 0.00% : 0.000037s : 1: opt.transform.loop_unroll_optimizer 0.00% : 0.000047s : 1: opt.transform.mutable_eliminate 7.56% : 0.077283s : 95: opt.transform.opt_a 0.01% : 0.000128s : 1: opt.transform.opt_after_cconv 0.01% : 0.000079s : 1: opt.transform.opt_after_jit_grad 0.05% : 0.000546s : 28: opt.transform.opt_b 0.02% : 0.000227s : 2: opt.transform.opt_trans_graph 0.02% : 0.000192s : 4: opt.transform.symbol_engine_opt 14.28% : 0.146108s : 1: opt_a 0.03% : 0.000316s : 1: opt_after_cconv 0.08% : 0.000821s : 1: opt_after_jit_grad 0.08% : 0.000789s : 1: opt_b 14.91% : 0.152519s : 1: optimize 0.00% : 0.000046s : 1: optimize_parallel_all_gather_comm 0.00% : 0.000019s : 1: order_py_execute_after_rewriter 0.01% : 0.000056s : 1: overlap_grad_flash_sp 0.00% : 0.000004s : 1: overlap_grad_matmul_and_grad_allreduce 0.00% : 0.000012s : 1: overlap_grad_ring_attention 0.00% : 0.000006s : 1: overlap_opt_shard_grad_in_pipeline 0.00% : 0.000004s : 1: overlap_opt_shard_in_pipeline 0.00% : 0.000006s : 1: overlap_param_gather 0.00% : 0.000004s : 1: overlap_recompute_allgather_and_fa_grad 0.00% : 0.000014s : 1: overlap_recompute_and_grad_model_parallel 0.00% : 0.000006s : 1: overlap_recompute_comm 0.00% : 0.000011s : 1: parallel-infer-symbol 0.00% : 0.000005s : 1: parallel-infer-symbol-second 0.00% : 0.000006s : 1: partial_unused_args_eliminate 0.00% : 0.000005s : 1: pipeline_parallel_scheduler 0.00% : 0.000005s : 1: pipeline_split 0.03% : 0.000267s : 1: pre_auto_parallel 0.02% : 0.000254s : 1: py_interpret_to_execute 0.00% : 0.000045s : 1: py_interpret_to_execute_after_opt_a 0.00% : 0.000004s : 1: remove_cast_before_assign_add 0.01% : 0.000114s : 1: remove_dup_value 4.40% : 0.044999s : 1: renormalize.infer 0.70% : 0.007208s : 1: renormalize.specialize 0.00% : 0.000008s : 1: reorder_send_recv_between_fp_bp 0.00% : 0.000010s : 1: rewriter_after_jit_bprop_graph 0.09% : 0.000890s : 1: rewriter_after_opt_a 0.08% : 0.000813s : 1: rewriter_before_opt_a 0.00% : 0.000007s : 1: slice_cell_reuse_recomputed_activation 0.00% : 0.000005s : 1: slice_recompute_activation 0.00% : 0.000005s : 1: split_layernorm_comm 0.00% : 0.000005s : 1: split_matmul_comm_elemetwise 0.00% : 0.000023s : 1: swap_dp_allreduce_reducescatter 0.06% : 0.000601s : 1: symbol_engine_optimizer 0.03% : 0.000272s : 1: tuple_transform 53.65% : 0.548772s : 1: type_inference . [hook] pytest_runtest_teardown:test_matmul_qkv_bf16_1[True-mstype0-k_n_shape0-256] tests/st/infer/ops/test_internal_ops/test_matmul_split.py::test_matmul_qkv_bf16_1[True-mstype0-k_n_shape0-256],max_mem:252.0M [WARNING] ME(168319:281472866918192,MainProcess):2026-01-29-17:44:25.930.672 [mindspore/context.py:1334] For 'context.set_context', the parameter 'device_target' will be deprecated and removed in a future version. Please use the api mindspore.set_device() instead. TotalTime = 0.75913, [21] [bootstrap]: 0.00069958 [type_inference]: 0.604378 [event_method]: 0.00020447 [auto_monad]: 0.00070358 [graph_reusing]: 4.72e-05 [inline]: 3.04001e-06 [add_attr]: 0.00525574, [1] [add_attr_with_inline]: 0.00523861, [1] [Cycle 1]: 0.00028161, [2] [tag_attr]: 0.00018667 [meta_addattr_fg_expand]: 5.1e-05 [parallel-infer-symbol]: 3.83001e-06 [pre_auto_parallel]: 0.00023577 [insert-virtual-dataset]: 2.65997e-06 [parallel-infer-symbol-second]: 1.15999e-06 [dataset_repeat_opt]: 1.92001e-06 [pipeline_split]: 1.94e-06 [optimize]: 0.146327, [53] [py_interpret_to_execute]: 0.00021014 [rewriter_before_opt_a]: 0.0129685 [opt_a]: 0.0756837, [2] [Cycle 1]: 0.0733791, [45] [expand_dump_flag]: 3.458e-05 [switch_simplify]: 0.00085493 [loop_unroll]: 0.0003183 [a_1]: 0.0559977 [with_stream_mark]: 5.44e-05 [recompute_prepare]: 3.937e-05 [updatestate_depend_eliminate]: 2.227e-05 [updatestate_assign_eliminate]: 1.572e-05 [updatestate_loads_eliminate]: 1.987e-05 [parameter_eliminate]: 2.88e-06 [a_2]: 0.00081272 [accelerated_algorithm]: 6.752e-05 [shard]: 2.74001e-06 [meta_shard_fg_expand]: 2.523e-05 [shard_inline]: 2.371e-05 [merge_send_recv]: 1.992e-05 [auto_parallel]: 2.084e-05 [parallel]: 3.593e-05 [flash_sp]: 1.607e-05 [merge_comm]: 1.449e-05 [allreduce_fusion]: 1.355e-05 [matmul_add_comm_reduction]: 2.136e-05 [allreduce_slice_to_reducescatter]: 9.20001e-07 [virtual_shard_identity]: 2.655e-05 [virtual_dataset]: 2.285e-05 [get_grad_eliminate_]: 2.197e-05 [virtual_output]: 2.181e-05 [merge_forward]: 1.364e-05 [cell_reuse_recompute_pass]: 1.47001e-06 [offload_activation]: 2.304e-05 [cell_reuse_handle_not_recompute_node_pass]: 4.729e-05 [merge_recompute_call_nodes]: 1.54998e-06 [before_grad]: 4.11e-05 [set_forward_comm_id_for_comm_node_pass]: 1.469e-05 [meta_fg_expand]: 1.556e-05 [flash_sp_send_recv_attached]: 6.24999e-06 [receive_attached]: 3.33998e-06 [after_resolve]: 2.938e-05 [a_after_grad]: 3.608e-05 [renormalize]: 0.0119469 [add_forward_monad_depend]: 1.968e-05 [auto_monad_grad]: 3.11001e-06 [auto_monad_eliminator]: 7.896e-05 [cse]: 0.00195063 [a_3]: 0.00018248 [Cycle 2]: 0.00228373, [45] [expand_dump_flag]: 2.91e-06 [switch_simplify]: 2.843e-05 [loop_unroll]: 2.175e-05 [a_1]: 0.00066566 [with_stream_mark]: 4.025e-05 [recompute_prepare]: 2.365e-05 [updatestate_depend_eliminate]: 1.778e-05 [updatestate_assign_eliminate]: 1.176e-05 [updatestate_loads_eliminate]: 1.421e-05 [parameter_eliminate]: 3.40998e-06 [a_2]: 0.00031578 [accelerated_algorithm]: 3.273e-05 [shard]: 3.85998e-06 [meta_shard_fg_expand]: 9.05001e-06 [shard_inline]: 2.21e-05 [merge_send_recv]: 2.172e-05 [auto_parallel]: 2.185e-05 [parallel]: 1.17e-05 [flash_sp]: 5.30999e-06 [merge_comm]: 1.291e-05 [allreduce_fusion]: 1.208e-05 [matmul_add_comm_reduction]: 2.431e-05 [allreduce_slice_to_reducescatter]: 1.64e-06 [virtual_shard_identity]: 2.509e-05 [virtual_dataset]: 2.074e-05 [get_grad_eliminate_]: 2.085e-05 [virtual_output]: 2.052e-05 [merge_forward]: 1.29e-05 [cell_reuse_recompute_pass]: 3.23e-06 [offload_activation]: 2.676e-05 [cell_reuse_handle_not_recompute_node_pass]: 4.331e-05 [merge_recompute_call_nodes]: 1.87999e-06 [before_grad]: 3.692e-05 [set_forward_comm_id_for_comm_node_pass]: 1.385e-05 [meta_fg_expand]: 1.025e-05 [flash_sp_send_recv_attached]: 1.83002e-06 [receive_attached]: 3.4e-06 [after_resolve]: 2.84e-05 [a_after_grad]: 3.403e-05 [renormalize]: 1.10012e-07 [add_forward_monad_depend]: 5.46998e-06 [auto_monad_grad]: 2.73998e-06 [auto_monad_eliminator]: 7.08e-05 [cse]: 0.00016747 [a_3]: 0.0001464 [py_interpret_to_execute_after_opt_a]: 3.395e-05 [slice_cell_reuse_recomputed_activation]: 2.39999e-06 [rewriter_after_opt_a]: 0.0008698 [convert_after_rewriter]: 2.618e-05 [order_py_execute_after_rewriter]: 1.468e-05 [mutable_eliminate]: 0.00078705 [opt_b]: 0.00076246, [1] [Cycle 1]: 0.0007538, [7] [b_1]: 0.00054307 [b_2]: 2.558e-05 [updatestate_depend_eliminate]: 2.027e-05 [updatestate_assign_eliminate]: 1.139e-05 [updatestate_loads_eliminate]: 1.463e-05 [renormalize]: 1.15999e-06 [cse]: 8.926e-05 [optimize_parallel_all_gather_comm]: 4.161e-05 [overlap_param_gather]: 1.98997e-06 [cconv]: 4.265e-05 [loop_unroll]: 0.00054673 [opt_after_cconv]: 0.00030796, [1] [Cycle 1]: 0.00029842, [7] [c_1]: 0.00012906 [parameter_eliminate]: 6.01e-06 [updatestate_depend_eliminate]: 2.073e-05 [updatestate_assign_eliminate]: 1.221e-05 [updatestate_loads_eliminate]: 1.306e-05 [cse]: 7.411e-05 [renormalize]: 6.99976e-07 [remove_dup_value]: 0.00010151 [tuple_transform]: 0.0525009, [1] [Cycle 1]: 0.0524912, [4] [d_1]: 0.0523829 [none_parameter_eliminate]: 6.29001e-06 [renormalize]: 4.59986e-07 [switch_simplify]: 3.514e-05 [partial_unused_args_eliminate]: 3.55003e-06 [add_recomputation]: 0.00018715 [cse_after_recomputation]: 0.00011795, [1] [Cycle 1]: 0.00010721, [1] [cse]: 9.613e-05 [environ_conv]: 2.606e-05 [swap_dp_allreduce_reducescatter]: 1.944e-05 [bias_add_comm_swap]: 3.56999e-06 [label_micro_interleaved_index]: 8.40001e-06 [label_fine_grained_interleaved_index]: 2.82002e-06 [merge_cast_opt]: 1.82999e-06 [slice_recompute_activation]: 2.47001e-06 [micro_interleaved_order_control]: 2.51e-06 [assign_add_opt]: 2.33002e-06 [ForceFp32Comm]: 9.50007e-07 [remove_cast_before_assign_add]: 1.14998e-06 [full_micro_interleaved_order_control]: 2.36e-06 [reorder_send_recv_between_fp_bp]: 2.81999e-06 [comm_op_add_attrs]: 1.13001e-06 [add_comm_op_reuse_tag]: 1.34e-06 [interleave_split_concat_branches]: 2.04e-06 [interleave_parallel_branches]: 1.13001e-06 [overlap_opt_shard_in_pipeline]: 1.38002e-06 [overlap_opt_shard_grad_in_pipeline]: 1.82999e-06 [control_data_broadcast_order]: 4.166e-05 [grouped_pairwise_exchange_alltoall]: 1.76998e-06 [offloading_packed_experts]: 1.125e-05 [overlap_recompute_and_grad_model_parallel]: 1.062e-05 [overlap_grad_matmul_and_grad_allreduce]: 1.35001e-06 [overlap_recompute_allgather_and_fa_grad]: 1.79e-06 [overlap_recompute_comm]: 2.37001e-06 [overlap_grad_ring_attention]: 9.79999e-06 [overlap_grad_flash_sp]: 5.201e-05 [begin_end_overlap_inline]: 5.50004e-07 [split_matmul_comm_elemetwise]: 2.34001e-06 [split_layernorm_comm]: 1.87001e-06 [handle_group_info]: 1.08001e-06 [symbol_engine_optimizer]: 0.00053976, [1] [Cycle 1]: 0.00053169, [6] [build]: 0.0002994 [elim_shapecalc]: 3.497e-05 [elim_not_effective]: 5.35e-05 [opt_reshape]: 3.519e-05 [fold_const_symbol]: 5.8e-05 [renormalize]: 4.99975e-07 [detach_backward]: 3.09001e-06 [pipeline_parallel_scheduler]: 1.65001e-06 [auto_monad_reorder]: 6.594e-05 [get_jit_bprop_graph]: 1.89e-06 [rewriter_after_jit_bprop_graph]: 6.04001e-06 [opt_after_jit_grad]: 0.00083885 [validate]: 9.59e-05 Sums bootstrap : 0.000700s : 0.09% type_inference : 0.604378s : 80.31% event_method : 0.000204s : 0.03% auto_monad : 0.000704s : 0.09% graph_reusing : 0.000047s : 0.01% inline : 0.000003s : 0.00% add_attr.add_attr_with_inline.tag_attr : 0.000187s : 0.02% add_attr.add_attr_with_inline.meta_addattr_fg_expand : 0.000051s : 0.01% parallel-infer-symbol : 0.000004s : 0.00% pre_auto_parallel : 0.000236s : 0.03% insert-virtual-dataset : 0.000003s : 0.00% parallel-infer-symbol-second : 0.000001s : 0.00% dataset_repeat_opt : 0.000002s : 0.00% pipeline_split : 0.000002s : 0.00% optimize.py_interpret_to_execute : 0.000210s : 0.03% optimize.rewriter_before_opt_a : 0.012968s : 1.72% optimize.opt_a.expand_dump_flag : 0.000037s : 0.00% optimize.opt_a.switch_simplify : 0.000883s : 0.12% optimize.opt_a.loop_unroll : 0.000340s : 0.05% optimize.opt_a.a_1 : 0.056663s : 7.53% optimize.opt_a.with_stream_mark : 0.000095s : 0.01% optimize.opt_a.recompute_prepare : 0.000063s : 0.01% optimize.opt_a.updatestate_depend_eliminate : 0.000040s : 0.01% optimize.opt_a.updatestate_assign_eliminate : 0.000027s : 0.00% optimize.opt_a.updatestate_loads_eliminate : 0.000034s : 0.00% optimize.opt_a.parameter_eliminate : 0.000006s : 0.00% optimize.opt_a.a_2 : 0.001129s : 0.15% optimize.opt_a.accelerated_algorithm : 0.000100s : 0.01% optimize.opt_a.shard : 0.000007s : 0.00% optimize.opt_a.meta_shard_fg_expand : 0.000034s : 0.00% optimize.opt_a.shard_inline : 0.000046s : 0.01% optimize.opt_a.merge_send_recv : 0.000042s : 0.01% optimize.opt_a.auto_parallel : 0.000043s : 0.01% optimize.opt_a.parallel : 0.000048s : 0.01% optimize.opt_a.flash_sp : 0.000021s : 0.00% optimize.opt_a.merge_comm : 0.000027s : 0.00% optimize.opt_a.allreduce_fusion : 0.000026s : 0.00% optimize.opt_a.matmul_add_comm_reduction : 0.000046s : 0.01% optimize.opt_a.allreduce_slice_to_reducescatter : 0.000003s : 0.00% optimize.opt_a.virtual_shard_identity : 0.000052s : 0.01% optimize.opt_a.virtual_dataset : 0.000044s : 0.01% optimize.opt_a.get_grad_eliminate_ : 0.000043s : 0.01% optimize.opt_a.virtual_output : 0.000042s : 0.01% optimize.opt_a.merge_forward : 0.000027s : 0.00% optimize.opt_a.cell_reuse_recompute_pass : 0.000005s : 0.00% optimize.opt_a.offload_activation : 0.000050s : 0.01% optimize.opt_a.cell_reuse_handle_not_recompute_node_pass : 0.000091s : 0.01% optimize.opt_a.merge_recompute_call_nodes : 0.000003s : 0.00% optimize.opt_a.before_grad : 0.000078s : 0.01% optimize.opt_a.set_forward_comm_id_for_comm_node_pass : 0.000029s : 0.00% optimize.opt_a.meta_fg_expand : 0.000026s : 0.00% optimize.opt_a.flash_sp_send_recv_attached : 0.000008s : 0.00% optimize.opt_a.receive_attached : 0.000007s : 0.00% optimize.opt_a.after_resolve : 0.000058s : 0.01% optimize.opt_a.a_after_grad : 0.000070s : 0.01% optimize.opt_a.renormalize : 0.011947s : 1.59% optimize.opt_a.add_forward_monad_depend : 0.000025s : 0.00% optimize.opt_a.auto_monad_grad : 0.000006s : 0.00% optimize.opt_a.auto_monad_eliminator : 0.000150s : 0.02% optimize.opt_a.cse : 0.002118s : 0.28% optimize.opt_a.a_3 : 0.000329s : 0.04% optimize.py_interpret_to_execute_after_opt_a : 0.000034s : 0.00% optimize.slice_cell_reuse_recomputed_activation : 0.000002s : 0.00% optimize.rewriter_after_opt_a : 0.000870s : 0.12% optimize.convert_after_rewriter : 0.000026s : 0.00% optimize.order_py_execute_after_rewriter : 0.000015s : 0.00% optimize.mutable_eliminate : 0.000787s : 0.10% optimize.opt_b.b_1 : 0.000543s : 0.07% optimize.opt_b.b_2 : 0.000026s : 0.00% optimize.opt_b.updatestate_depend_eliminate : 0.000020s : 0.00% optimize.opt_b.updatestate_assign_eliminate : 0.000011s : 0.00% optimize.opt_b.updatestate_loads_eliminate : 0.000015s : 0.00% optimize.opt_b.renormalize : 0.000001s : 0.00% optimize.opt_b.cse : 0.000089s : 0.01% optimize.optimize_parallel_all_gather_comm : 0.000042s : 0.01% optimize.overlap_param_gather : 0.000002s : 0.00% optimize.cconv : 0.000043s : 0.01% optimize.loop_unroll : 0.000547s : 0.07% optimize.opt_after_cconv.c_1 : 0.000129s : 0.02% optimize.opt_after_cconv.parameter_eliminate : 0.000006s : 0.00% optimize.opt_after_cconv.updatestate_depend_eliminate : 0.000021s : 0.00% optimize.opt_after_cconv.updatestate_assign_eliminate : 0.000012s : 0.00% optimize.opt_after_cconv.updatestate_loads_eliminate : 0.000013s : 0.00% optimize.opt_after_cconv.cse : 0.000074s : 0.01% optimize.opt_after_cconv.renormalize : 0.000001s : 0.00% optimize.remove_dup_value : 0.000102s : 0.01% optimize.tuple_transform.d_1 : 0.052383s : 6.96% optimize.tuple_transform.none_parameter_eliminate : 0.000006s : 0.00% optimize.tuple_transform.renormalize : 0.000000s : 0.00% optimize.tuple_transform.switch_simplify : 0.000035s : 0.00% optimize.partial_unused_args_eliminate : 0.000004s : 0.00% optimize.add_recomputation : 0.000187s : 0.02% optimize.cse_after_recomputation.cse : 0.000096s : 0.01% optimize.environ_conv : 0.000026s : 0.00% optimize.swap_dp_allreduce_reducescatter : 0.000019s : 0.00% optimize.bias_add_comm_swap : 0.000004s : 0.00% optimize.label_micro_interleaved_index : 0.000008s : 0.00% optimize.label_fine_grained_interleaved_index : 0.000003s : 0.00% optimize.merge_cast_opt : 0.000002s : 0.00% optimize.slice_recompute_activation : 0.000002s : 0.00% optimize.micro_interleaved_order_control : 0.000003s : 0.00% optimize.assign_add_opt : 0.000002s : 0.00% optimize.ForceFp32Comm : 0.000001s : 0.00% optimize.remove_cast_before_assign_add : 0.000001s : 0.00% optimize.full_micro_interleaved_order_control : 0.000002s : 0.00% optimize.reorder_send_recv_between_fp_bp : 0.000003s : 0.00% optimize.comm_op_add_attrs : 0.000001s : 0.00% optimize.add_comm_op_reuse_tag : 0.000001s : 0.00% optimize.interleave_split_concat_branches : 0.000002s : 0.00% optimize.interleave_parallel_branches : 0.000001s : 0.00% optimize.overlap_opt_shard_in_pipeline : 0.000001s : 0.00% optimize.overlap_opt_shard_grad_in_pipeline : 0.000002s : 0.00% optimize.control_data_broadcast_order : 0.000042s : 0.01% optimize.grouped_pairwise_exchange_alltoall : 0.000002s : 0.00% optimize.offloading_packed_experts : 0.000011s : 0.00% optimize.overlap_recompute_and_grad_model_parallel : 0.000011s : 0.00% optimize.overlap_grad_matmul_and_grad_allreduce : 0.000001s : 0.00% optimize.overlap_recompute_allgather_and_fa_grad : 0.000002s : 0.00% optimize.overlap_recompute_comm : 0.000002s : 0.00% optimize.overlap_grad_ring_attention : 0.000010s : 0.00% optimize.overlap_grad_flash_sp : 0.000052s : 0.01% optimize.begin_end_overlap_inline : 0.000001s : 0.00% optimize.split_matmul_comm_elemetwise : 0.000002s : 0.00% optimize.split_layernorm_comm : 0.000002s : 0.00% optimize.handle_group_info : 0.000001s : 0.00% optimize.symbol_engine_optimizer.build : 0.000299s : 0.04% optimize.symbol_engine_optimizer.elim_shapecalc : 0.000035s : 0.00% optimize.symbol_engine_optimizer.elim_not_effective : 0.000053s : 0.01% optimize.symbol_engine_optimizer.opt_reshape : 0.000035s : 0.00% optimize.symbol_engine_optimizer.fold_const_symbol : 0.000058s : 0.01% optimize.symbol_engine_optimizer.renormalize : 0.000000s : 0.00% detach_backward : 0.000003s : 0.00% pipeline_parallel_scheduler : 0.000002s : 0.00% auto_monad_reorder : 0.000066s : 0.01% get_jit_bprop_graph : 0.000002s : 0.00% rewriter_after_jit_bprop_graph : 0.000006s : 0.00% opt_after_jit_grad : 0.000839s : 0.11% validate : 0.000096s : 0.01% Time group info: ------[substitution.] 0.003796 569 0.44% : 0.000017s : 8: substitution.depend_value_elim 0.34% : 0.000013s : 14: substitution.elim_not_effective 0.49% : 0.000019s : 12: substitution.float_tuple_getitem_switch 0.70% : 0.000027s : 14: substitution.fold_const_symbol 0.74% : 0.000028s : 17: substitution.graph_param_transform 67.43% : 0.002560s : 71: substitution.inline 0.36% : 0.000014s : 29: substitution.j_node_and_user_rematch 0.97% : 0.000037s : 6: substitution.less_batch_normalization 1.60% : 0.000061s : 2: substitution.list_to_tuple_eliminator_ 0.16% : 0.000006s : 6: substitution.load_eliminater 0.91% : 0.000034s : 21: substitution.minmaximum_grad 0.23% : 0.000009s : 2: substitution.opt_reshape 0.55% : 0.000021s : 29: substitution.remove_not_recompute_node 0.18% : 0.000007s : 4: substitution.replace_old_param 1.87% : 0.000071s : 18: substitution.reshape_eliminate 1.36% : 0.000052s : 24: substitution.switch_simplify 4.90% : 0.000186s : 51: substitution.tuple_list_convert_item_index_to_positive 2.50% : 0.000095s : 51: substitution.tuple_list_get_item_const_eliminator 3.29% : 0.000125s : 51: substitution.tuple_list_get_item_depend_reorder 6.47% : 0.000246s : 73: substitution.tuple_list_get_item_eliminator 3.32% : 0.000126s : 51: substitution.tuple_list_get_set_item_eliminator 0.61% : 0.000023s : 6: substitution.updatestate_pure_node_eliminater 0.58% : 0.000022s : 9: substitution.updatestate_useless_node_eliminater ------[type_inference.] 0.603953 2 97.09% : 0.586356s : 1: type_inference.infer 2.91% : 0.017597s : 1: type_inference.specialize ------[replace.] 0.001232 115 3.53% : 0.000043s : 5: replace.depend_value_elim 58.64% : 0.000722s : 71: replace.inline 3.50% : 0.000043s : 2: replace.list_to_tuple_eliminator_ 24.86% : 0.000306s : 24: replace.switch_simplify 9.47% : 0.000117s : 13: replace.tuple_list_get_item_eliminator ------[match.] 0.002648 115 0.09% : 0.000002s : 5: match.depend_value_elim 94.91% : 0.002513s : 71: match.inline 2.20% : 0.000058s : 2: match.list_to_tuple_eliminator_ 1.35% : 0.000036s : 24: match.switch_simplify 1.45% : 0.000039s : 13: match.tuple_list_get_item_eliminator ------[predicate.] 0.001673 9701 1.18% : 0.000020s : 123: predicate.accumulaten_eliminater 0.38% : 0.000006s : 17: predicate.ad_related_special_op_eliminate 0.49% : 0.000008s : 58: predicate.addn_check_dump 1.50% : 0.000025s : 123: predicate.addn_zero_filter 1.07% : 0.000018s : 123: predicate.adjust_all_reduce_mul_add 2.28% : 0.000038s : 181: predicate.arithmetic_simplify 1.14% : 0.000019s : 123: predicate.cast_eliminate 0.31% : 0.000005s : 34: predicate.check_bprop_eliminate 0.49% : 0.000008s : 58: predicate.compare_switch_simplify 0.08% : 0.000001s : 17: predicate.const_output_eliminate 0.61% : 0.000010s : 61: predicate.depend_value_elim 1.15% : 0.000019s : 123: predicate.dict_get_item_const_eliminator 1.24% : 0.000021s : 123: predicate.dict_get_item_eliminator 1.14% : 0.000019s : 123: predicate.dict_set_item_eliminator 0.37% : 0.000006s : 34: predicate.dumpgradient_eliminate 0.10% : 0.000002s : 17: predicate.elim_not_effective 0.21% : 0.000004s : 17: predicate.elim_shapecalc_of_broadcastargs 1.32% : 0.000022s : 140: predicate.environ_add_const_eliminate 1.21% : 0.000020s : 140: predicate.environ_get_add_eliminate 1.21% : 0.000020s : 140: predicate.environ_get_depend_swap 1.73% : 0.000029s : 198: predicate.environ_get_eliminate 1.25% : 0.000021s : 140: predicate.environ_get_set_eliminate 1.85% : 0.000031s : 209: predicate.exchange_switch_depend_value 2.86% : 0.000048s : 209: predicate.float_depend_g_call 0.49% : 0.000008s : 58: predicate.float_environ_get_switch 0.71% : 0.000012s : 75: predicate.float_tuple_getitem_switch 0.07% : 0.000001s : 17: predicate.fold_const_symbol 0.33% : 0.000005s : 35: predicate.get_grad_eliminate 0.15% : 0.000002s : 17: predicate.graph_param_transform 0.50% : 0.000008s : 58: predicate.incorporate_call 0.45% : 0.000008s : 58: predicate.incorporate_call_switch 6.04% : 0.000101s : 459: predicate.inline 0.43% : 0.000007s : 35: predicate.inline_without_move 0.15% : 0.000003s : 35: predicate.j_node_and_user_rematch 0.51% : 0.000009s : 36: predicate.less_batch_normalization 1.75% : 0.000029s : 172: predicate.list_to_tuple_eliminator_ 2.74% : 0.000046s : 295: predicate.load_eliminater 0.42% : 0.000007s : 17: predicate.loop_unroll_after_grad 3.01% : 0.000050s : 314: predicate.loop_unroll_before_grad 1.50% : 0.000025s : 157: predicate.make_slice_get_slice_eliminator 0.52% : 0.000009s : 58: predicate.merge_addn 0.29% : 0.000005s : 34: predicate.micro_step_allgather_replace 0.30% : 0.000005s : 34: predicate.mini_step_allgather_replace 1.20% : 0.000020s : 123: predicate.minmaximum_grad 0.41% : 0.000007s : 17: predicate.mutable_eliminate 0.17% : 0.000003s : 17: predicate.opt_reshape 0.20% : 0.000003s : 17: predicate.parallel_virtual_node 3.29% : 0.000055s : 209: predicate.partial_defer_inline 1.49% : 0.000025s : 155: predicate.partial_eliminate 1.13% : 0.000019s : 123: predicate.print_const_string_wrapper 0.47% : 0.000008s : 53: predicate.reduce_all_const_elim 1.57% : 0.000026s : 123: predicate.reduce_eliminate 2.60% : 0.000043s : 295: predicate.redundant_stop_gradient_eliminater 0.18% : 0.000003s : 35: predicate.remove_not_recompute_node 1.07% : 0.000018s : 172: predicate.replace_applicator 0.17% : 0.000003s : 35: predicate.replace_old_param 0.10% : 0.000002s : 17: predicate.reset_defer_inline 1.28% : 0.000021s : 123: predicate.reshape_eliminate 0.32% : 0.000005s : 34: predicate.row_tensor_add_zeros_like 0.21% : 0.000003s : 17: predicate.row_tensor_eliminate 0.54% : 0.000009s : 34: predicate.same_eliminate 0.23% : 0.000004s : 40: predicate.set_cell_output_no_recompute 0.39% : 0.000006s : 35: predicate.shard_identity_eliminate 0.32% : 0.000005s : 34: predicate.special_op_eliminate 0.60% : 0.000010s : 58: predicate.specialize_transform 0.41% : 0.000007s : 34: predicate.split_environ_get_set_with_tuple_value 0.37% : 0.000006s : 35: predicate.stack_unstack_eliminate 0.17% : 0.000003s : 17: predicate.switch_call_monad_eliminater 2.10% : 0.000035s : 209: predicate.switch_defer_inline 2.38% : 0.000040s : 243: predicate.switch_layer_defer_inline 7.95% : 0.000133s : 646: predicate.switch_simplify 1.23% : 0.000021s : 123: predicate.tile_eliminate 1.16% : 0.000019s : 123: predicate.transpose_eliminate 1.64% : 0.000027s : 157: predicate.tuple_list_convert_item_index_to_positive 1.76% : 0.000029s : 157: predicate.tuple_list_get_item_const_eliminator 1.65% : 0.000028s : 157: predicate.tuple_list_get_item_depend_reorder 3.38% : 0.000056s : 228: predicate.tuple_list_get_item_eliminator 1.58% : 0.000026s : 157: predicate.tuple_list_get_set_item_eliminator 2.27% : 0.000038s : 215: predicate.tuple_list_set_item_eliminator 1.53% : 0.000026s : 170: predicate.tuple_to_list_eliminator_ 2.49% : 0.000042s : 295: predicate.updatestate_pure_node_eliminater 3.23% : 0.000054s : 353: predicate.updatestate_useless_node_eliminater 0.20% : 0.000003s : 17: predicate.value_based_eliminate 0.33% : 0.000006s : 35: predicate.virtual_dataset_eliminate 0.35% : 0.000006s : 35: predicate.virtual_output_eliminate 0.15% : 0.000003s : 17: predicate.virtual_view_grad_eliminate 0.19% : 0.000003s : 17: predicate.zero_like_fill_zero ------[func_graph_cloner_run.] 0.015027 135 58.94% : 0.008857s : 55: func_graph_cloner_run.FuncGraphClonerGraph 41.06% : 0.006171s : 80: func_graph_cloner_run.FuncGraphSpecializer ------[meta_graph.] 0.000000 0 ------[manager.] 0.000000 0 ------[pynative] 0.000000 0 ------[others.] 1.035671 209 0.00% : 0.000004s : 1: ForceFp32Comm 0.51% : 0.005262s : 1: add_attr 0.51% : 0.005244s : 1: add_attr_with_inline 0.00% : 0.000006s : 1: add_comm_op_reuse_tag 0.02% : 0.000194s : 1: add_recomputation 0.00% : 0.000006s : 1: assign_add_opt 0.07% : 0.000722s : 1: auto_monad 0.01% : 0.000072s : 1: auto_monad_reorder 0.00% : 0.000005s : 1: begin_end_overlap_inline 0.00% : 0.000007s : 1: bias_add_comm_swap 0.07% : 0.000748s : 1: bootstrap 0.00% : 0.000047s : 1: cconv 0.00% : 0.000004s : 1: comm_op_add_attrs 0.00% : 0.000046s : 1: control_data_broadcast_order 0.00% : 0.000032s : 1: convert_after_rewriter 0.01% : 0.000121s : 1: cse_after_recomputation 0.00% : 0.000005s : 1: dataset_repeat_opt 0.00% : 0.000006s : 1: detach_backward 0.00% : 0.000030s : 1: environ_conv 0.02% : 0.000218s : 1: event_method 0.00% : 0.000007s : 1: full_micro_interleaved_order_control 0.00% : 0.000006s : 1: get_jit_bprop_graph 0.01% : 0.000054s : 1: graph_reusing 0.00% : 0.000005s : 1: grouped_pairwise_exchange_alltoall 0.00% : 0.000006s : 1: handle_group_info 0.00% : 0.000007s : 1: inline 0.00% : 0.000007s : 1: insert-virtual-dataset 0.00% : 0.000004s : 1: interleave_parallel_branches 0.00% : 0.000005s : 1: interleave_split_concat_branches 0.00% : 0.000006s : 1: label_fine_grained_interleaved_index 0.00% : 0.000012s : 1: label_micro_interleaved_index 0.05% : 0.000559s : 1: loop_unroll 0.00% : 0.000007s : 1: merge_cast_opt 0.00% : 0.000005s : 1: micro_interleaved_order_control 0.08% : 0.000799s : 1: mutable_eliminate 0.00% : 0.000014s : 1: offloading_packed_experts 0.00% : 0.000039s : 1: opt.transform.loop_unroll_optimizer 0.00% : 0.000044s : 1: opt.transform.mutable_eliminate 5.78% : 0.059872s : 95: opt.transform.opt_a 0.01% : 0.000127s : 1: opt.transform.opt_after_cconv 0.01% : 0.000078s : 1: opt.transform.opt_after_jit_grad 0.05% : 0.000531s : 28: opt.transform.opt_b 5.06% : 0.052407s : 2: opt.transform.opt_trans_graph 0.02% : 0.000176s : 4: opt.transform.symbol_engine_opt 7.31% : 0.075690s : 1: opt_a 0.03% : 0.000312s : 1: opt_after_cconv 0.08% : 0.000851s : 1: opt_after_jit_grad 0.07% : 0.000766s : 1: opt_b 14.13% : 0.146334s : 1: optimize 0.00% : 0.000046s : 1: optimize_parallel_all_gather_comm 0.00% : 0.000019s : 1: order_py_execute_after_rewriter 0.01% : 0.000057s : 1: overlap_grad_flash_sp 0.00% : 0.000004s : 1: overlap_grad_matmul_and_grad_allreduce 0.00% : 0.000013s : 1: overlap_grad_ring_attention 0.00% : 0.000006s : 1: overlap_opt_shard_grad_in_pipeline 0.00% : 0.000004s : 1: overlap_opt_shard_in_pipeline 0.00% : 0.000006s : 1: overlap_param_gather 0.00% : 0.000004s : 1: overlap_recompute_allgather_and_fa_grad 0.00% : 0.000014s : 1: overlap_recompute_and_grad_model_parallel 0.00% : 0.000006s : 1: overlap_recompute_comm 0.00% : 0.000009s : 1: parallel-infer-symbol 0.00% : 0.000005s : 1: parallel-infer-symbol-second 0.00% : 0.000009s : 1: partial_unused_args_eliminate 0.00% : 0.000005s : 1: pipeline_parallel_scheduler 0.00% : 0.000005s : 1: pipeline_split 0.02% : 0.000245s : 1: pre_auto_parallel 0.02% : 0.000220s : 1: py_interpret_to_execute 0.00% : 0.000038s : 1: py_interpret_to_execute_after_opt_a 0.00% : 0.000004s : 1: remove_cast_before_assign_add 0.01% : 0.000109s : 1: remove_dup_value 0.60% : 0.006191s : 1: renormalize.infer 0.55% : 0.005736s : 1: renormalize.specialize 0.00% : 0.000008s : 1: reorder_send_recv_between_fp_bp 0.00% : 0.000009s : 1: rewriter_after_jit_bprop_graph 0.09% : 0.000882s : 1: rewriter_after_opt_a 1.25% : 0.012996s : 1: rewriter_before_opt_a 0.00% : 0.000005s : 1: slice_cell_reuse_recomputed_activation 0.00% : 0.000005s : 1: slice_recompute_activation 0.00% : 0.000004s : 1: split_layernorm_comm 0.00% : 0.000005s : 1: split_matmul_comm_elemetwise 0.00% : 0.000023s : 1: swap_dp_allreduce_reducescatter 0.05% : 0.000543s : 1: symbol_engine_optimizer 5.07% : 0.052509s : 1: tuple_transform 58.36% : 0.604417s : 1: type_inference . [hook] pytest_runtest_teardown:test_matmul_qkv_bf16_1[True-mstype0-k_n_shape0-1024] tests/st/infer/ops/test_internal_ops/test_matmul_split.py::test_matmul_qkv_bf16_1[True-mstype0-k_n_shape0-1024],max_mem:208.0M [WARNING] ME(168319:281472866918192,MainProcess):2026-01-29-17:44:34.121.939 [mindspore/context.py:1334] For 'context.set_context', the parameter 'device_target' will be deprecated and removed in a future version. Please use the api mindspore.set_device() instead. TotalTime = 0.349656, [21] [bootstrap]: 0.00067554 [type_inference]: 0.307762 [event_method]: 0.00020791 [auto_monad]: 0.00071358 [graph_reusing]: 4.648e-05 [inline]: 3.30003e-06 [add_attr]: 0.00546533, [1] [add_attr_with_inline]: 0.005447, [1] [Cycle 1]: 0.00029423, [2] [tag_attr]: 0.00019396 [meta_addattr_fg_expand]: 5.269e-05 [parallel-infer-symbol]: 4.15999e-06 [pre_auto_parallel]: 0.00024702 [insert-virtual-dataset]: 3.06001e-06 [parallel-infer-symbol-second]: 1.32e-06 [dataset_repeat_opt]: 2.19001e-06 [pipeline_split]: 1.76e-06 [optimize]: 0.0335027, [53] [py_interpret_to_execute]: 0.00020522 [rewriter_before_opt_a]: 0.00078183 [opt_a]: 0.0277034, [2] [Cycle 1]: 0.0254112, [45] [expand_dump_flag]: 1.729e-05 [switch_simplify]: 0.00082441 [loop_unroll]: 0.00040615 [a_1]: 0.00926937 [with_stream_mark]: 5.014e-05 [recompute_prepare]: 3.595e-05 [updatestate_depend_eliminate]: 1.952e-05 [updatestate_assign_eliminate]: 1.492e-05 [updatestate_loads_eliminate]: 1.986e-05 [parameter_eliminate]: 2.92002e-06 [a_2]: 0.00076321 [accelerated_algorithm]: 6.292e-05 [shard]: 2.47001e-06 [meta_shard_fg_expand]: 1.616e-05 [shard_inline]: 2.231e-05 [merge_send_recv]: 1.978e-05 [auto_parallel]: 1.99e-05 [parallel]: 3.509e-05 [flash_sp]: 1.505e-05 [merge_comm]: 1.424e-05 [allreduce_fusion]: 1.319e-05 [matmul_add_comm_reduction]: 2.204e-05 [allreduce_slice_to_reducescatter]: 1.13001e-06 [virtual_shard_identity]: 2.387e-05 [virtual_dataset]: 2.272e-05 [get_grad_eliminate_]: 2.117e-05 [virtual_output]: 2.08e-05 [merge_forward]: 1.312e-05 [cell_reuse_recompute_pass]: 1.91e-06 [offload_activation]: 2.342e-05 [cell_reuse_handle_not_recompute_node_pass]: 4.737e-05 [merge_recompute_call_nodes]: 1.60999e-06 [before_grad]: 3.934e-05 [set_forward_comm_id_for_comm_node_pass]: 1.422e-05 [meta_fg_expand]: 1.522e-05 [flash_sp_send_recv_attached]: 6.29001e-06 [receive_attached]: 2.83e-06 [after_resolve]: 2.99e-05 [a_after_grad]: 3.528e-05 [renormalize]: 0.0111733 [add_forward_monad_depend]: 2.083e-05 [auto_monad_grad]: 2.90998e-06 [auto_monad_eliminator]: 7.861e-05 [cse]: 0.00157016 [a_3]: 0.00017703 [Cycle 2]: 0.00227098, [45] [expand_dump_flag]: 2.96001e-06 [switch_simplify]: 2.804e-05 [loop_unroll]: 2.312e-05 [a_1]: 0.00064837 [with_stream_mark]: 2.84e-05 [recompute_prepare]: 2.339e-05 [updatestate_depend_eliminate]: 1.611e-05 [updatestate_assign_eliminate]: 1.183e-05 [updatestate_loads_eliminate]: 1.507e-05 [parameter_eliminate]: 2.17001e-06 [a_2]: 0.00032413 [accelerated_algorithm]: 3.296e-05 [shard]: 2.35002e-06 [meta_shard_fg_expand]: 6.95002e-06 [shard_inline]: 2.352e-05 [merge_send_recv]: 1.952e-05 [auto_parallel]: 1.995e-05 [parallel]: 9.39998e-06 [flash_sp]: 4.34997e-06 [merge_comm]: 1.281e-05 [allreduce_fusion]: 1.175e-05 [matmul_add_comm_reduction]: 2.228e-05 [allreduce_slice_to_reducescatter]: 1.38002e-06 [virtual_shard_identity]: 2.298e-05 [virtual_dataset]: 2.024e-05 [get_grad_eliminate_]: 2.357e-05 [virtual_output]: 2.122e-05 [merge_forward]: 1.282e-05 [cell_reuse_recompute_pass]: 3.33e-06 [offload_activation]: 2.442e-05 [cell_reuse_handle_not_recompute_node_pass]: 4.286e-05 [merge_recompute_call_nodes]: 1.49998e-06 [before_grad]: 3.714e-05 [set_forward_comm_id_for_comm_node_pass]: 1.319e-05 [meta_fg_expand]: 9.96e-06 [flash_sp_send_recv_attached]: 1.97001e-06 [receive_attached]: 3.21999e-06 [after_resolve]: 2.927e-05 [a_after_grad]: 3.369e-05 [renormalize]: 8.9989e-08 [add_forward_monad_depend]: 2.51998e-06 [auto_monad_grad]: 1.81e-06 [auto_monad_eliminator]: 4.76e-05 [cse]: 7.623e-05 [a_3]: 0.00014541 [py_interpret_to_execute_after_opt_a]: 4.069e-05 [slice_cell_reuse_recomputed_activation]: 2.42001e-06 [rewriter_after_opt_a]: 0.00085196 [convert_after_rewriter]: 2.243e-05 [order_py_execute_after_rewriter]: 1.487e-05 [mutable_eliminate]: 0.00079996 [opt_b]: 0.00072473, [1] [Cycle 1]: 0.0007162, [7] [b_1]: 0.00053257 [b_2]: 2.289e-05 [updatestate_depend_eliminate]: 1.668e-05 [updatestate_assign_eliminate]: 1.099e-05 [updatestate_loads_eliminate]: 1.313e-05 [renormalize]: 5.59987e-07 [cse]: 7.666e-05 [optimize_parallel_all_gather_comm]: 3.653e-05 [overlap_param_gather]: 2.42001e-06 [cconv]: 3.739e-05 [loop_unroll]: 0.00048995 [opt_after_cconv]: 0.00027885, [1] [Cycle 1]: 0.00027067, [7] [c_1]: 0.00012214 [parameter_eliminate]: 2.93e-06 [updatestate_depend_eliminate]: 1.664e-05 [updatestate_assign_eliminate]: 1.1e-05 [updatestate_loads_eliminate]: 1.224e-05 [cse]: 6.767e-05 [renormalize]: 3.10014e-07 [remove_dup_value]: 8.884e-05 [tuple_transform]: 0.00024594, [1] [Cycle 1]: 0.00023982, [4] [d_1]: 0.0001949 [none_parameter_eliminate]: 2.36998e-06 [renormalize]: 2.69996e-07 [switch_simplify]: 2.187e-05 [partial_unused_args_eliminate]: 1.84998e-06 [add_recomputation]: 0.00013499 [cse_after_recomputation]: 6.656e-05, [1] [Cycle 1]: 6.033e-05, [1] [cse]: 5.375e-05 [environ_conv]: 1.868e-05 [swap_dp_allreduce_reducescatter]: 1.776e-05 [bias_add_comm_swap]: 2.95998e-06 [label_micro_interleaved_index]: 4.69002e-06 [label_fine_grained_interleaved_index]: 2.86e-06 [merge_cast_opt]: 1.42999e-06 [slice_recompute_activation]: 2.14999e-06 [micro_interleaved_order_control]: 3.41001e-06 [assign_add_opt]: 1.32e-06 [ForceFp32Comm]: 8.39995e-07 [remove_cast_before_assign_add]: 1.14e-06 [full_micro_interleaved_order_control]: 2.27999e-06 [reorder_send_recv_between_fp_bp]: 2.79999e-06 [comm_op_add_attrs]: 1.05999e-06 [add_comm_op_reuse_tag]: 1.05999e-06 [interleave_split_concat_branches]: 1.35999e-06 [interleave_parallel_branches]: 1.15999e-06 [overlap_opt_shard_in_pipeline]: 1.59e-06 [overlap_opt_shard_grad_in_pipeline]: 1.71e-06 [control_data_broadcast_order]: 3.574e-05 [grouped_pairwise_exchange_alltoall]: 2.04999e-06 [offloading_packed_experts]: 1.095e-05 [overlap_recompute_and_grad_model_parallel]: 1.01e-05 [overlap_grad_matmul_and_grad_allreduce]: 1.30999e-06 [overlap_recompute_allgather_and_fa_grad]: 1.47999e-06 [overlap_recompute_comm]: 2.71999e-06 [overlap_grad_ring_attention]: 9.14e-06 [overlap_grad_flash_sp]: 4.94e-05 [begin_end_overlap_inline]: 6.29982e-07 [split_matmul_comm_elemetwise]: 2.21998e-06 [split_layernorm_comm]: 2.01e-06 [handle_group_info]: 1.00001e-06 [symbol_engine_optimizer]: 0.00047904, [1] [Cycle 1]: 0.00047309, [6] [build]: 0.00027038 [elim_shapecalc]: 2.987e-05 [elim_not_effective]: 4.738e-05 [opt_reshape]: 3.151e-05 [fold_const_symbol]: 5.423e-05 [renormalize]: 3.29979e-07 [detach_backward]: 2.69999e-06 [pipeline_parallel_scheduler]: 1.83002e-06 [auto_monad_reorder]: 6.028e-05 [get_jit_bprop_graph]: 1.90001e-06 [rewriter_after_jit_bprop_graph]: 4.98001e-06 [opt_after_jit_grad]: 0.00062088 [validate]: 9.817e-05 Sums bootstrap : 0.000676s : 0.20% type_inference : 0.307762s : 89.75% event_method : 0.000208s : 0.06% auto_monad : 0.000714s : 0.21% graph_reusing : 0.000046s : 0.01% inline : 0.000003s : 0.00% add_attr.add_attr_with_inline.tag_attr : 0.000194s : 0.06% add_attr.add_attr_with_inline.meta_addattr_fg_expand : 0.000053s : 0.02% parallel-infer-symbol : 0.000004s : 0.00% pre_auto_parallel : 0.000247s : 0.07% insert-virtual-dataset : 0.000003s : 0.00% parallel-infer-symbol-second : 0.000001s : 0.00% dataset_repeat_opt : 0.000002s : 0.00% pipeline_split : 0.000002s : 0.00% optimize.py_interpret_to_execute : 0.000205s : 0.06% optimize.rewriter_before_opt_a : 0.000782s : 0.23% optimize.opt_a.expand_dump_flag : 0.000020s : 0.01% optimize.opt_a.switch_simplify : 0.000852s : 0.25% optimize.opt_a.loop_unroll : 0.000429s : 0.13% optimize.opt_a.a_1 : 0.009918s : 2.89% optimize.opt_a.with_stream_mark : 0.000079s : 0.02% optimize.opt_a.recompute_prepare : 0.000059s : 0.02% optimize.opt_a.updatestate_depend_eliminate : 0.000036s : 0.01% optimize.opt_a.updatestate_assign_eliminate : 0.000027s : 0.01% optimize.opt_a.updatestate_loads_eliminate : 0.000035s : 0.01% optimize.opt_a.parameter_eliminate : 0.000005s : 0.00% optimize.opt_a.a_2 : 0.001087s : 0.32% optimize.opt_a.accelerated_algorithm : 0.000096s : 0.03% optimize.opt_a.shard : 0.000005s : 0.00% optimize.opt_a.meta_shard_fg_expand : 0.000023s : 0.01% optimize.opt_a.shard_inline : 0.000046s : 0.01% optimize.opt_a.merge_send_recv : 0.000039s : 0.01% optimize.opt_a.auto_parallel : 0.000040s : 0.01% optimize.opt_a.parallel : 0.000044s : 0.01% optimize.opt_a.flash_sp : 0.000019s : 0.01% optimize.opt_a.merge_comm : 0.000027s : 0.01% optimize.opt_a.allreduce_fusion : 0.000025s : 0.01% optimize.opt_a.matmul_add_comm_reduction : 0.000044s : 0.01% optimize.opt_a.allreduce_slice_to_reducescatter : 0.000003s : 0.00% optimize.opt_a.virtual_shard_identity : 0.000047s : 0.01% optimize.opt_a.virtual_dataset : 0.000043s : 0.01% optimize.opt_a.get_grad_eliminate_ : 0.000045s : 0.01% optimize.opt_a.virtual_output : 0.000042s : 0.01% optimize.opt_a.merge_forward : 0.000026s : 0.01% optimize.opt_a.cell_reuse_recompute_pass : 0.000005s : 0.00% optimize.opt_a.offload_activation : 0.000048s : 0.01% optimize.opt_a.cell_reuse_handle_not_recompute_node_pass : 0.000090s : 0.03% optimize.opt_a.merge_recompute_call_nodes : 0.000003s : 0.00% optimize.opt_a.before_grad : 0.000076s : 0.02% optimize.opt_a.set_forward_comm_id_for_comm_node_pass : 0.000027s : 0.01% optimize.opt_a.meta_fg_expand : 0.000025s : 0.01% optimize.opt_a.flash_sp_send_recv_attached : 0.000008s : 0.00% optimize.opt_a.receive_attached : 0.000006s : 0.00% optimize.opt_a.after_resolve : 0.000059s : 0.02% optimize.opt_a.a_after_grad : 0.000069s : 0.02% optimize.opt_a.renormalize : 0.011173s : 3.26% optimize.opt_a.add_forward_monad_depend : 0.000023s : 0.01% optimize.opt_a.auto_monad_grad : 0.000005s : 0.00% optimize.opt_a.auto_monad_eliminator : 0.000126s : 0.04% optimize.opt_a.cse : 0.001646s : 0.48% optimize.opt_a.a_3 : 0.000322s : 0.09% optimize.py_interpret_to_execute_after_opt_a : 0.000041s : 0.01% optimize.slice_cell_reuse_recomputed_activation : 0.000002s : 0.00% optimize.rewriter_after_opt_a : 0.000852s : 0.25% optimize.convert_after_rewriter : 0.000022s : 0.01% optimize.order_py_execute_after_rewriter : 0.000015s : 0.00% optimize.mutable_eliminate : 0.000800s : 0.23% optimize.opt_b.b_1 : 0.000533s : 0.16% optimize.opt_b.b_2 : 0.000023s : 0.01% optimize.opt_b.updatestate_depend_eliminate : 0.000017s : 0.00% optimize.opt_b.updatestate_assign_eliminate : 0.000011s : 0.00% optimize.opt_b.updatestate_loads_eliminate : 0.000013s : 0.00% optimize.opt_b.renormalize : 0.000001s : 0.00% optimize.opt_b.cse : 0.000077s : 0.02% optimize.optimize_parallel_all_gather_comm : 0.000037s : 0.01% optimize.overlap_param_gather : 0.000002s : 0.00% optimize.cconv : 0.000037s : 0.01% optimize.loop_unroll : 0.000490s : 0.14% optimize.opt_after_cconv.c_1 : 0.000122s : 0.04% optimize.opt_after_cconv.parameter_eliminate : 0.000003s : 0.00% optimize.opt_after_cconv.updatestate_depend_eliminate : 0.000017s : 0.00% optimize.opt_after_cconv.updatestate_assign_eliminate : 0.000011s : 0.00% optimize.opt_after_cconv.updatestate_loads_eliminate : 0.000012s : 0.00% optimize.opt_after_cconv.cse : 0.000068s : 0.02% optimize.opt_after_cconv.renormalize : 0.000000s : 0.00% optimize.remove_dup_value : 0.000089s : 0.03% optimize.tuple_transform.d_1 : 0.000195s : 0.06% optimize.tuple_transform.none_parameter_eliminate : 0.000002s : 0.00% optimize.tuple_transform.renormalize : 0.000000s : 0.00% optimize.tuple_transform.switch_simplify : 0.000022s : 0.01% optimize.partial_unused_args_eliminate : 0.000002s : 0.00% optimize.add_recomputation : 0.000135s : 0.04% optimize.cse_after_recomputation.cse : 0.000054s : 0.02% optimize.environ_conv : 0.000019s : 0.01% optimize.swap_dp_allreduce_reducescatter : 0.000018s : 0.01% optimize.bias_add_comm_swap : 0.000003s : 0.00% optimize.label_micro_interleaved_index : 0.000005s : 0.00% optimize.label_fine_grained_interleaved_index : 0.000003s : 0.00% optimize.merge_cast_opt : 0.000001s : 0.00% optimize.slice_recompute_activation : 0.000002s : 0.00% optimize.micro_interleaved_order_control : 0.000003s : 0.00% optimize.assign_add_opt : 0.000001s : 0.00% optimize.ForceFp32Comm : 0.000001s : 0.00% optimize.remove_cast_before_assign_add : 0.000001s : 0.00% optimize.full_micro_interleaved_order_control : 0.000002s : 0.00% optimize.reorder_send_recv_between_fp_bp : 0.000003s : 0.00% optimize.comm_op_add_attrs : 0.000001s : 0.00% optimize.add_comm_op_reuse_tag : 0.000001s : 0.00% optimize.interleave_split_concat_branches : 0.000001s : 0.00% optimize.interleave_parallel_branches : 0.000001s : 0.00% optimize.overlap_opt_shard_in_pipeline : 0.000002s : 0.00% optimize.overlap_opt_shard_grad_in_pipeline : 0.000002s : 0.00% optimize.control_data_broadcast_order : 0.000036s : 0.01% optimize.grouped_pairwise_exchange_alltoall : 0.000002s : 0.00% optimize.offloading_packed_experts : 0.000011s : 0.00% optimize.overlap_recompute_and_grad_model_parallel : 0.000010s : 0.00% optimize.overlap_grad_matmul_and_grad_allreduce : 0.000001s : 0.00% optimize.overlap_recompute_allgather_and_fa_grad : 0.000001s : 0.00% optimize.overlap_recompute_comm : 0.000003s : 0.00% optimize.overlap_grad_ring_attention : 0.000009s : 0.00% optimize.overlap_grad_flash_sp : 0.000049s : 0.01% optimize.begin_end_overlap_inline : 0.000001s : 0.00% optimize.split_matmul_comm_elemetwise : 0.000002s : 0.00% optimize.split_layernorm_comm : 0.000002s : 0.00% optimize.handle_group_info : 0.000001s : 0.00% optimize.symbol_engine_optimizer.build : 0.000270s : 0.08% optimize.symbol_engine_optimizer.elim_shapecalc : 0.000030s : 0.01% optimize.symbol_engine_optimizer.elim_not_effective : 0.000047s : 0.01% optimize.symbol_engine_optimizer.opt_reshape : 0.000032s : 0.01% optimize.symbol_engine_optimizer.fold_const_symbol : 0.000054s : 0.02% optimize.symbol_engine_optimizer.renormalize : 0.000000s : 0.00% detach_backward : 0.000003s : 0.00% pipeline_parallel_scheduler : 0.000002s : 0.00% auto_monad_reorder : 0.000060s : 0.02% get_jit_bprop_graph : 0.000002s : 0.00% rewriter_after_jit_bprop_graph : 0.000005s : 0.00% opt_after_jit_grad : 0.000621s : 0.18% validate : 0.000098s : 0.03% Time group info: ------[substitution.] 0.003176 569 0.56% : 0.000018s : 8: substitution.depend_value_elim 0.37% : 0.000012s : 14: substitution.elim_not_effective 0.59% : 0.000019s : 12: substitution.float_tuple_getitem_switch 0.70% : 0.000022s : 14: substitution.fold_const_symbol 0.46% : 0.000015s : 17: substitution.graph_param_transform 68.23% : 0.002167s : 71: substitution.inline 0.42% : 0.000013s : 29: substitution.j_node_and_user_rematch 1.09% : 0.000035s : 6: substitution.less_batch_normalization 1.72% : 0.000055s : 2: substitution.list_to_tuple_eliminator_ 0.20% : 0.000006s : 6: substitution.load_eliminater 0.99% : 0.000031s : 21: substitution.minmaximum_grad 0.24% : 0.000008s : 2: substitution.opt_reshape 0.65% : 0.000021s : 29: substitution.remove_not_recompute_node 0.24% : 0.000008s : 4: substitution.replace_old_param 2.04% : 0.000065s : 18: substitution.reshape_eliminate 1.64% : 0.000052s : 24: substitution.switch_simplify 4.44% : 0.000141s : 51: substitution.tuple_list_convert_item_index_to_positive 2.38% : 0.000075s : 51: substitution.tuple_list_get_item_const_eliminator 3.03% : 0.000096s : 51: substitution.tuple_list_get_item_depend_reorder 5.77% : 0.000183s : 73: substitution.tuple_list_get_item_eliminator 3.08% : 0.000098s : 51: substitution.tuple_list_get_set_item_eliminator 0.59% : 0.000019s : 6: substitution.updatestate_pure_node_eliminater 0.57% : 0.000018s : 9: substitution.updatestate_useless_node_eliminater ------[type_inference.] 0.307373 2 94.44% : 0.290270s : 1: type_inference.infer 5.56% : 0.017103s : 1: type_inference.specialize ------[replace.] 0.001064 115 3.63% : 0.000039s : 5: replace.depend_value_elim 55.38% : 0.000589s : 71: replace.inline 3.79% : 0.000040s : 2: replace.list_to_tuple_eliminator_ 27.67% : 0.000294s : 24: replace.switch_simplify 9.52% : 0.000101s : 13: replace.tuple_list_get_item_eliminator ------[match.] 0.002249 115 0.10% : 0.000002s : 5: match.depend_value_elim 94.53% : 0.002126s : 71: match.inline 2.32% : 0.000052s : 2: match.list_to_tuple_eliminator_ 1.66% : 0.000037s : 24: match.switch_simplify 1.39% : 0.000031s : 13: match.tuple_list_get_item_eliminator ------[predicate.] 0.001787 9701 1.07% : 0.000019s : 123: predicate.accumulaten_eliminater 0.31% : 0.000006s : 17: predicate.ad_related_special_op_eliminate 0.44% : 0.000008s : 58: predicate.addn_check_dump 1.12% : 0.000020s : 123: predicate.addn_zero_filter 0.97% : 0.000017s : 123: predicate.adjust_all_reduce_mul_add 2.01% : 0.000036s : 181: predicate.arithmetic_simplify 1.03% : 0.000018s : 123: predicate.cast_eliminate 0.32% : 0.000006s : 34: predicate.check_bprop_eliminate 0.46% : 0.000008s : 58: predicate.compare_switch_simplify 0.07% : 0.000001s : 17: predicate.const_output_eliminate 0.53% : 0.000010s : 61: predicate.depend_value_elim 1.09% : 0.000020s : 123: predicate.dict_get_item_const_eliminator 1.42% : 0.000025s : 123: predicate.dict_get_item_eliminator 1.00% : 0.000018s : 123: predicate.dict_set_item_eliminator 0.36% : 0.000006s : 34: predicate.dumpgradient_eliminate 0.07% : 0.000001s : 17: predicate.elim_not_effective 0.17% : 0.000003s : 17: predicate.elim_shapecalc_of_broadcastargs 4.25% : 0.000076s : 140: predicate.environ_add_const_eliminate 1.11% : 0.000020s : 140: predicate.environ_get_add_eliminate 1.14% : 0.000020s : 140: predicate.environ_get_depend_swap 1.59% : 0.000028s : 198: predicate.environ_get_eliminate 1.12% : 0.000020s : 140: predicate.environ_get_set_eliminate 1.70% : 0.000030s : 209: predicate.exchange_switch_depend_value 2.51% : 0.000045s : 209: predicate.float_depend_g_call 0.47% : 0.000008s : 58: predicate.float_environ_get_switch 0.63% : 0.000011s : 75: predicate.float_tuple_getitem_switch 0.07% : 0.000001s : 17: predicate.fold_const_symbol 0.32% : 0.000006s : 35: predicate.get_grad_eliminate 0.09% : 0.000002s : 17: predicate.graph_param_transform 0.46% : 0.000008s : 58: predicate.incorporate_call 0.42% : 0.000008s : 58: predicate.incorporate_call_switch 5.14% : 0.000092s : 459: predicate.inline 0.37% : 0.000007s : 35: predicate.inline_without_move 0.15% : 0.000003s : 35: predicate.j_node_and_user_rematch 0.43% : 0.000008s : 36: predicate.less_batch_normalization 1.53% : 0.000027s : 172: predicate.list_to_tuple_eliminator_ 2.47% : 0.000044s : 295: predicate.load_eliminater 0.34% : 0.000006s : 17: predicate.loop_unroll_after_grad 3.15% : 0.000056s : 314: predicate.loop_unroll_before_grad 1.37% : 0.000024s : 157: predicate.make_slice_get_slice_eliminator 0.48% : 0.000009s : 58: predicate.merge_addn 0.28% : 0.000005s : 34: predicate.micro_step_allgather_replace 0.27% : 0.000005s : 34: predicate.mini_step_allgather_replace 1.03% : 0.000018s : 123: predicate.minmaximum_grad 0.37% : 0.000007s : 17: predicate.mutable_eliminate 0.16% : 0.000003s : 17: predicate.opt_reshape 0.16% : 0.000003s : 17: predicate.parallel_virtual_node 2.88% : 0.000052s : 209: predicate.partial_defer_inline 9.79% : 0.000175s : 155: predicate.partial_eliminate 1.03% : 0.000018s : 123: predicate.print_const_string_wrapper 0.41% : 0.000007s : 53: predicate.reduce_all_const_elim 1.41% : 0.000025s : 123: predicate.reduce_eliminate 2.30% : 0.000041s : 295: predicate.redundant_stop_gradient_eliminater 0.14% : 0.000003s : 35: predicate.remove_not_recompute_node 1.02% : 0.000018s : 172: predicate.replace_applicator 0.20% : 0.000004s : 35: predicate.replace_old_param 0.11% : 0.000002s : 17: predicate.reset_defer_inline 1.14% : 0.000020s : 123: predicate.reshape_eliminate 0.30% : 0.000005s : 34: predicate.row_tensor_add_zeros_like 0.16% : 0.000003s : 17: predicate.row_tensor_eliminate 0.41% : 0.000007s : 34: predicate.same_eliminate 0.20% : 0.000004s : 40: predicate.set_cell_output_no_recompute 0.33% : 0.000006s : 35: predicate.shard_identity_eliminate 0.30% : 0.000005s : 34: predicate.special_op_eliminate 0.55% : 0.000010s : 58: predicate.specialize_transform 0.40% : 0.000007s : 34: predicate.split_environ_get_set_with_tuple_value 0.34% : 0.000006s : 35: predicate.stack_unstack_eliminate 0.14% : 0.000003s : 17: predicate.switch_call_monad_eliminater 1.89% : 0.000034s : 209: predicate.switch_defer_inline 2.10% : 0.000038s : 243: predicate.switch_layer_defer_inline 6.07% : 0.000109s : 646: predicate.switch_simplify 1.03% : 0.000018s : 123: predicate.tile_eliminate 1.03% : 0.000018s : 123: predicate.transpose_eliminate 1.56% : 0.000028s : 157: predicate.tuple_list_convert_item_index_to_positive 1.55% : 0.000028s : 157: predicate.tuple_list_get_item_const_eliminator 1.48% : 0.000026s : 157: predicate.tuple_list_get_item_depend_reorder 2.66% : 0.000048s : 228: predicate.tuple_list_get_item_eliminator 1.44% : 0.000026s : 157: predicate.tuple_list_get_set_item_eliminator 1.98% : 0.000035s : 215: predicate.tuple_list_set_item_eliminator 1.40% : 0.000025s : 170: predicate.tuple_to_list_eliminator_ 2.29% : 0.000041s : 295: predicate.updatestate_pure_node_eliminater 2.84% : 0.000051s : 353: predicate.updatestate_useless_node_eliminater 0.16% : 0.000003s : 17: predicate.value_based_eliminate 0.29% : 0.000005s : 35: predicate.virtual_dataset_eliminate 0.31% : 0.000006s : 35: predicate.virtual_output_eliminate 0.14% : 0.000003s : 17: predicate.virtual_view_grad_eliminate 0.17% : 0.000003s : 17: predicate.zero_like_fill_zero ------[func_graph_cloner_run.] 0.013640 135 60.56% : 0.008260s : 55: func_graph_cloner_run.FuncGraphClonerGraph 39.44% : 0.005380s : 80: func_graph_cloner_run.FuncGraphSpecializer ------[meta_graph.] 0.000000 0 ------[manager.] 0.000000 0 ------[pynative] 0.000000 0 ------[others.] 0.413904 209 0.00% : 0.000004s : 1: ForceFp32Comm 1.32% : 0.005472s : 1: add_attr 1.32% : 0.005453s : 1: add_attr_with_inline 0.00% : 0.000005s : 1: add_comm_op_reuse_tag 0.03% : 0.000140s : 1: add_recomputation 0.00% : 0.000005s : 1: assign_add_opt 0.18% : 0.000729s : 1: auto_monad 0.02% : 0.000065s : 1: auto_monad_reorder 0.00% : 0.000005s : 1: begin_end_overlap_inline 0.00% : 0.000006s : 1: bias_add_comm_swap 0.17% : 0.000719s : 1: bootstrap 0.01% : 0.000041s : 1: cconv 0.00% : 0.000004s : 1: comm_op_add_attrs 0.01% : 0.000039s : 1: control_data_broadcast_order 0.01% : 0.000027s : 1: convert_after_rewriter 0.02% : 0.000070s : 1: cse_after_recomputation 0.00% : 0.000005s : 1: dataset_repeat_opt 0.00% : 0.000006s : 1: detach_backward 0.01% : 0.000022s : 1: environ_conv 0.05% : 0.000221s : 1: event_method 0.00% : 0.000006s : 1: full_micro_interleaved_order_control 0.00% : 0.000007s : 1: get_jit_bprop_graph 0.01% : 0.000054s : 1: graph_reusing 0.00% : 0.000005s : 1: grouped_pairwise_exchange_alltoall 0.00% : 0.000005s : 1: handle_group_info 0.00% : 0.000007s : 1: inline 0.00% : 0.000007s : 1: insert-virtual-dataset 0.00% : 0.000004s : 1: interleave_parallel_branches 0.00% : 0.000004s : 1: interleave_split_concat_branches 0.00% : 0.000006s : 1: label_fine_grained_interleaved_index 0.00% : 0.000008s : 1: label_micro_interleaved_index 0.12% : 0.000499s : 1: loop_unroll 0.00% : 0.000005s : 1: merge_cast_opt 0.00% : 0.000006s : 1: micro_interleaved_order_control 0.20% : 0.000811s : 1: mutable_eliminate 0.00% : 0.000014s : 1: offloading_packed_experts 0.01% : 0.000035s : 1: opt.transform.loop_unroll_optimizer 0.01% : 0.000039s : 1: opt.transform.mutable_eliminate 3.17% : 0.013141s : 95: opt.transform.opt_a 0.03% : 0.000121s : 1: opt.transform.opt_after_cconv 0.03% : 0.000121s : 1: opt.transform.opt_after_jit_grad 0.13% : 0.000521s : 28: opt.transform.opt_b 0.05% : 0.000214s : 2: opt.transform.opt_trans_graph 0.04% : 0.000158s : 4: opt.transform.symbol_engine_opt 6.69% : 0.027709s : 1: opt_a 0.07% : 0.000282s : 1: opt_after_cconv 0.15% : 0.000633s : 1: opt_after_jit_grad 0.18% : 0.000729s : 1: opt_b 8.10% : 0.033510s : 1: optimize 0.01% : 0.000041s : 1: optimize_parallel_all_gather_comm 0.00% : 0.000019s : 1: order_py_execute_after_rewriter 0.01% : 0.000053s : 1: overlap_grad_flash_sp 0.00% : 0.000004s : 1: overlap_grad_matmul_and_grad_allreduce 0.00% : 0.000012s : 1: overlap_grad_ring_attention 0.00% : 0.000006s : 1: overlap_opt_shard_grad_in_pipeline 0.00% : 0.000004s : 1: overlap_opt_shard_in_pipeline 0.00% : 0.000007s : 1: overlap_param_gather 0.00% : 0.000004s : 1: overlap_recompute_allgather_and_fa_grad 0.00% : 0.000013s : 1: overlap_recompute_and_grad_model_parallel 0.00% : 0.000007s : 1: overlap_recompute_comm 0.00% : 0.000010s : 1: parallel-infer-symbol 0.00% : 0.000005s : 1: parallel-infer-symbol-second 0.00% : 0.000005s : 1: partial_unused_args_eliminate 0.00% : 0.000005s : 1: pipeline_parallel_scheduler 0.00% : 0.000005s : 1: pipeline_split 0.06% : 0.000256s : 1: pre_auto_parallel 0.05% : 0.000215s : 1: py_interpret_to_execute 0.01% : 0.000046s : 1: py_interpret_to_execute_after_opt_a 0.00% : 0.000004s : 1: remove_cast_before_assign_add 0.02% : 0.000094s : 1: remove_dup_value 1.41% : 0.005851s : 1: renormalize.infer 1.28% : 0.005305s : 1: renormalize.specialize 0.00% : 0.000007s : 1: reorder_send_recv_between_fp_bp 0.00% : 0.000008s : 1: rewriter_after_jit_bprop_graph 0.21% : 0.000860s : 1: rewriter_after_opt_a 0.19% : 0.000798s : 1: rewriter_before_opt_a 0.00% : 0.000006s : 1: slice_cell_reuse_recomputed_activation 0.00% : 0.000005s : 1: slice_recompute_activation 0.00% : 0.000005s : 1: split_layernorm_comm 0.00% : 0.000005s : 1: split_matmul_comm_elemetwise 0.01% : 0.000021s : 1: swap_dp_allreduce_reducescatter 0.12% : 0.000482s : 1: symbol_engine_optimizer 0.06% : 0.000250s : 1: tuple_transform 74.36% : 0.307791s : 1: type_inference . [hook] pytest_runtest_teardown:test_matmul_qkv_bf16_1[True-mstype0-k_n_shape1-256] tests/st/infer/ops/test_internal_ops/test_matmul_split.py::test_matmul_qkv_bf16_1[True-mstype0-k_n_shape1-256],max_mem:146.0M [WARNING] ME(168319:281472866918192,MainProcess):2026-01-29-17:44:42.140.342 [mindspore/context.py:1334] For 'context.set_context', the parameter 'device_target' will be deprecated and removed in a future version. Please use the api mindspore.set_device() instead. TotalTime = 0.788447, [21] [bootstrap]: 0.00063959 [type_inference]: 0.649158 [event_method]: 0.00020614 [auto_monad]: 0.0376363 [graph_reusing]: 5.947e-05 [inline]: 2.90998e-06 [add_attr]: 0.00556554, [1] [add_attr_with_inline]: 0.00554561, [1] [Cycle 1]: 0.00032987, [2] [tag_attr]: 0.00021904 [meta_addattr_fg_expand]: 5.126e-05 [parallel-infer-symbol]: 4.53999e-06 [pre_auto_parallel]: 0.00025491 [insert-virtual-dataset]: 2.84999e-06 [parallel-infer-symbol-second]: 1.62001e-06 [dataset_repeat_opt]: 2.39999e-06 [pipeline_split]: 2.05002e-06 [optimize]: 0.0936913, [53] [py_interpret_to_execute]: 0.00023821 [rewriter_before_opt_a]: 0.0009461 [opt_a]: 0.0871578, [2] [Cycle 1]: 0.0849846, [45] [expand_dump_flag]: 2.299e-05 [switch_simplify]: 0.00082477 [loop_unroll]: 0.00032183 [a_1]: 0.0107222 [with_stream_mark]: 5.746e-05 [recompute_prepare]: 3.632e-05 [updatestate_depend_eliminate]: 1.999e-05 [updatestate_assign_eliminate]: 1.534e-05 [updatestate_loads_eliminate]: 1.883e-05 [parameter_eliminate]: 2.83e-06 [a_2]: 0.00076748 [accelerated_algorithm]: 6.523e-05 [shard]: 2.29001e-06 [meta_shard_fg_expand]: 2.392e-05 [shard_inline]: 2.394e-05 [merge_send_recv]: 2.128e-05 [auto_parallel]: 2.019e-05 [parallel]: 3.82e-05 [flash_sp]: 1.639e-05 [merge_comm]: 1.421e-05 [allreduce_fusion]: 1.265e-05 [matmul_add_comm_reduction]: 2.141e-05 [allreduce_slice_to_reducescatter]: 1.04e-06 [virtual_shard_identity]: 2.784e-05 [virtual_dataset]: 2.399e-05 [get_grad_eliminate_]: 2.095e-05 [virtual_output]: 2.217e-05 [merge_forward]: 1.345e-05 [cell_reuse_recompute_pass]: 2.73e-06 [offload_activation]: 2.344e-05 [cell_reuse_handle_not_recompute_node_pass]: 4.44e-05 [merge_recompute_call_nodes]: 1.99e-06 [before_grad]: 4.004e-05 [set_forward_comm_id_for_comm_node_pass]: 1.37e-05 [meta_fg_expand]: 1.656e-05 [flash_sp_send_recv_attached]: 6.28998e-06 [receive_attached]: 2.65997e-06 [after_resolve]: 2.893e-05 [a_after_grad]: 3.635e-05 [renormalize]: 0.0695838 [add_forward_monad_depend]: 2.021e-05 [auto_monad_grad]: 3.03e-06 [auto_monad_eliminator]: 8.01e-05 [cse]: 0.00123665 [a_3]: 0.00017448 [Cycle 2]: 0.00215353, [45] [expand_dump_flag]: 3.16999e-06 [switch_simplify]: 2.617e-05 [loop_unroll]: 2.205e-05 [a_1]: 0.00065334 [with_stream_mark]: 3.39e-05 [recompute_prepare]: 2.497e-05 [updatestate_depend_eliminate]: 1.63e-05 [updatestate_assign_eliminate]: 1.132e-05 [updatestate_loads_eliminate]: 1.486e-05 [parameter_eliminate]: 2.41e-06 [a_2]: 0.0003088 [accelerated_algorithm]: 3.201e-05 [shard]: 2.27001e-06 [meta_shard_fg_expand]: 8.71002e-06 [shard_inline]: 2.141e-05 [merge_send_recv]: 2.092e-05 [auto_parallel]: 1.963e-05 [parallel]: 9.64e-06 [flash_sp]: 5.76e-06 [merge_comm]: 1.246e-05 [allreduce_fusion]: 1.177e-05 [matmul_add_comm_reduction]: 2.29e-05 [allreduce_slice_to_reducescatter]: 7.60017e-07 [virtual_shard_identity]: 2.418e-05 [virtual_dataset]: 1.972e-05 [get_grad_eliminate_]: 2.076e-05 [virtual_output]: 1.979e-05 [merge_forward]: 1.247e-05 [cell_reuse_recompute_pass]: 3.36001e-06 [offload_activation]: 2.447e-05 [cell_reuse_handle_not_recompute_node_pass]: 3.881e-05 [merge_recompute_call_nodes]: 1.75001e-06 [before_grad]: 3.713e-05 [set_forward_comm_id_for_comm_node_pass]: 1.307e-05 [meta_fg_expand]: 9.34e-06 [flash_sp_send_recv_attached]: 2.61999e-06 [receive_attached]: 3.01999e-06 [after_resolve]: 2.68e-05 [a_after_grad]: 3.261e-05 [renormalize]: 1.09983e-07 [add_forward_monad_depend]: 3.25998e-06 [auto_monad_grad]: 2.63e-06 [auto_monad_eliminator]: 5.777e-05 [cse]: 0.0001248 [a_3]: 0.0001388 [py_interpret_to_execute_after_opt_a]: 3.071e-05 [slice_cell_reuse_recomputed_activation]: 2.48002e-06 [rewriter_after_opt_a]: 0.00096022 [convert_after_rewriter]: 2.405e-05 [order_py_execute_after_rewriter]: 1.528e-05 [mutable_eliminate]: 0.00076096 [opt_b]: 0.00083568, [1] [Cycle 1]: 0.00082549, [7] [b_1]: 0.00060411 [b_2]: 2.487e-05 [updatestate_depend_eliminate]: 1.838e-05 [updatestate_assign_eliminate]: 1.162e-05 [updatestate_loads_eliminate]: 1.496e-05 [renormalize]: 9.70002e-07 [cse]: 9.987e-05 [optimize_parallel_all_gather_comm]: 4.165e-05 [overlap_param_gather]: 6.98e-06 [cconv]: 3.838e-05 [loop_unroll]: 0.00060625 [opt_after_cconv]: 0.00031069, [1] [Cycle 1]: 0.00030068, [7] [c_1]: 0.00012939 [parameter_eliminate]: 5.24e-06 [updatestate_depend_eliminate]: 1.936e-05 [updatestate_assign_eliminate]: 1.128e-05 [updatestate_loads_eliminate]: 1.319e-05 [cse]: 7.978e-05 [renormalize]: 7.39994e-07 [remove_dup_value]: 9.498e-05 [tuple_transform]: 0.00025931, [1] [Cycle 1]: 0.00025318, [4] [d_1]: 0.00020517 [none_parameter_eliminate]: 2.43e-06 [renormalize]: 2.9002e-07 [switch_simplify]: 2.278e-05 [partial_unused_args_eliminate]: 2.34999e-06 [add_recomputation]: 0.00016968 [cse_after_recomputation]: 7.985e-05, [1] [Cycle 1]: 7.234e-05, [1] [cse]: 6.364e-05 [environ_conv]: 2.377e-05 [swap_dp_allreduce_reducescatter]: 1.896e-05 [bias_add_comm_swap]: 3.35e-06 [label_micro_interleaved_index]: 6.34001e-06 [label_fine_grained_interleaved_index]: 2.78e-06 [merge_cast_opt]: 2.12999e-06 [slice_recompute_activation]: 2.39999e-06 [micro_interleaved_order_control]: 3.13e-06 [assign_add_opt]: 1.52999e-06 [ForceFp32Comm]: 9.10019e-07 [remove_cast_before_assign_add]: 1.30999e-06 [full_micro_interleaved_order_control]: 2.36e-06 [reorder_send_recv_between_fp_bp]: 2.80997e-06 [comm_op_add_attrs]: 1.13001e-06 [add_comm_op_reuse_tag]: 1.09003e-06 [interleave_split_concat_branches]: 1.27e-06 [interleave_parallel_branches]: 1.14e-06 [overlap_opt_shard_in_pipeline]: 4.48001e-06 [overlap_opt_shard_grad_in_pipeline]: 1.79998e-06 [control_data_broadcast_order]: 3.971e-05 [grouped_pairwise_exchange_alltoall]: 1.96998e-06 [offloading_packed_experts]: 1.064e-05 [overlap_recompute_and_grad_model_parallel]: 1.046e-05 [overlap_grad_matmul_and_grad_allreduce]: 1.27999e-06 [overlap_recompute_allgather_and_fa_grad]: 1.55001e-06 [overlap_recompute_comm]: 2.37001e-06 [overlap_grad_ring_attention]: 9.51003e-06 [overlap_grad_flash_sp]: 5.012e-05 [begin_end_overlap_inline]: 7.89994e-07 [split_matmul_comm_elemetwise]: 2.09999e-06 [split_layernorm_comm]: 1.87001e-06 [handle_group_info]: 1.07e-06 [symbol_engine_optimizer]: 0.00055913, [1] [Cycle 1]: 0.000552, [6] [build]: 0.00031509 [elim_shapecalc]: 3.719e-05 [elim_not_effective]: 5.381e-05 [opt_reshape]: 3.364e-05 [fold_const_symbol]: 6.115e-05 [renormalize]: 3.30008e-07 [detach_backward]: 3.10002e-06 [pipeline_parallel_scheduler]: 1.66998e-06 [auto_monad_reorder]: 6.464e-05 [get_jit_bprop_graph]: 2.49999e-06 [rewriter_after_jit_bprop_graph]: 5.12e-06 [opt_after_jit_grad]: 0.00075325 [validate]: 0.00010837 Sums bootstrap : 0.000640s : 0.08% type_inference : 0.649158s : 83.06% event_method : 0.000206s : 0.03% auto_monad : 0.037636s : 4.82% graph_reusing : 0.000059s : 0.01% inline : 0.000003s : 0.00% add_attr.add_attr_with_inline.tag_attr : 0.000219s : 0.03% add_attr.add_attr_with_inline.meta_addattr_fg_expand : 0.000051s : 0.01% parallel-infer-symbol : 0.000005s : 0.00% pre_auto_parallel : 0.000255s : 0.03% insert-virtual-dataset : 0.000003s : 0.00% parallel-infer-symbol-second : 0.000002s : 0.00% dataset_repeat_opt : 0.000002s : 0.00% pipeline_split : 0.000002s : 0.00% optimize.py_interpret_to_execute : 0.000238s : 0.03% optimize.rewriter_before_opt_a : 0.000946s : 0.12% optimize.opt_a.expand_dump_flag : 0.000026s : 0.00% optimize.opt_a.switch_simplify : 0.000851s : 0.11% optimize.opt_a.loop_unroll : 0.000344s : 0.04% optimize.opt_a.a_1 : 0.011376s : 1.46% optimize.opt_a.with_stream_mark : 0.000091s : 0.01% optimize.opt_a.recompute_prepare : 0.000061s : 0.01% optimize.opt_a.updatestate_depend_eliminate : 0.000036s : 0.00% optimize.opt_a.updatestate_assign_eliminate : 0.000027s : 0.00% optimize.opt_a.updatestate_loads_eliminate : 0.000034s : 0.00% optimize.opt_a.parameter_eliminate : 0.000005s : 0.00% optimize.opt_a.a_2 : 0.001076s : 0.14% optimize.opt_a.accelerated_algorithm : 0.000097s : 0.01% optimize.opt_a.shard : 0.000005s : 0.00% optimize.opt_a.meta_shard_fg_expand : 0.000033s : 0.00% optimize.opt_a.shard_inline : 0.000045s : 0.01% optimize.opt_a.merge_send_recv : 0.000042s : 0.01% optimize.opt_a.auto_parallel : 0.000040s : 0.01% optimize.opt_a.parallel : 0.000048s : 0.01% optimize.opt_a.flash_sp : 0.000022s : 0.00% optimize.opt_a.merge_comm : 0.000027s : 0.00% optimize.opt_a.allreduce_fusion : 0.000024s : 0.00% optimize.opt_a.matmul_add_comm_reduction : 0.000044s : 0.01% optimize.opt_a.allreduce_slice_to_reducescatter : 0.000002s : 0.00% optimize.opt_a.virtual_shard_identity : 0.000052s : 0.01% optimize.opt_a.virtual_dataset : 0.000044s : 0.01% optimize.opt_a.get_grad_eliminate_ : 0.000042s : 0.01% optimize.opt_a.virtual_output : 0.000042s : 0.01% optimize.opt_a.merge_forward : 0.000026s : 0.00% optimize.opt_a.cell_reuse_recompute_pass : 0.000006s : 0.00% optimize.opt_a.offload_activation : 0.000048s : 0.01% optimize.opt_a.cell_reuse_handle_not_recompute_node_pass : 0.000083s : 0.01% optimize.opt_a.merge_recompute_call_nodes : 0.000004s : 0.00% optimize.opt_a.before_grad : 0.000077s : 0.01% optimize.opt_a.set_forward_comm_id_for_comm_node_pass : 0.000027s : 0.00% optimize.opt_a.meta_fg_expand : 0.000026s : 0.00% optimize.opt_a.flash_sp_send_recv_attached : 0.000009s : 0.00% optimize.opt_a.receive_attached : 0.000006s : 0.00% optimize.opt_a.after_resolve : 0.000056s : 0.01% optimize.opt_a.a_after_grad : 0.000069s : 0.01% optimize.opt_a.renormalize : 0.069584s : 8.90% optimize.opt_a.add_forward_monad_depend : 0.000023s : 0.00% optimize.opt_a.auto_monad_grad : 0.000006s : 0.00% optimize.opt_a.auto_monad_eliminator : 0.000138s : 0.02% optimize.opt_a.cse : 0.001361s : 0.17% optimize.opt_a.a_3 : 0.000313s : 0.04% optimize.py_interpret_to_execute_after_opt_a : 0.000031s : 0.00% optimize.slice_cell_reuse_recomputed_activation : 0.000002s : 0.00% optimize.rewriter_after_opt_a : 0.000960s : 0.12% optimize.convert_after_rewriter : 0.000024s : 0.00% optimize.order_py_execute_after_rewriter : 0.000015s : 0.00% optimize.mutable_eliminate : 0.000761s : 0.10% optimize.opt_b.b_1 : 0.000604s : 0.08% optimize.opt_b.b_2 : 0.000025s : 0.00% optimize.opt_b.updatestate_depend_eliminate : 0.000018s : 0.00% optimize.opt_b.updatestate_assign_eliminate : 0.000012s : 0.00% optimize.opt_b.updatestate_loads_eliminate : 0.000015s : 0.00% optimize.opt_b.renormalize : 0.000001s : 0.00% optimize.opt_b.cse : 0.000100s : 0.01% optimize.optimize_parallel_all_gather_comm : 0.000042s : 0.01% optimize.overlap_param_gather : 0.000007s : 0.00% optimize.cconv : 0.000038s : 0.00% optimize.loop_unroll : 0.000606s : 0.08% optimize.opt_after_cconv.c_1 : 0.000129s : 0.02% optimize.opt_after_cconv.parameter_eliminate : 0.000005s : 0.00% optimize.opt_after_cconv.updatestate_depend_eliminate : 0.000019s : 0.00% optimize.opt_after_cconv.updatestate_assign_eliminate : 0.000011s : 0.00% optimize.opt_after_cconv.updatestate_loads_eliminate : 0.000013s : 0.00% optimize.opt_after_cconv.cse : 0.000080s : 0.01% optimize.opt_after_cconv.renormalize : 0.000001s : 0.00% optimize.remove_dup_value : 0.000095s : 0.01% optimize.tuple_transform.d_1 : 0.000205s : 0.03% optimize.tuple_transform.none_parameter_eliminate : 0.000002s : 0.00% optimize.tuple_transform.renormalize : 0.000000s : 0.00% optimize.tuple_transform.switch_simplify : 0.000023s : 0.00% optimize.partial_unused_args_eliminate : 0.000002s : 0.00% optimize.add_recomputation : 0.000170s : 0.02% optimize.cse_after_recomputation.cse : 0.000064s : 0.01% optimize.environ_conv : 0.000024s : 0.00% optimize.swap_dp_allreduce_reducescatter : 0.000019s : 0.00% optimize.bias_add_comm_swap : 0.000003s : 0.00% optimize.label_micro_interleaved_index : 0.000006s : 0.00% optimize.label_fine_grained_interleaved_index : 0.000003s : 0.00% optimize.merge_cast_opt : 0.000002s : 0.00% optimize.slice_recompute_activation : 0.000002s : 0.00% optimize.micro_interleaved_order_control : 0.000003s : 0.00% optimize.assign_add_opt : 0.000002s : 0.00% optimize.ForceFp32Comm : 0.000001s : 0.00% optimize.remove_cast_before_assign_add : 0.000001s : 0.00% optimize.full_micro_interleaved_order_control : 0.000002s : 0.00% optimize.reorder_send_recv_between_fp_bp : 0.000003s : 0.00% optimize.comm_op_add_attrs : 0.000001s : 0.00% optimize.add_comm_op_reuse_tag : 0.000001s : 0.00% optimize.interleave_split_concat_branches : 0.000001s : 0.00% optimize.interleave_parallel_branches : 0.000001s : 0.00% optimize.overlap_opt_shard_in_pipeline : 0.000004s : 0.00% optimize.overlap_opt_shard_grad_in_pipeline : 0.000002s : 0.00% optimize.control_data_broadcast_order : 0.000040s : 0.01% optimize.grouped_pairwise_exchange_alltoall : 0.000002s : 0.00% optimize.offloading_packed_experts : 0.000011s : 0.00% optimize.overlap_recompute_and_grad_model_parallel : 0.000010s : 0.00% optimize.overlap_grad_matmul_and_grad_allreduce : 0.000001s : 0.00% optimize.overlap_recompute_allgather_and_fa_grad : 0.000002s : 0.00% optimize.overlap_recompute_comm : 0.000002s : 0.00% optimize.overlap_grad_ring_attention : 0.000010s : 0.00% optimize.overlap_grad_flash_sp : 0.000050s : 0.01% optimize.begin_end_overlap_inline : 0.000001s : 0.00% optimize.split_matmul_comm_elemetwise : 0.000002s : 0.00% optimize.split_layernorm_comm : 0.000002s : 0.00% optimize.handle_group_info : 0.000001s : 0.00% optimize.symbol_engine_optimizer.build : 0.000315s : 0.04% optimize.symbol_engine_optimizer.elim_shapecalc : 0.000037s : 0.00% optimize.symbol_engine_optimizer.elim_not_effective : 0.000054s : 0.01% optimize.symbol_engine_optimizer.opt_reshape : 0.000034s : 0.00% optimize.symbol_engine_optimizer.fold_const_symbol : 0.000061s : 0.01% optimize.symbol_engine_optimizer.renormalize : 0.000000s : 0.00% detach_backward : 0.000003s : 0.00% pipeline_parallel_scheduler : 0.000002s : 0.00% auto_monad_reorder : 0.000065s : 0.01% get_jit_bprop_graph : 0.000002s : 0.00% rewriter_after_jit_bprop_graph : 0.000005s : 0.00% opt_after_jit_grad : 0.000753s : 0.10% validate : 0.000108s : 0.01% Time group info: ------[substitution.] 0.003905 569 0.43% : 0.000017s : 8: substitution.depend_value_elim 0.39% : 0.000015s : 14: substitution.elim_not_effective 0.47% : 0.000018s : 12: substitution.float_tuple_getitem_switch 0.74% : 0.000029s : 14: substitution.fold_const_symbol 0.37% : 0.000015s : 17: substitution.graph_param_transform 70.48% : 0.002752s : 71: substitution.inline 0.33% : 0.000013s : 29: substitution.j_node_and_user_rematch 0.87% : 0.000034s : 6: substitution.less_batch_normalization 1.48% : 0.000058s : 2: substitution.list_to_tuple_eliminator_ 0.17% : 0.000007s : 6: substitution.load_eliminater 0.79% : 0.000031s : 21: substitution.minmaximum_grad 0.22% : 0.000008s : 2: substitution.opt_reshape 0.48% : 0.000019s : 29: substitution.remove_not_recompute_node 0.18% : 0.000007s : 4: substitution.replace_old_param 1.73% : 0.000068s : 18: substitution.reshape_eliminate 1.32% : 0.000051s : 24: substitution.switch_simplify 4.33% : 0.000169s : 51: substitution.tuple_list_convert_item_index_to_positive 2.23% : 0.000087s : 51: substitution.tuple_list_get_item_const_eliminator 2.98% : 0.000116s : 51: substitution.tuple_list_get_item_depend_reorder 6.02% : 0.000235s : 73: substitution.tuple_list_get_item_eliminator 2.91% : 0.000114s : 51: substitution.tuple_list_get_set_item_eliminator 0.55% : 0.000021s : 6: substitution.updatestate_pure_node_eliminater 0.55% : 0.000021s : 9: substitution.updatestate_useless_node_eliminater ------[type_inference.] 0.648814 2 95.30% : 0.618321s : 1: type_inference.infer 4.70% : 0.030493s : 1: type_inference.specialize ------[replace.] 0.001348 115 3.01% : 0.000041s : 5: replace.depend_value_elim 61.85% : 0.000834s : 71: replace.inline 3.41% : 0.000046s : 2: replace.list_to_tuple_eliminator_ 22.52% : 0.000303s : 24: replace.switch_simplify 9.22% : 0.000124s : 13: replace.tuple_list_get_item_eliminator ------[match.] 0.002840 115 0.07% : 0.000002s : 5: match.depend_value_elim 95.15% : 0.002702s : 71: match.inline 1.94% : 0.000055s : 2: match.list_to_tuple_eliminator_ 1.32% : 0.000038s : 24: match.switch_simplify 1.51% : 0.000043s : 13: match.tuple_list_get_item_eliminator ------[predicate.] 0.001606 9701 1.16% : 0.000019s : 123: predicate.accumulaten_eliminater 0.35% : 0.000006s : 17: predicate.ad_related_special_op_eliminate 0.50% : 0.000008s : 58: predicate.addn_check_dump 1.16% : 0.000019s : 123: predicate.addn_zero_filter 1.08% : 0.000017s : 123: predicate.adjust_all_reduce_mul_add 2.31% : 0.000037s : 181: predicate.arithmetic_simplify 1.18% : 0.000019s : 123: predicate.cast_eliminate 0.31% : 0.000005s : 34: predicate.check_bprop_eliminate 0.48% : 0.000008s : 58: predicate.compare_switch_simplify 0.08% : 0.000001s : 17: predicate.const_output_eliminate 0.61% : 0.000010s : 61: predicate.depend_value_elim 1.30% : 0.000021s : 123: predicate.dict_get_item_const_eliminator 1.31% : 0.000021s : 123: predicate.dict_get_item_eliminator 1.12% : 0.000018s : 123: predicate.dict_set_item_eliminator 0.39% : 0.000006s : 34: predicate.dumpgradient_eliminate 0.10% : 0.000002s : 17: predicate.elim_not_effective 0.20% : 0.000003s : 17: predicate.elim_shapecalc_of_broadcastargs 1.21% : 0.000019s : 140: predicate.environ_add_const_eliminate 1.27% : 0.000020s : 140: predicate.environ_get_add_eliminate 1.27% : 0.000020s : 140: predicate.environ_get_depend_swap 1.80% : 0.000029s : 198: predicate.environ_get_eliminate 1.25% : 0.000020s : 140: predicate.environ_get_set_eliminate 1.93% : 0.000031s : 209: predicate.exchange_switch_depend_value 2.79% : 0.000045s : 209: predicate.float_depend_g_call 0.49% : 0.000008s : 58: predicate.float_environ_get_switch 0.67% : 0.000011s : 75: predicate.float_tuple_getitem_switch 0.08% : 0.000001s : 17: predicate.fold_const_symbol 0.32% : 0.000005s : 35: predicate.get_grad_eliminate 0.09% : 0.000001s : 17: predicate.graph_param_transform 0.52% : 0.000008s : 58: predicate.incorporate_call 0.46% : 0.000007s : 58: predicate.incorporate_call_switch 6.07% : 0.000098s : 459: predicate.inline 0.44% : 0.000007s : 35: predicate.inline_without_move 0.16% : 0.000003s : 35: predicate.j_node_and_user_rematch 0.48% : 0.000008s : 36: predicate.less_batch_normalization 1.70% : 0.000027s : 172: predicate.list_to_tuple_eliminator_ 2.76% : 0.000044s : 295: predicate.load_eliminater 0.38% : 0.000006s : 17: predicate.loop_unroll_after_grad 3.30% : 0.000053s : 314: predicate.loop_unroll_before_grad 1.53% : 0.000025s : 157: predicate.make_slice_get_slice_eliminator 0.50% : 0.000008s : 58: predicate.merge_addn 0.28% : 0.000005s : 34: predicate.micro_step_allgather_replace 0.30% : 0.000005s : 34: predicate.mini_step_allgather_replace 1.19% : 0.000019s : 123: predicate.minmaximum_grad 0.36% : 0.000006s : 17: predicate.mutable_eliminate 0.18% : 0.000003s : 17: predicate.opt_reshape 0.16% : 0.000003s : 17: predicate.parallel_virtual_node 3.63% : 0.000058s : 209: predicate.partial_defer_inline 1.53% : 0.000025s : 155: predicate.partial_eliminate 1.23% : 0.000020s : 123: predicate.print_const_string_wrapper 0.47% : 0.000007s : 53: predicate.reduce_all_const_elim 1.77% : 0.000028s : 123: predicate.reduce_eliminate 2.66% : 0.000043s : 295: predicate.redundant_stop_gradient_eliminater 0.15% : 0.000002s : 35: predicate.remove_not_recompute_node 1.06% : 0.000017s : 172: predicate.replace_applicator 0.19% : 0.000003s : 35: predicate.replace_old_param 0.09% : 0.000002s : 17: predicate.reset_defer_inline 1.34% : 0.000022s : 123: predicate.reshape_eliminate 0.30% : 0.000005s : 34: predicate.row_tensor_add_zeros_like 0.19% : 0.000003s : 17: predicate.row_tensor_eliminate 0.44% : 0.000007s : 34: predicate.same_eliminate 0.23% : 0.000004s : 40: predicate.set_cell_output_no_recompute 0.37% : 0.000006s : 35: predicate.shard_identity_eliminate 0.36% : 0.000006s : 34: predicate.special_op_eliminate 0.58% : 0.000009s : 58: predicate.specialize_transform 0.41% : 0.000007s : 34: predicate.split_environ_get_set_with_tuple_value 0.38% : 0.000006s : 35: predicate.stack_unstack_eliminate 0.18% : 0.000003s : 17: predicate.switch_call_monad_eliminater 2.11% : 0.000034s : 209: predicate.switch_defer_inline 2.38% : 0.000038s : 243: predicate.switch_layer_defer_inline 6.98% : 0.000112s : 646: predicate.switch_simplify 1.24% : 0.000020s : 123: predicate.tile_eliminate 1.28% : 0.000021s : 123: predicate.transpose_eliminate 1.71% : 0.000027s : 157: predicate.tuple_list_convert_item_index_to_positive 1.82% : 0.000029s : 157: predicate.tuple_list_get_item_const_eliminator 1.59% : 0.000025s : 157: predicate.tuple_list_get_item_depend_reorder 3.39% : 0.000054s : 228: predicate.tuple_list_get_item_eliminator 1.63% : 0.000026s : 157: predicate.tuple_list_get_set_item_eliminator 2.31% : 0.000037s : 215: predicate.tuple_list_set_item_eliminator 1.57% : 0.000025s : 170: predicate.tuple_to_list_eliminator_ 2.52% : 0.000040s : 295: predicate.updatestate_pure_node_eliminater 3.11% : 0.000050s : 353: predicate.updatestate_useless_node_eliminater 0.16% : 0.000003s : 17: predicate.value_based_eliminate 0.36% : 0.000006s : 35: predicate.virtual_dataset_eliminate 0.34% : 0.000005s : 35: predicate.virtual_output_eliminate 0.14% : 0.000002s : 17: predicate.virtual_view_grad_eliminate 0.19% : 0.000003s : 17: predicate.zero_like_fill_zero ------[func_graph_cloner_run.] 0.016758 135 57.78% : 0.009682s : 55: func_graph_cloner_run.FuncGraphClonerGraph 42.22% : 0.007076s : 80: func_graph_cloner_run.FuncGraphSpecializer ------[meta_graph.] 0.000000 0 ------[manager.] 0.000000 0 ------[pynative] 0.000000 0 ------[others.] 0.972780 209 0.00% : 0.000004s : 1: ForceFp32Comm 0.57% : 0.005573s : 1: add_attr 0.57% : 0.005552s : 1: add_attr_with_inline 0.00% : 0.000005s : 1: add_comm_op_reuse_tag 0.02% : 0.000176s : 1: add_recomputation 0.00% : 0.000005s : 1: assign_add_opt 3.87% : 0.037682s : 1: auto_monad 0.01% : 0.000071s : 1: auto_monad_reorder 0.00% : 0.000005s : 1: begin_end_overlap_inline 0.00% : 0.000006s : 1: bias_add_comm_swap 0.07% : 0.000675s : 1: bootstrap 0.00% : 0.000042s : 1: cconv 0.00% : 0.000004s : 1: comm_op_add_attrs 0.00% : 0.000043s : 1: control_data_broadcast_order 0.00% : 0.000031s : 1: convert_after_rewriter 0.01% : 0.000083s : 1: cse_after_recomputation 0.00% : 0.000005s : 1: dataset_repeat_opt 0.00% : 0.000006s : 1: detach_backward 0.00% : 0.000027s : 1: environ_conv 0.02% : 0.000219s : 1: event_method 0.00% : 0.000007s : 1: full_micro_interleaved_order_control 0.00% : 0.000007s : 1: get_jit_bprop_graph 0.01% : 0.000069s : 1: graph_reusing 0.00% : 0.000005s : 1: grouped_pairwise_exchange_alltoall 0.00% : 0.000005s : 1: handle_group_info 0.00% : 0.000007s : 1: inline 0.00% : 0.000007s : 1: insert-virtual-dataset 0.00% : 0.000004s : 1: interleave_parallel_branches 0.00% : 0.000004s : 1: interleave_split_concat_branches 0.00% : 0.000006s : 1: label_fine_grained_interleaved_index 0.00% : 0.000010s : 1: label_micro_interleaved_index 0.06% : 0.000619s : 1: loop_unroll 0.00% : 0.000006s : 1: merge_cast_opt 0.00% : 0.000006s : 1: micro_interleaved_order_control 0.08% : 0.000772s : 1: mutable_eliminate 0.00% : 0.000014s : 1: offloading_packed_experts 0.00% : 0.000040s : 1: opt.transform.loop_unroll_optimizer 0.00% : 0.000043s : 1: opt.transform.mutable_eliminate 1.49% : 0.014484s : 95: opt.transform.opt_a 0.01% : 0.000128s : 1: opt.transform.opt_after_cconv 0.01% : 0.000082s : 1: opt.transform.opt_after_jit_grad 0.06% : 0.000590s : 28: opt.transform.opt_b 0.02% : 0.000225s : 2: opt.transform.opt_trans_graph 0.02% : 0.000179s : 4: opt.transform.symbol_engine_opt 8.96% : 0.087163s : 1: opt_a 0.03% : 0.000314s : 1: opt_after_cconv 0.08% : 0.000768s : 1: opt_after_jit_grad 0.09% : 0.000841s : 1: opt_b 9.63% : 0.093698s : 1: optimize 0.00% : 0.000046s : 1: optimize_parallel_all_gather_comm 0.00% : 0.000020s : 1: order_py_execute_after_rewriter 0.01% : 0.000054s : 1: overlap_grad_flash_sp 0.00% : 0.000004s : 1: overlap_grad_matmul_and_grad_allreduce 0.00% : 0.000013s : 1: overlap_grad_ring_attention 0.00% : 0.000006s : 1: overlap_opt_shard_grad_in_pipeline 0.00% : 0.000007s : 1: overlap_opt_shard_in_pipeline 0.00% : 0.000012s : 1: overlap_param_gather 0.00% : 0.000004s : 1: overlap_recompute_allgather_and_fa_grad 0.00% : 0.000013s : 1: overlap_recompute_and_grad_model_parallel 0.00% : 0.000007s : 1: overlap_recompute_comm 0.00% : 0.000010s : 1: parallel-infer-symbol 0.00% : 0.000005s : 1: parallel-infer-symbol-second 0.00% : 0.000006s : 1: partial_unused_args_eliminate 0.00% : 0.000005s : 1: pipeline_parallel_scheduler 0.00% : 0.000005s : 1: pipeline_split 0.03% : 0.000266s : 1: pre_auto_parallel 0.03% : 0.000250s : 1: py_interpret_to_execute 0.00% : 0.000036s : 1: py_interpret_to_execute_after_opt_a 0.00% : 0.000004s : 1: remove_cast_before_assign_add 0.01% : 0.000103s : 1: remove_dup_value 6.45% : 0.062718s : 1: renormalize.infer 0.70% : 0.006845s : 1: renormalize.specialize 0.00% : 0.000007s : 1: reorder_send_recv_between_fp_bp 0.00% : 0.000008s : 1: rewriter_after_jit_bprop_graph 0.10% : 0.000974s : 1: rewriter_after_opt_a 0.10% : 0.000969s : 1: rewriter_before_opt_a 0.00% : 0.000005s : 1: slice_cell_reuse_recomputed_activation 0.00% : 0.000005s : 1: slice_recompute_activation 0.00% : 0.000004s : 1: split_layernorm_comm 0.00% : 0.000005s : 1: split_matmul_comm_elemetwise 0.00% : 0.000023s : 1: swap_dp_allreduce_reducescatter 0.06% : 0.000563s : 1: symbol_engine_optimizer 0.03% : 0.000263s : 1: tuple_transform 66.74% : 0.649198s : 1: type_inference . [hook] pytest_runtest_teardown:test_matmul_qkv_bf16_1[True-mstype0-k_n_shape1-1024] tests/st/infer/ops/test_internal_ops/test_matmul_split.py::test_matmul_qkv_bf16_1[True-mstype0-k_n_shape1-1024],max_mem:170.0M [WARNING] ME(168319:281472866918192,MainProcess):2026-01-29-17:44:54.689.178 [mindspore/context.py:1334] For 'context.set_context', the parameter 'device_target' will be deprecated and removed in a future version. Please use the api mindspore.set_device() instead. TotalTime = 0.591928, [21] [bootstrap]: 0.0208269 [type_inference]: 0.527469 [event_method]: 0.00021145 [auto_monad]: 0.00080184 [graph_reusing]: 5.078e-05 [inline]: 3.76999e-06 [add_attr]: 0.00527665, [1] [add_attr_with_inline]: 0.00525757, [1] [Cycle 1]: 0.0003053, [2] [tag_attr]: 0.00020793 [meta_addattr_fg_expand]: 5.867e-05 [parallel-infer-symbol]: 3.92998e-06 [pre_auto_parallel]: 0.0002484 [insert-virtual-dataset]: 3.25e-06 [parallel-infer-symbol-second]: 1.17999e-06 [dataset_repeat_opt]: 2.24001e-06 [pipeline_split]: 1.94999e-06 [optimize]: 0.0358859, [53] [py_interpret_to_execute]: 0.00023116 [rewriter_before_opt_a]: 0.00077992 [opt_a]: 0.0298994, [2] [Cycle 1]: 0.0277353, [45] [expand_dump_flag]: 1.793e-05 [switch_simplify]: 0.00083251 [loop_unroll]: 0.00034571 [a_1]: 0.00910404 [with_stream_mark]: 5.74e-05 [recompute_prepare]: 3.812e-05 [updatestate_depend_eliminate]: 1.983e-05 [updatestate_assign_eliminate]: 1.992e-05 [updatestate_loads_eliminate]: 1.975e-05 [parameter_eliminate]: 3.00002e-06 [a_2]: 0.00077374 [accelerated_algorithm]: 6.122e-05 [shard]: 2.44999e-06 [meta_shard_fg_expand]: 2.239e-05 [shard_inline]: 2.278e-05 [merge_send_recv]: 2.145e-05 [auto_parallel]: 2.093e-05 [parallel]: 3.387e-05 [flash_sp]: 1.527e-05 [merge_comm]: 1.418e-05 [allreduce_fusion]: 1.31e-05 [matmul_add_comm_reduction]: 2.224e-05 [allreduce_slice_to_reducescatter]: 9.39996e-07 [virtual_shard_identity]: 2.425e-05 [virtual_dataset]: 2.243e-05 [get_grad_eliminate_]: 2.096e-05 [virtual_output]: 2.107e-05 [merge_forward]: 1.292e-05 [cell_reuse_recompute_pass]: 2.15002e-06 [offload_activation]: 2.291e-05 [cell_reuse_handle_not_recompute_node_pass]: 4.595e-05 [merge_recompute_call_nodes]: 1.52999e-06 [before_grad]: 3.954e-05 [set_forward_comm_id_for_comm_node_pass]: 1.359e-05 [meta_fg_expand]: 1.602e-05 [flash_sp_send_recv_attached]: 5.49998e-06 [receive_attached]: 2.96001e-06 [after_resolve]: 2.906e-05 [a_after_grad]: 3.581e-05 [renormalize]: 0.0140861 [add_forward_monad_depend]: 2.113e-05 [auto_monad_grad]: 3.29001e-06 [auto_monad_eliminator]: 7.922e-05 [cse]: 0.00117931 [a_3]: 0.00018085 [Cycle 2]: 0.00214896, [45] [expand_dump_flag]: 3.04999e-06 [switch_simplify]: 2.801e-05 [loop_unroll]: 2.218e-05 [a_1]: 0.00066135 [with_stream_mark]: 2.907e-05 [recompute_prepare]: 2.133e-05 [updatestate_depend_eliminate]: 1.553e-05 [updatestate_assign_eliminate]: 1.167e-05 [updatestate_loads_eliminate]: 4.15e-05 [parameter_eliminate]: 3.03e-06 [a_2]: 0.00032364 [accelerated_algorithm]: 3.012e-05 [shard]: 2.39001e-06 [meta_shard_fg_expand]: 7.11999e-06 [shard_inline]: 2.328e-05 [merge_send_recv]: 2.067e-05 [auto_parallel]: 2.114e-05 [parallel]: 1.081e-05 [flash_sp]: 4.45999e-06 [merge_comm]: 1.337e-05 [allreduce_fusion]: 1.237e-05 [matmul_add_comm_reduction]: 2.341e-05 [allreduce_slice_to_reducescatter]: 9.00007e-07 [virtual_shard_identity]: 2.346e-05 [virtual_dataset]: 2.092e-05 [get_grad_eliminate_]: 2.161e-05 [virtual_output]: 1.996e-05 [merge_forward]: 1.246e-05 [cell_reuse_recompute_pass]: 3.65e-06 [offload_activation]: 2.732e-05 [cell_reuse_handle_not_recompute_node_pass]: 3.88e-05 [merge_recompute_call_nodes]: 1.47001e-06 [before_grad]: 3.82e-05 [set_forward_comm_id_for_comm_node_pass]: 1.377e-05 [meta_fg_expand]: 9.73002e-06 [flash_sp_send_recv_attached]: 2.94001e-06 [receive_attached]: 3.55998e-06 [after_resolve]: 2.971e-05 [a_after_grad]: 3.353e-05 [renormalize]: 9.00181e-08 [add_forward_monad_depend]: 2.46e-06 [auto_monad_grad]: 2.82002e-06 [auto_monad_eliminator]: 5.408e-05 [cse]: 7.913e-05 [a_3]: 0.0001408 [py_interpret_to_execute_after_opt_a]: 3.348e-05 [slice_cell_reuse_recomputed_activation]: 2.16e-06 [rewriter_after_opt_a]: 0.00085369 [convert_after_rewriter]: 2.301e-05 [order_py_execute_after_rewriter]: 1.516e-05 [mutable_eliminate]: 0.00079374 [opt_b]: 0.00072716, [1] [Cycle 1]: 0.00071809, [7] [b_1]: 0.0005341 [b_2]: 2.278e-05 [updatestate_depend_eliminate]: 1.747e-05 [updatestate_assign_eliminate]: 1.215e-05 [updatestate_loads_eliminate]: 1.352e-05 [renormalize]: 6.10016e-07 [cse]: 7.53e-05 [optimize_parallel_all_gather_comm]: 4.288e-05 [overlap_param_gather]: 2.00002e-06 [cconv]: 3.833e-05 [loop_unroll]: 0.00055675 [opt_after_cconv]: 0.00028435, [1] [Cycle 1]: 0.00027541, [7] [c_1]: 0.00012189 [parameter_eliminate]: 3.64002e-06 [updatestate_depend_eliminate]: 1.782e-05 [updatestate_assign_eliminate]: 1.068e-05 [updatestate_loads_eliminate]: 1.348e-05 [cse]: 6.972e-05 [renormalize]: 7.50006e-07 [remove_dup_value]: 9.369e-05 [tuple_transform]: 0.00025013, [1] [Cycle 1]: 0.00024449, [4] [d_1]: 0.00019761 [none_parameter_eliminate]: 2.65997e-06 [renormalize]: 2.00002e-07 [switch_simplify]: 2.295e-05 [partial_unused_args_eliminate]: 2.32999e-06 [add_recomputation]: 0.00016119 [cse_after_recomputation]: 7.554e-05, [1] [Cycle 1]: 6.863e-05, [1] [cse]: 6.177e-05 [environ_conv]: 2.378e-05 [swap_dp_allreduce_reducescatter]: 1.916e-05 [bias_add_comm_swap]: 3.09001e-06 [label_micro_interleaved_index]: 6.46e-06 [label_fine_grained_interleaved_index]: 2.63e-06 [merge_cast_opt]: 1.62999e-06 [slice_recompute_activation]: 2.56e-06 [micro_interleaved_order_control]: 2.89999e-06 [assign_add_opt]: 1.42999e-06 [ForceFp32Comm]: 1.01002e-06 [remove_cast_before_assign_add]: 1.22999e-06 [full_micro_interleaved_order_control]: 2.93998e-06 [reorder_send_recv_between_fp_bp]: 3.06001e-06 [comm_op_add_attrs]: 1.19e-06 [add_comm_op_reuse_tag]: 1.30999e-06 [interleave_split_concat_branches]: 1.40999e-06 [interleave_parallel_branches]: 1.14e-06 [overlap_opt_shard_in_pipeline]: 1.30999e-06 [overlap_opt_shard_grad_in_pipeline]: 1.78002e-06 [control_data_broadcast_order]: 3.811e-05 [grouped_pairwise_exchange_alltoall]: 1.81e-06 [offloading_packed_experts]: 1.217e-05 [overlap_recompute_and_grad_model_parallel]: 1.133e-05 [overlap_grad_matmul_and_grad_allreduce]: 1.52999e-06 [overlap_recompute_allgather_and_fa_grad]: 1.74e-06 [overlap_recompute_comm]: 2.41e-06 [overlap_grad_ring_attention]: 9.02e-06 [overlap_grad_flash_sp]: 5.201e-05 [begin_end_overlap_inline]: 7.00005e-07 [split_matmul_comm_elemetwise]: 2.73003e-06 [split_layernorm_comm]: 1.84e-06 [handle_group_info]: 1.02e-06 [symbol_engine_optimizer]: 0.00050984, [1] [Cycle 1]: 0.00050426, [6] [build]: 0.00029127 [elim_shapecalc]: 3.017e-05 [elim_not_effective]: 5.008e-05 [opt_reshape]: 3.269e-05 [fold_const_symbol]: 5.728e-05 [renormalize]: 3.60014e-07 [detach_backward]: 2.48e-06 [pipeline_parallel_scheduler]: 1.64998e-06 [auto_monad_reorder]: 6.008e-05 [get_jit_bprop_graph]: 1.99999e-06 [rewriter_after_jit_bprop_graph]: 6.16e-06 [opt_after_jit_grad]: 0.00069269 [validate]: 0.00011191 Sums bootstrap : 0.020827s : 3.56% type_inference : 0.527469s : 90.08% event_method : 0.000211s : 0.04% auto_monad : 0.000802s : 0.14% graph_reusing : 0.000051s : 0.01% inline : 0.000004s : 0.00% add_attr.add_attr_with_inline.tag_attr : 0.000208s : 0.04% add_attr.add_attr_with_inline.meta_addattr_fg_expand : 0.000059s : 0.01% parallel-infer-symbol : 0.000004s : 0.00% pre_auto_parallel : 0.000248s : 0.04% insert-virtual-dataset : 0.000003s : 0.00% parallel-infer-symbol-second : 0.000001s : 0.00% dataset_repeat_opt : 0.000002s : 0.00% pipeline_split : 0.000002s : 0.00% optimize.py_interpret_to_execute : 0.000231s : 0.04% optimize.rewriter_before_opt_a : 0.000780s : 0.13% optimize.opt_a.expand_dump_flag : 0.000021s : 0.00% optimize.opt_a.switch_simplify : 0.000861s : 0.15% optimize.opt_a.loop_unroll : 0.000368s : 0.06% optimize.opt_a.a_1 : 0.009765s : 1.67% optimize.opt_a.with_stream_mark : 0.000086s : 0.01% optimize.opt_a.recompute_prepare : 0.000059s : 0.01% optimize.opt_a.updatestate_depend_eliminate : 0.000035s : 0.01% optimize.opt_a.updatestate_assign_eliminate : 0.000032s : 0.01% optimize.opt_a.updatestate_loads_eliminate : 0.000061s : 0.01% optimize.opt_a.parameter_eliminate : 0.000006s : 0.00% optimize.opt_a.a_2 : 0.001097s : 0.19% optimize.opt_a.accelerated_algorithm : 0.000091s : 0.02% optimize.opt_a.shard : 0.000005s : 0.00% optimize.opt_a.meta_shard_fg_expand : 0.000030s : 0.01% optimize.opt_a.shard_inline : 0.000046s : 0.01% optimize.opt_a.merge_send_recv : 0.000042s : 0.01% optimize.opt_a.auto_parallel : 0.000042s : 0.01% optimize.opt_a.parallel : 0.000045s : 0.01% optimize.opt_a.flash_sp : 0.000020s : 0.00% optimize.opt_a.merge_comm : 0.000028s : 0.00% optimize.opt_a.allreduce_fusion : 0.000025s : 0.00% optimize.opt_a.matmul_add_comm_reduction : 0.000046s : 0.01% optimize.opt_a.allreduce_slice_to_reducescatter : 0.000002s : 0.00% optimize.opt_a.virtual_shard_identity : 0.000048s : 0.01% optimize.opt_a.virtual_dataset : 0.000043s : 0.01% optimize.opt_a.get_grad_eliminate_ : 0.000043s : 0.01% optimize.opt_a.virtual_output : 0.000041s : 0.01% optimize.opt_a.merge_forward : 0.000025s : 0.00% optimize.opt_a.cell_reuse_recompute_pass : 0.000006s : 0.00% optimize.opt_a.offload_activation : 0.000050s : 0.01% optimize.opt_a.cell_reuse_handle_not_recompute_node_pass : 0.000085s : 0.01% optimize.opt_a.merge_recompute_call_nodes : 0.000003s : 0.00% optimize.opt_a.before_grad : 0.000078s : 0.01% optimize.opt_a.set_forward_comm_id_for_comm_node_pass : 0.000027s : 0.00% optimize.opt_a.meta_fg_expand : 0.000026s : 0.00% optimize.opt_a.flash_sp_send_recv_attached : 0.000008s : 0.00% optimize.opt_a.receive_attached : 0.000007s : 0.00% optimize.opt_a.after_resolve : 0.000059s : 0.01% optimize.opt_a.a_after_grad : 0.000069s : 0.01% optimize.opt_a.renormalize : 0.014086s : 2.41% optimize.opt_a.add_forward_monad_depend : 0.000024s : 0.00% optimize.opt_a.auto_monad_grad : 0.000006s : 0.00% optimize.opt_a.auto_monad_eliminator : 0.000133s : 0.02% optimize.opt_a.cse : 0.001258s : 0.21% optimize.opt_a.a_3 : 0.000322s : 0.05% optimize.py_interpret_to_execute_after_opt_a : 0.000033s : 0.01% optimize.slice_cell_reuse_recomputed_activation : 0.000002s : 0.00% optimize.rewriter_after_opt_a : 0.000854s : 0.15% optimize.convert_after_rewriter : 0.000023s : 0.00% optimize.order_py_execute_after_rewriter : 0.000015s : 0.00% optimize.mutable_eliminate : 0.000794s : 0.14% optimize.opt_b.b_1 : 0.000534s : 0.09% optimize.opt_b.b_2 : 0.000023s : 0.00% optimize.opt_b.updatestate_depend_eliminate : 0.000017s : 0.00% optimize.opt_b.updatestate_assign_eliminate : 0.000012s : 0.00% optimize.opt_b.updatestate_loads_eliminate : 0.000014s : 0.00% optimize.opt_b.renormalize : 0.000001s : 0.00% optimize.opt_b.cse : 0.000075s : 0.01% optimize.optimize_parallel_all_gather_comm : 0.000043s : 0.01% optimize.overlap_param_gather : 0.000002s : 0.00% optimize.cconv : 0.000038s : 0.01% optimize.loop_unroll : 0.000557s : 0.10% optimize.opt_after_cconv.c_1 : 0.000122s : 0.02% optimize.opt_after_cconv.parameter_eliminate : 0.000004s : 0.00% optimize.opt_after_cconv.updatestate_depend_eliminate : 0.000018s : 0.00% optimize.opt_after_cconv.updatestate_assign_eliminate : 0.000011s : 0.00% optimize.opt_after_cconv.updatestate_loads_eliminate : 0.000013s : 0.00% optimize.opt_after_cconv.cse : 0.000070s : 0.01% optimize.opt_after_cconv.renormalize : 0.000001s : 0.00% optimize.remove_dup_value : 0.000094s : 0.02% optimize.tuple_transform.d_1 : 0.000198s : 0.03% optimize.tuple_transform.none_parameter_eliminate : 0.000003s : 0.00% optimize.tuple_transform.renormalize : 0.000000s : 0.00% optimize.tuple_transform.switch_simplify : 0.000023s : 0.00% optimize.partial_unused_args_eliminate : 0.000002s : 0.00% optimize.add_recomputation : 0.000161s : 0.03% optimize.cse_after_recomputation.cse : 0.000062s : 0.01% optimize.environ_conv : 0.000024s : 0.00% optimize.swap_dp_allreduce_reducescatter : 0.000019s : 0.00% optimize.bias_add_comm_swap : 0.000003s : 0.00% optimize.label_micro_interleaved_index : 0.000006s : 0.00% optimize.label_fine_grained_interleaved_index : 0.000003s : 0.00% optimize.merge_cast_opt : 0.000002s : 0.00% optimize.slice_recompute_activation : 0.000003s : 0.00% optimize.micro_interleaved_order_control : 0.000003s : 0.00% optimize.assign_add_opt : 0.000001s : 0.00% optimize.ForceFp32Comm : 0.000001s : 0.00% optimize.remove_cast_before_assign_add : 0.000001s : 0.00% optimize.full_micro_interleaved_order_control : 0.000003s : 0.00% optimize.reorder_send_recv_between_fp_bp : 0.000003s : 0.00% optimize.comm_op_add_attrs : 0.000001s : 0.00% optimize.add_comm_op_reuse_tag : 0.000001s : 0.00% optimize.interleave_split_concat_branches : 0.000001s : 0.00% optimize.interleave_parallel_branches : 0.000001s : 0.00% optimize.overlap_opt_shard_in_pipeline : 0.000001s : 0.00% optimize.overlap_opt_shard_grad_in_pipeline : 0.000002s : 0.00% optimize.control_data_broadcast_order : 0.000038s : 0.01% optimize.grouped_pairwise_exchange_alltoall : 0.000002s : 0.00% optimize.offloading_packed_experts : 0.000012s : 0.00% optimize.overlap_recompute_and_grad_model_parallel : 0.000011s : 0.00% optimize.overlap_grad_matmul_and_grad_allreduce : 0.000002s : 0.00% optimize.overlap_recompute_allgather_and_fa_grad : 0.000002s : 0.00% optimize.overlap_recompute_comm : 0.000002s : 0.00% optimize.overlap_grad_ring_attention : 0.000009s : 0.00% optimize.overlap_grad_flash_sp : 0.000052s : 0.01% optimize.begin_end_overlap_inline : 0.000001s : 0.00% optimize.split_matmul_comm_elemetwise : 0.000003s : 0.00% optimize.split_layernorm_comm : 0.000002s : 0.00% optimize.handle_group_info : 0.000001s : 0.00% optimize.symbol_engine_optimizer.build : 0.000291s : 0.05% optimize.symbol_engine_optimizer.elim_shapecalc : 0.000030s : 0.01% optimize.symbol_engine_optimizer.elim_not_effective : 0.000050s : 0.01% optimize.symbol_engine_optimizer.opt_reshape : 0.000033s : 0.01% optimize.symbol_engine_optimizer.fold_const_symbol : 0.000057s : 0.01% optimize.symbol_engine_optimizer.renormalize : 0.000000s : 0.00% detach_backward : 0.000002s : 0.00% pipeline_parallel_scheduler : 0.000002s : 0.00% auto_monad_reorder : 0.000060s : 0.01% get_jit_bprop_graph : 0.000002s : 0.00% rewriter_after_jit_bprop_graph : 0.000006s : 0.00% opt_after_jit_grad : 0.000693s : 0.12% validate : 0.000112s : 0.02% Time group info: ------[substitution.] 0.003358 569 0.53% : 0.000018s : 8: substitution.depend_value_elim 0.35% : 0.000012s : 14: substitution.elim_not_effective 0.52% : 0.000018s : 12: substitution.float_tuple_getitem_switch 0.77% : 0.000026s : 14: substitution.fold_const_symbol 0.45% : 0.000015s : 17: substitution.graph_param_transform 68.61% : 0.002304s : 71: substitution.inline 0.41% : 0.000014s : 29: substitution.j_node_and_user_rematch 0.96% : 0.000032s : 6: substitution.less_batch_normalization 1.52% : 0.000051s : 2: substitution.list_to_tuple_eliminator_ 0.18% : 0.000006s : 6: substitution.load_eliminater 1.07% : 0.000036s : 21: substitution.minmaximum_grad 0.23% : 0.000008s : 2: substitution.opt_reshape 0.59% : 0.000020s : 29: substitution.remove_not_recompute_node 0.24% : 0.000008s : 4: substitution.replace_old_param 1.81% : 0.000061s : 18: substitution.reshape_eliminate 1.47% : 0.000049s : 24: substitution.switch_simplify 4.52% : 0.000152s : 51: substitution.tuple_list_convert_item_index_to_positive 2.41% : 0.000081s : 51: substitution.tuple_list_get_item_const_eliminator 3.11% : 0.000105s : 51: substitution.tuple_list_get_item_depend_reorder 5.82% : 0.000196s : 73: substitution.tuple_list_get_item_eliminator 3.17% : 0.000106s : 51: substitution.tuple_list_get_set_item_eliminator 0.66% : 0.000022s : 6: substitution.updatestate_pure_node_eliminater 0.59% : 0.000020s : 9: substitution.updatestate_useless_node_eliminater ------[type_inference.] 0.527107 2 96.58% : 0.509071s : 1: type_inference.infer 3.42% : 0.018036s : 1: type_inference.specialize ------[replace.] 0.001125 115 3.74% : 0.000042s : 5: replace.depend_value_elim 54.72% : 0.000616s : 71: replace.inline 3.61% : 0.000041s : 2: replace.list_to_tuple_eliminator_ 28.46% : 0.000320s : 24: replace.switch_simplify 9.46% : 0.000106s : 13: replace.tuple_list_get_item_eliminator ------[match.] 0.002380 115 0.09% : 0.000002s : 5: match.depend_value_elim 94.99% : 0.002261s : 71: match.inline 2.06% : 0.000049s : 2: match.list_to_tuple_eliminator_ 1.46% : 0.000035s : 24: match.switch_simplify 1.40% : 0.000033s : 13: match.tuple_list_get_item_eliminator ------[predicate.] 0.001587 9701 1.17% : 0.000018s : 123: predicate.accumulaten_eliminater 0.34% : 0.000005s : 17: predicate.ad_related_special_op_eliminate 0.49% : 0.000008s : 58: predicate.addn_check_dump 1.20% : 0.000019s : 123: predicate.addn_zero_filter 1.09% : 0.000017s : 123: predicate.adjust_all_reduce_mul_add 2.31% : 0.000037s : 181: predicate.arithmetic_simplify 1.15% : 0.000018s : 123: predicate.cast_eliminate 0.32% : 0.000005s : 34: predicate.check_bprop_eliminate 0.51% : 0.000008s : 58: predicate.compare_switch_simplify 0.08% : 0.000001s : 17: predicate.const_output_eliminate 0.59% : 0.000009s : 61: predicate.depend_value_elim 1.23% : 0.000019s : 123: predicate.dict_get_item_const_eliminator 1.39% : 0.000022s : 123: predicate.dict_get_item_eliminator 1.17% : 0.000019s : 123: predicate.dict_set_item_eliminator 0.40% : 0.000006s : 34: predicate.dumpgradient_eliminate 0.09% : 0.000001s : 17: predicate.elim_not_effective 0.20% : 0.000003s : 17: predicate.elim_shapecalc_of_broadcastargs 1.28% : 0.000020s : 140: predicate.environ_add_const_eliminate 1.22% : 0.000019s : 140: predicate.environ_get_add_eliminate 1.25% : 0.000020s : 140: predicate.environ_get_depend_swap 1.78% : 0.000028s : 198: predicate.environ_get_eliminate 1.28% : 0.000020s : 140: predicate.environ_get_set_eliminate 1.92% : 0.000030s : 209: predicate.exchange_switch_depend_value 2.67% : 0.000042s : 209: predicate.float_depend_g_call 0.51% : 0.000008s : 58: predicate.float_environ_get_switch 0.70% : 0.000011s : 75: predicate.float_tuple_getitem_switch 0.08% : 0.000001s : 17: predicate.fold_const_symbol 0.34% : 0.000005s : 35: predicate.get_grad_eliminate 0.10% : 0.000002s : 17: predicate.graph_param_transform 0.53% : 0.000008s : 58: predicate.incorporate_call 0.48% : 0.000008s : 58: predicate.incorporate_call_switch 5.81% : 0.000092s : 459: predicate.inline 0.46% : 0.000007s : 35: predicate.inline_without_move 0.16% : 0.000002s : 35: predicate.j_node_and_user_rematch 0.50% : 0.000008s : 36: predicate.less_batch_normalization 1.65% : 0.000026s : 172: predicate.list_to_tuple_eliminator_ 2.86% : 0.000045s : 295: predicate.load_eliminater 0.40% : 0.000006s : 17: predicate.loop_unroll_after_grad 3.57% : 0.000057s : 314: predicate.loop_unroll_before_grad 1.51% : 0.000024s : 157: predicate.make_slice_get_slice_eliminator 0.52% : 0.000008s : 58: predicate.merge_addn 0.33% : 0.000005s : 34: predicate.micro_step_allgather_replace 0.33% : 0.000005s : 34: predicate.mini_step_allgather_replace 1.10% : 0.000018s : 123: predicate.minmaximum_grad 0.38% : 0.000006s : 17: predicate.mutable_eliminate 0.18% : 0.000003s : 17: predicate.opt_reshape 0.18% : 0.000003s : 17: predicate.parallel_virtual_node 3.27% : 0.000052s : 209: predicate.partial_defer_inline 1.53% : 0.000024s : 155: predicate.partial_eliminate 1.19% : 0.000019s : 123: predicate.print_const_string_wrapper 0.48% : 0.000008s : 53: predicate.reduce_all_const_elim 1.73% : 0.000027s : 123: predicate.reduce_eliminate 2.67% : 0.000042s : 295: predicate.redundant_stop_gradient_eliminater 0.17% : 0.000003s : 35: predicate.remove_not_recompute_node 1.11% : 0.000018s : 172: predicate.replace_applicator 0.20% : 0.000003s : 35: predicate.replace_old_param 0.11% : 0.000002s : 17: predicate.reset_defer_inline 1.25% : 0.000020s : 123: predicate.reshape_eliminate 0.34% : 0.000005s : 34: predicate.row_tensor_add_zeros_like 0.19% : 0.000003s : 17: predicate.row_tensor_eliminate 0.50% : 0.000008s : 34: predicate.same_eliminate 0.22% : 0.000004s : 40: predicate.set_cell_output_no_recompute 0.40% : 0.000006s : 35: predicate.shard_identity_eliminate 0.35% : 0.000006s : 34: predicate.special_op_eliminate 0.62% : 0.000010s : 58: predicate.specialize_transform 0.45% : 0.000007s : 34: predicate.split_environ_get_set_with_tuple_value 0.37% : 0.000006s : 35: predicate.stack_unstack_eliminate 0.17% : 0.000003s : 17: predicate.switch_call_monad_eliminater 2.12% : 0.000034s : 209: predicate.switch_defer_inline 2.36% : 0.000038s : 243: predicate.switch_layer_defer_inline 6.98% : 0.000111s : 646: predicate.switch_simplify 1.31% : 0.000021s : 123: predicate.tile_eliminate 1.16% : 0.000018s : 123: predicate.transpose_eliminate 1.72% : 0.000027s : 157: predicate.tuple_list_convert_item_index_to_positive 1.74% : 0.000028s : 157: predicate.tuple_list_get_item_const_eliminator 1.61% : 0.000026s : 157: predicate.tuple_list_get_item_depend_reorder 3.15% : 0.000050s : 228: predicate.tuple_list_get_item_eliminator 1.77% : 0.000028s : 157: predicate.tuple_list_get_set_item_eliminator 2.37% : 0.000038s : 215: predicate.tuple_list_set_item_eliminator 1.58% : 0.000025s : 170: predicate.tuple_to_list_eliminator_ 2.58% : 0.000041s : 295: predicate.updatestate_pure_node_eliminater 3.20% : 0.000051s : 353: predicate.updatestate_useless_node_eliminater 0.20% : 0.000003s : 17: predicate.value_based_eliminate 0.37% : 0.000006s : 35: predicate.virtual_dataset_eliminate 0.35% : 0.000006s : 35: predicate.virtual_output_eliminate 0.14% : 0.000002s : 17: predicate.virtual_view_grad_eliminate 0.19% : 0.000003s : 17: predicate.zero_like_fill_zero ------[func_graph_cloner_run.] 0.015762 135 56.90% : 0.008968s : 55: func_graph_cloner_run.FuncGraphClonerGraph 43.10% : 0.006794s : 80: func_graph_cloner_run.FuncGraphSpecializer ------[meta_graph.] 0.000000 0 ------[manager.] 0.000000 0 ------[pynative] 0.000000 0 ------[others.] 0.661013 209 0.00% : 0.000004s : 1: ForceFp32Comm 0.80% : 0.005284s : 1: add_attr 0.80% : 0.005264s : 1: add_attr_with_inline 0.00% : 0.000005s : 1: add_comm_op_reuse_tag 0.03% : 0.000167s : 1: add_recomputation 0.00% : 0.000005s : 1: assign_add_opt 0.12% : 0.000819s : 1: auto_monad 0.01% : 0.000065s : 1: auto_monad_reorder 0.00% : 0.000004s : 1: begin_end_overlap_inline 0.00% : 0.000006s : 1: bias_add_comm_swap 3.16% : 0.020891s : 1: bootstrap 0.01% : 0.000042s : 1: cconv 0.00% : 0.000004s : 1: comm_op_add_attrs 0.01% : 0.000042s : 1: control_data_broadcast_order 0.00% : 0.000028s : 1: convert_after_rewriter 0.01% : 0.000078s : 1: cse_after_recomputation 0.00% : 0.000006s : 1: dataset_repeat_opt 0.00% : 0.000006s : 1: detach_backward 0.00% : 0.000027s : 1: environ_conv 0.03% : 0.000224s : 1: event_method 0.00% : 0.000007s : 1: full_micro_interleaved_order_control 0.00% : 0.000007s : 1: get_jit_bprop_graph 0.01% : 0.000059s : 1: graph_reusing 0.00% : 0.000005s : 1: grouped_pairwise_exchange_alltoall 0.00% : 0.000005s : 1: handle_group_info 0.00% : 0.000007s : 1: inline 0.00% : 0.000007s : 1: insert-virtual-dataset 0.00% : 0.000004s : 1: interleave_parallel_branches 0.00% : 0.000004s : 1: interleave_split_concat_branches 0.00% : 0.000006s : 1: label_fine_grained_interleaved_index 0.00% : 0.000009s : 1: label_micro_interleaved_index 0.09% : 0.000566s : 1: loop_unroll 0.00% : 0.000006s : 1: merge_cast_opt 0.00% : 0.000006s : 1: micro_interleaved_order_control 0.12% : 0.000804s : 1: mutable_eliminate 0.00% : 0.000015s : 1: offloading_packed_experts 0.01% : 0.000035s : 1: opt.transform.loop_unroll_optimizer 0.01% : 0.000038s : 1: opt.transform.mutable_eliminate 1.96% : 0.012933s : 95: opt.transform.opt_a 0.02% : 0.000120s : 1: opt.transform.opt_after_cconv 0.01% : 0.000075s : 1: opt.transform.opt_after_jit_grad 0.08% : 0.000521s : 28: opt.transform.opt_b 0.03% : 0.000217s : 2: opt.transform.opt_trans_graph 0.02% : 0.000165s : 4: opt.transform.symbol_engine_opt 4.52% : 0.029904s : 1: opt_a 0.04% : 0.000288s : 1: opt_after_cconv 0.11% : 0.000705s : 1: opt_after_jit_grad 0.11% : 0.000731s : 1: opt_b 5.43% : 0.035893s : 1: optimize 0.01% : 0.000048s : 1: optimize_parallel_all_gather_comm 0.00% : 0.000019s : 1: order_py_execute_after_rewriter 0.01% : 0.000056s : 1: overlap_grad_flash_sp 0.00% : 0.000004s : 1: overlap_grad_matmul_and_grad_allreduce 0.00% : 0.000012s : 1: overlap_grad_ring_attention 0.00% : 0.000006s : 1: overlap_opt_shard_grad_in_pipeline 0.00% : 0.000004s : 1: overlap_opt_shard_in_pipeline 0.00% : 0.000007s : 1: overlap_param_gather 0.00% : 0.000004s : 1: overlap_recompute_allgather_and_fa_grad 0.00% : 0.000014s : 1: overlap_recompute_and_grad_model_parallel 0.00% : 0.000006s : 1: overlap_recompute_comm 0.00% : 0.000009s : 1: parallel-infer-symbol 0.00% : 0.000004s : 1: parallel-infer-symbol-second 0.00% : 0.000006s : 1: partial_unused_args_eliminate 0.00% : 0.000005s : 1: pipeline_parallel_scheduler 0.00% : 0.000005s : 1: pipeline_split 0.04% : 0.000256s : 1: pre_auto_parallel 0.04% : 0.000241s : 1: py_interpret_to_execute 0.01% : 0.000037s : 1: py_interpret_to_execute_after_opt_a 0.00% : 0.000004s : 1: remove_cast_before_assign_add 0.01% : 0.000099s : 1: remove_dup_value 1.12% : 0.007413s : 1: renormalize.infer 1.01% : 0.006652s : 1: renormalize.specialize 0.00% : 0.000007s : 1: reorder_send_recv_between_fp_bp 0.00% : 0.000009s : 1: rewriter_after_jit_bprop_graph 0.13% : 0.000863s : 1: rewriter_after_opt_a 0.12% : 0.000792s : 1: rewriter_before_opt_a 0.00% : 0.000005s : 1: slice_cell_reuse_recomputed_activation 0.00% : 0.000005s : 1: slice_recompute_activation 0.00% : 0.000005s : 1: split_layernorm_comm 0.00% : 0.000006s : 1: split_matmul_comm_elemetwise 0.00% : 0.000022s : 1: swap_dp_allreduce_reducescatter 0.08% : 0.000513s : 1: symbol_engine_optimizer 0.04% : 0.000255s : 1: tuple_transform 79.80% : 0.527506s : 1: type_inference . [hook] pytest_runtest_teardown:test_matmul_qkv_bf16_1[True-mstype0-k_n_shape2-256] tests/st/infer/ops/test_internal_ops/test_matmul_split.py::test_matmul_qkv_bf16_1[True-mstype0-k_n_shape2-256],max_mem:212.0M [WARNING] ME(168319:281472866918192,MainProcess):2026-01-29-17:45:09.196.626 [mindspore/context.py:1334] For 'context.set_context', the parameter 'device_target' will be deprecated and removed in a future version. Please use the api mindspore.set_device() instead. TotalTime = 0.33226, [21] [bootstrap]: 0.00062734 [type_inference]: 0.286365 [event_method]: 0.00021624 [auto_monad]: 0.00076984 [graph_reusing]: 5.069e-05 [inline]: 2.99999e-06 [add_attr]: 0.00535267, [1] [add_attr_with_inline]: 0.00533395, [1] [Cycle 1]: 0.00031146, [2] [tag_attr]: 0.00020547 [meta_addattr_fg_expand]: 6.048e-05 [parallel-infer-symbol]: 3.87998e-06 [pre_auto_parallel]: 0.00024555 [insert-virtual-dataset]: 2.58e-06 [parallel-infer-symbol-second]: 1.16002e-06 [dataset_repeat_opt]: 2.51e-06 [pipeline_split]: 1.86e-06 [optimize]: 0.0367173, [53] [py_interpret_to_execute]: 0.00025162 [rewriter_before_opt_a]: 0.00079572 [opt_a]: 0.0300548, [2] [Cycle 1]: 0.0278463, [45] [expand_dump_flag]: 2.673e-05 [switch_simplify]: 0.00085022 [loop_unroll]: 0.00034777 [a_1]: 0.0096265 [with_stream_mark]: 4.68e-05 [recompute_prepare]: 3.747e-05 [updatestate_depend_eliminate]: 2.15e-05 [updatestate_assign_eliminate]: 1.628e-05 [updatestate_loads_eliminate]: 2.083e-05 [parameter_eliminate]: 2.47001e-06 [a_2]: 0.00078385 [accelerated_algorithm]: 6.547e-05 [shard]: 2.71e-06 [meta_shard_fg_expand]: 2.446e-05 [shard_inline]: 2.312e-05 [merge_send_recv]: 2.094e-05 [auto_parallel]: 2.041e-05 [parallel]: 3.842e-05 [flash_sp]: 1.532e-05 [merge_comm]: 1.353e-05 [allreduce_fusion]: 1.272e-05 [matmul_add_comm_reduction]: 2.211e-05 [allreduce_slice_to_reducescatter]: 7.2e-07 [virtual_shard_identity]: 2.552e-05 [virtual_dataset]: 2.38e-05 [get_grad_eliminate_]: 2.123e-05 [virtual_output]: 2.131e-05 [merge_forward]: 1.304e-05 [cell_reuse_recompute_pass]: 1.96e-06 [offload_activation]: 2.368e-05 [cell_reuse_handle_not_recompute_node_pass]: 4.52e-05 [merge_recompute_call_nodes]: 1.91e-06 [before_grad]: 3.867e-05 [set_forward_comm_id_for_comm_node_pass]: 1.432e-05 [meta_fg_expand]: 2.065e-05 [flash_sp_send_recv_attached]: 5.54e-06 [receive_attached]: 2.79001e-06 [after_resolve]: 2.968e-05 [a_after_grad]: 0.00014348 [renormalize]: 0.0123081 [add_forward_monad_depend]: 2.031e-05 [auto_monad_grad]: 3.03e-06 [auto_monad_eliminator]: 8.137e-05 [cse]: 0.00236057 [a_3]: 0.00017785 [Cycle 2]: 0.00218922, [45] [expand_dump_flag]: 3.66999e-06 [switch_simplify]: 2.812e-05 [loop_unroll]: 2.168e-05 [a_1]: 0.00065912 [with_stream_mark]: 6.398e-05 [recompute_prepare]: 2.632e-05 [updatestate_depend_eliminate]: 1.736e-05 [updatestate_assign_eliminate]: 1.233e-05 [updatestate_loads_eliminate]: 1.511e-05 [parameter_eliminate]: 3.16999e-06 [a_2]: 0.00032477 [accelerated_algorithm]: 3.272e-05 [shard]: 2.72001e-06 [meta_shard_fg_expand]: 9.13002e-06 [shard_inline]: 2.13e-05 [merge_send_recv]: 2.036e-05 [auto_parallel]: 2.021e-05 [parallel]: 1.027e-05 [flash_sp]: 4.38999e-06 [merge_comm]: 1.361e-05 [allreduce_fusion]: 1.258e-05 [matmul_add_comm_reduction]: 2.308e-05 [allreduce_slice_to_reducescatter]: 7.90023e-07 [virtual_shard_identity]: 2.496e-05 [virtual_dataset]: 2.098e-05 [get_grad_eliminate_]: 2.152e-05 [virtual_output]: 1.947e-05 [merge_forward]: 1.285e-05 [cell_reuse_recompute_pass]: 3.20002e-06 [offload_activation]: 2.43e-05 [cell_reuse_handle_not_recompute_node_pass]: 4.21e-05 [merge_recompute_call_nodes]: 1.81e-06 [before_grad]: 3.691e-05 [set_forward_comm_id_for_comm_node_pass]: 1.413e-05 [meta_fg_expand]: 1.05e-05 [flash_sp_send_recv_attached]: 2.19001e-06 [receive_attached]: 2.64999e-06 [after_resolve]: 2.88e-05 [a_after_grad]: 3.304e-05 [renormalize]: 8.00064e-08 [add_forward_monad_depend]: 3.14001e-06 [auto_monad_grad]: 2.41e-06 [auto_monad_eliminator]: 5.776e-05 [cse]: 8.771e-05 [a_3]: 0.00013994 [py_interpret_to_execute_after_opt_a]: 3.263e-05 [slice_cell_reuse_recomputed_activation]: 2.63e-06 [rewriter_after_opt_a]: 0.00089255 [convert_after_rewriter]: 2.549e-05 [order_py_execute_after_rewriter]: 1.587e-05 [mutable_eliminate]: 0.00082914 [opt_b]: 0.00089858, [1] [Cycle 1]: 0.00088824, [7] [b_1]: 0.00057823 [b_2]: 2.907e-05 [updatestate_depend_eliminate]: 2.103e-05 [updatestate_assign_eliminate]: 1.177e-05 [updatestate_loads_eliminate]: 1.405e-05 [renormalize]: 6.80011e-07 [cse]: 0.00012872 [optimize_parallel_all_gather_comm]: 4.578e-05 [overlap_param_gather]: 2.34999e-06 [cconv]: 4.225e-05 [loop_unroll]: 0.00072937 [opt_after_cconv]: 0.00034166, [1] [Cycle 1]: 0.00033115, [7] [c_1]: 0.00014656 [parameter_eliminate]: 6.51e-06 [updatestate_depend_eliminate]: 2.192e-05 [updatestate_assign_eliminate]: 1.147e-05 [updatestate_loads_eliminate]: 1.409e-05 [cse]: 8.719e-05 [renormalize]: 7.40023e-07 [remove_dup_value]: 0.00010628 [tuple_transform]: 0.00028407, [1] [Cycle 1]: 0.00027688, [4] [d_1]: 0.00022664 [none_parameter_eliminate]: 2.26998e-06 [renormalize]: 2.69996e-07 [switch_simplify]: 2.517e-05 [partial_unused_args_eliminate]: 2.37999e-06 [add_recomputation]: 0.00016913 [cse_after_recomputation]: 7.774e-05, [1] [Cycle 1]: 7.053e-05, [1] [cse]: 6.249e-05 [environ_conv]: 2.459e-05 [swap_dp_allreduce_reducescatter]: 2.067e-05 [bias_add_comm_swap]: 3.56001e-06 [label_micro_interleaved_index]: 7.46001e-06 [label_fine_grained_interleaved_index]: 2.96999e-06 [merge_cast_opt]: 1.57999e-06 [slice_recompute_activation]: 2.48e-06 [micro_interleaved_order_control]: 3.22002e-06 [assign_add_opt]: 1.60999e-06 [ForceFp32Comm]: 1.07e-06 [remove_cast_before_assign_add]: 1.22e-06 [full_micro_interleaved_order_control]: 2.27001e-06 [reorder_send_recv_between_fp_bp]: 2.79999e-06 [comm_op_add_attrs]: 1.10001e-06 [add_comm_op_reuse_tag]: 1.19e-06 [interleave_split_concat_branches]: 1.49e-06 [interleave_parallel_branches]: 1.49998e-06 [overlap_opt_shard_in_pipeline]: 1.74e-06 [overlap_opt_shard_grad_in_pipeline]: 2.17001e-06 [control_data_broadcast_order]: 4.072e-05 [grouped_pairwise_exchange_alltoall]: 1.92001e-06 [offloading_packed_experts]: 1.261e-05 [overlap_recompute_and_grad_model_parallel]: 1.198e-05 [overlap_grad_matmul_and_grad_allreduce]: 1.47999e-06 [overlap_recompute_allgather_and_fa_grad]: 1.55999e-06 [overlap_recompute_comm]: 2.92002e-06 [overlap_grad_ring_attention]: 1.069e-05 [overlap_grad_flash_sp]: 5.159e-05 [begin_end_overlap_inline]: 5.50004e-07 [split_matmul_comm_elemetwise]: 2.83e-06 [split_layernorm_comm]: 2.31e-06 [handle_group_info]: 1.42999e-06 [symbol_engine_optimizer]: 0.00055855, [1] [Cycle 1]: 0.0005517, [6] [build]: 0.00030498 [elim_shapecalc]: 3.55e-05 [elim_not_effective]: 5.949e-05 [opt_reshape]: 3.897e-05 [fold_const_symbol]: 6.665e-05 [renormalize]: 4.89992e-07 [detach_backward]: 2.51998e-06 [pipeline_parallel_scheduler]: 1.79e-06 [auto_monad_reorder]: 6.501e-05 [get_jit_bprop_graph]: 1.74998e-06 [rewriter_after_jit_bprop_graph]: 5.34e-06 [opt_after_jit_grad]: 0.00078445 [validate]: 0.00081595 Sums bootstrap : 0.000627s : 0.19% type_inference : 0.286365s : 87.93% event_method : 0.000216s : 0.07% auto_monad : 0.000770s : 0.24% graph_reusing : 0.000051s : 0.02% inline : 0.000003s : 0.00% add_attr.add_attr_with_inline.tag_attr : 0.000205s : 0.06% add_attr.add_attr_with_inline.meta_addattr_fg_expand : 0.000060s : 0.02% parallel-infer-symbol : 0.000004s : 0.00% pre_auto_parallel : 0.000246s : 0.08% insert-virtual-dataset : 0.000003s : 0.00% parallel-infer-symbol-second : 0.000001s : 0.00% dataset_repeat_opt : 0.000003s : 0.00% pipeline_split : 0.000002s : 0.00% optimize.py_interpret_to_execute : 0.000252s : 0.08% optimize.rewriter_before_opt_a : 0.000796s : 0.24% optimize.opt_a.expand_dump_flag : 0.000030s : 0.01% optimize.opt_a.switch_simplify : 0.000878s : 0.27% optimize.opt_a.loop_unroll : 0.000369s : 0.11% optimize.opt_a.a_1 : 0.010286s : 3.16% optimize.opt_a.with_stream_mark : 0.000111s : 0.03% optimize.opt_a.recompute_prepare : 0.000064s : 0.02% optimize.opt_a.updatestate_depend_eliminate : 0.000039s : 0.01% optimize.opt_a.updatestate_assign_eliminate : 0.000029s : 0.01% optimize.opt_a.updatestate_loads_eliminate : 0.000036s : 0.01% optimize.opt_a.parameter_eliminate : 0.000006s : 0.00% optimize.opt_a.a_2 : 0.001109s : 0.34% optimize.opt_a.accelerated_algorithm : 0.000098s : 0.03% optimize.opt_a.shard : 0.000005s : 0.00% optimize.opt_a.meta_shard_fg_expand : 0.000034s : 0.01% optimize.opt_a.shard_inline : 0.000044s : 0.01% optimize.opt_a.merge_send_recv : 0.000041s : 0.01% optimize.opt_a.auto_parallel : 0.000041s : 0.01% optimize.opt_a.parallel : 0.000049s : 0.01% optimize.opt_a.flash_sp : 0.000020s : 0.01% optimize.opt_a.merge_comm : 0.000027s : 0.01% optimize.opt_a.allreduce_fusion : 0.000025s : 0.01% optimize.opt_a.matmul_add_comm_reduction : 0.000045s : 0.01% optimize.opt_a.allreduce_slice_to_reducescatter : 0.000002s : 0.00% optimize.opt_a.virtual_shard_identity : 0.000050s : 0.02% optimize.opt_a.virtual_dataset : 0.000045s : 0.01% optimize.opt_a.get_grad_eliminate_ : 0.000043s : 0.01% optimize.opt_a.virtual_output : 0.000041s : 0.01% optimize.opt_a.merge_forward : 0.000026s : 0.01% optimize.opt_a.cell_reuse_recompute_pass : 0.000005s : 0.00% optimize.opt_a.offload_activation : 0.000048s : 0.01% optimize.opt_a.cell_reuse_handle_not_recompute_node_pass : 0.000087s : 0.03% optimize.opt_a.merge_recompute_call_nodes : 0.000004s : 0.00% optimize.opt_a.before_grad : 0.000076s : 0.02% optimize.opt_a.set_forward_comm_id_for_comm_node_pass : 0.000028s : 0.01% optimize.opt_a.meta_fg_expand : 0.000031s : 0.01% optimize.opt_a.flash_sp_send_recv_attached : 0.000008s : 0.00% optimize.opt_a.receive_attached : 0.000005s : 0.00% optimize.opt_a.after_resolve : 0.000058s : 0.02% optimize.opt_a.a_after_grad : 0.000177s : 0.05% optimize.opt_a.renormalize : 0.012308s : 3.78% optimize.opt_a.add_forward_monad_depend : 0.000023s : 0.01% optimize.opt_a.auto_monad_grad : 0.000005s : 0.00% optimize.opt_a.auto_monad_eliminator : 0.000139s : 0.04% optimize.opt_a.cse : 0.002448s : 0.75% optimize.opt_a.a_3 : 0.000318s : 0.10% optimize.py_interpret_to_execute_after_opt_a : 0.000033s : 0.01% optimize.slice_cell_reuse_recomputed_activation : 0.000003s : 0.00% optimize.rewriter_after_opt_a : 0.000893s : 0.27% optimize.convert_after_rewriter : 0.000025s : 0.01% optimize.order_py_execute_after_rewriter : 0.000016s : 0.00% optimize.mutable_eliminate : 0.000829s : 0.25% optimize.opt_b.b_1 : 0.000578s : 0.18% optimize.opt_b.b_2 : 0.000029s : 0.01% optimize.opt_b.updatestate_depend_eliminate : 0.000021s : 0.01% optimize.opt_b.updatestate_assign_eliminate : 0.000012s : 0.00% optimize.opt_b.updatestate_loads_eliminate : 0.000014s : 0.00% optimize.opt_b.renormalize : 0.000001s : 0.00% optimize.opt_b.cse : 0.000129s : 0.04% optimize.optimize_parallel_all_gather_comm : 0.000046s : 0.01% optimize.overlap_param_gather : 0.000002s : 0.00% optimize.cconv : 0.000042s : 0.01% optimize.loop_unroll : 0.000729s : 0.22% optimize.opt_after_cconv.c_1 : 0.000147s : 0.05% optimize.opt_after_cconv.parameter_eliminate : 0.000007s : 0.00% optimize.opt_after_cconv.updatestate_depend_eliminate : 0.000022s : 0.01% optimize.opt_after_cconv.updatestate_assign_eliminate : 0.000011s : 0.00% optimize.opt_after_cconv.updatestate_loads_eliminate : 0.000014s : 0.00% optimize.opt_after_cconv.cse : 0.000087s : 0.03% optimize.opt_after_cconv.renormalize : 0.000001s : 0.00% optimize.remove_dup_value : 0.000106s : 0.03% optimize.tuple_transform.d_1 : 0.000227s : 0.07% optimize.tuple_transform.none_parameter_eliminate : 0.000002s : 0.00% optimize.tuple_transform.renormalize : 0.000000s : 0.00% optimize.tuple_transform.switch_simplify : 0.000025s : 0.01% optimize.partial_unused_args_eliminate : 0.000002s : 0.00% optimize.add_recomputation : 0.000169s : 0.05% optimize.cse_after_recomputation.cse : 0.000062s : 0.02% optimize.environ_conv : 0.000025s : 0.01% optimize.swap_dp_allreduce_reducescatter : 0.000021s : 0.01% optimize.bias_add_comm_swap : 0.000004s : 0.00% optimize.label_micro_interleaved_index : 0.000007s : 0.00% optimize.label_fine_grained_interleaved_index : 0.000003s : 0.00% optimize.merge_cast_opt : 0.000002s : 0.00% optimize.slice_recompute_activation : 0.000002s : 0.00% optimize.micro_interleaved_order_control : 0.000003s : 0.00% optimize.assign_add_opt : 0.000002s : 0.00% optimize.ForceFp32Comm : 0.000001s : 0.00% optimize.remove_cast_before_assign_add : 0.000001s : 0.00% optimize.full_micro_interleaved_order_control : 0.000002s : 0.00% optimize.reorder_send_recv_between_fp_bp : 0.000003s : 0.00% optimize.comm_op_add_attrs : 0.000001s : 0.00% optimize.add_comm_op_reuse_tag : 0.000001s : 0.00% optimize.interleave_split_concat_branches : 0.000001s : 0.00% optimize.interleave_parallel_branches : 0.000001s : 0.00% optimize.overlap_opt_shard_in_pipeline : 0.000002s : 0.00% optimize.overlap_opt_shard_grad_in_pipeline : 0.000002s : 0.00% optimize.control_data_broadcast_order : 0.000041s : 0.01% optimize.grouped_pairwise_exchange_alltoall : 0.000002s : 0.00% optimize.offloading_packed_experts : 0.000013s : 0.00% optimize.overlap_recompute_and_grad_model_parallel : 0.000012s : 0.00% optimize.overlap_grad_matmul_and_grad_allreduce : 0.000001s : 0.00% optimize.overlap_recompute_allgather_and_fa_grad : 0.000002s : 0.00% optimize.overlap_recompute_comm : 0.000003s : 0.00% optimize.overlap_grad_ring_attention : 0.000011s : 0.00% optimize.overlap_grad_flash_sp : 0.000052s : 0.02% optimize.begin_end_overlap_inline : 0.000001s : 0.00% optimize.split_matmul_comm_elemetwise : 0.000003s : 0.00% optimize.split_layernorm_comm : 0.000002s : 0.00% optimize.handle_group_info : 0.000001s : 0.00% optimize.symbol_engine_optimizer.build : 0.000305s : 0.09% optimize.symbol_engine_optimizer.elim_shapecalc : 0.000036s : 0.01% optimize.symbol_engine_optimizer.elim_not_effective : 0.000059s : 0.02% optimize.symbol_engine_optimizer.opt_reshape : 0.000039s : 0.01% optimize.symbol_engine_optimizer.fold_const_symbol : 0.000067s : 0.02% optimize.symbol_engine_optimizer.renormalize : 0.000000s : 0.00% detach_backward : 0.000003s : 0.00% pipeline_parallel_scheduler : 0.000002s : 0.00% auto_monad_reorder : 0.000065s : 0.02% get_jit_bprop_graph : 0.000002s : 0.00% rewriter_after_jit_bprop_graph : 0.000005s : 0.00% opt_after_jit_grad : 0.000784s : 0.24% validate : 0.000816s : 0.25% Time group info: ------[substitution.] 0.003591 569 0.47% : 0.000017s : 8: substitution.depend_value_elim 0.41% : 0.000015s : 14: substitution.elim_not_effective 0.50% : 0.000018s : 12: substitution.float_tuple_getitem_switch 0.82% : 0.000029s : 14: substitution.fold_const_symbol 0.46% : 0.000017s : 17: substitution.graph_param_transform 69.30% : 0.002488s : 71: substitution.inline 0.36% : 0.000013s : 29: substitution.j_node_and_user_rematch 1.00% : 0.000036s : 6: substitution.less_batch_normalization 1.49% : 0.000053s : 2: substitution.list_to_tuple_eliminator_ 0.19% : 0.000007s : 6: substitution.load_eliminater 0.91% : 0.000033s : 21: substitution.minmaximum_grad 0.26% : 0.000009s : 2: substitution.opt_reshape 0.56% : 0.000020s : 29: substitution.remove_not_recompute_node 0.20% : 0.000007s : 4: substitution.replace_old_param 1.71% : 0.000061s : 18: substitution.reshape_eliminate 1.40% : 0.000050s : 24: substitution.switch_simplify 4.50% : 0.000161s : 51: substitution.tuple_list_convert_item_index_to_positive 2.36% : 0.000085s : 51: substitution.tuple_list_get_item_const_eliminator 3.06% : 0.000110s : 51: substitution.tuple_list_get_item_depend_reorder 5.78% : 0.000208s : 73: substitution.tuple_list_get_item_eliminator 2.98% : 0.000107s : 51: substitution.tuple_list_get_set_item_eliminator 0.71% : 0.000025s : 6: substitution.updatestate_pure_node_eliminater 0.59% : 0.000021s : 9: substitution.updatestate_useless_node_eliminater ------[type_inference.] 0.285968 2 93.39% : 0.267060s : 1: type_inference.infer 6.61% : 0.018908s : 1: type_inference.specialize ------[replace.] 0.001125 115 3.61% : 0.000041s : 5: replace.depend_value_elim 57.10% : 0.000642s : 71: replace.inline 3.67% : 0.000041s : 2: replace.list_to_tuple_eliminator_ 26.12% : 0.000294s : 24: replace.switch_simplify 9.51% : 0.000107s : 13: replace.tuple_list_get_item_eliminator ------[match.] 0.002565 115 0.09% : 0.000002s : 5: match.depend_value_elim 95.25% : 0.002443s : 71: match.inline 1.97% : 0.000050s : 2: match.list_to_tuple_eliminator_ 1.39% : 0.000036s : 24: match.switch_simplify 1.30% : 0.000033s : 13: match.tuple_list_get_item_eliminator ------[predicate.] 0.001707 9701 1.21% : 0.000021s : 123: predicate.accumulaten_eliminater 0.37% : 0.000006s : 17: predicate.ad_related_special_op_eliminate 0.50% : 0.000008s : 58: predicate.addn_check_dump 1.23% : 0.000021s : 123: predicate.addn_zero_filter 1.16% : 0.000020s : 123: predicate.adjust_all_reduce_mul_add 2.11% : 0.000036s : 181: predicate.arithmetic_simplify 1.12% : 0.000019s : 123: predicate.cast_eliminate 0.29% : 0.000005s : 34: predicate.check_bprop_eliminate 0.48% : 0.000008s : 58: predicate.compare_switch_simplify 0.09% : 0.000001s : 17: predicate.const_output_eliminate 0.61% : 0.000010s : 61: predicate.depend_value_elim 1.26% : 0.000022s : 123: predicate.dict_get_item_const_eliminator 1.20% : 0.000020s : 123: predicate.dict_get_item_eliminator 1.16% : 0.000020s : 123: predicate.dict_set_item_eliminator 0.35% : 0.000006s : 34: predicate.dumpgradient_eliminate 0.10% : 0.000002s : 17: predicate.elim_not_effective 0.24% : 0.000004s : 17: predicate.elim_shapecalc_of_broadcastargs 1.28% : 0.000022s : 140: predicate.environ_add_const_eliminate 1.18% : 0.000020s : 140: predicate.environ_get_add_eliminate 1.16% : 0.000020s : 140: predicate.environ_get_depend_swap 1.71% : 0.000029s : 198: predicate.environ_get_eliminate 1.19% : 0.000020s : 140: predicate.environ_get_set_eliminate 1.81% : 0.000031s : 209: predicate.exchange_switch_depend_value 2.58% : 0.000044s : 209: predicate.float_depend_g_call 0.50% : 0.000008s : 58: predicate.float_environ_get_switch 0.66% : 0.000011s : 75: predicate.float_tuple_getitem_switch 0.09% : 0.000002s : 17: predicate.fold_const_symbol 0.31% : 0.000005s : 35: predicate.get_grad_eliminate 0.09% : 0.000001s : 17: predicate.graph_param_transform 0.49% : 0.000008s : 58: predicate.incorporate_call 0.44% : 0.000007s : 58: predicate.incorporate_call_switch 5.38% : 0.000092s : 459: predicate.inline 0.45% : 0.000008s : 35: predicate.inline_without_move 0.15% : 0.000003s : 35: predicate.j_node_and_user_rematch 0.47% : 0.000008s : 36: predicate.less_batch_normalization 3.83% : 0.000065s : 172: predicate.list_to_tuple_eliminator_ 2.66% : 0.000045s : 295: predicate.load_eliminater 0.40% : 0.000007s : 17: predicate.loop_unroll_after_grad 3.43% : 0.000059s : 314: predicate.loop_unroll_before_grad 1.49% : 0.000025s : 157: predicate.make_slice_get_slice_eliminator 0.50% : 0.000008s : 58: predicate.merge_addn 0.28% : 0.000005s : 34: predicate.micro_step_allgather_replace 0.29% : 0.000005s : 34: predicate.mini_step_allgather_replace 1.08% : 0.000018s : 123: predicate.minmaximum_grad 0.47% : 0.000008s : 17: predicate.mutable_eliminate 0.19% : 0.000003s : 17: predicate.opt_reshape 0.17% : 0.000003s : 17: predicate.parallel_virtual_node 3.37% : 0.000058s : 209: predicate.partial_defer_inline 1.48% : 0.000025s : 155: predicate.partial_eliminate 1.12% : 0.000019s : 123: predicate.print_const_string_wrapper 0.46% : 0.000008s : 53: predicate.reduce_all_const_elim 1.54% : 0.000026s : 123: predicate.reduce_eliminate 2.58% : 0.000044s : 295: predicate.redundant_stop_gradient_eliminater 0.18% : 0.000003s : 35: predicate.remove_not_recompute_node 1.01% : 0.000017s : 172: predicate.replace_applicator 0.16% : 0.000003s : 35: predicate.replace_old_param 0.08% : 0.000001s : 17: predicate.reset_defer_inline 1.25% : 0.000021s : 123: predicate.reshape_eliminate 0.30% : 0.000005s : 34: predicate.row_tensor_add_zeros_like 0.22% : 0.000004s : 17: predicate.row_tensor_eliminate 0.47% : 0.000008s : 34: predicate.same_eliminate 0.20% : 0.000003s : 40: predicate.set_cell_output_no_recompute 0.36% : 0.000006s : 35: predicate.shard_identity_eliminate 0.34% : 0.000006s : 34: predicate.special_op_eliminate 0.56% : 0.000010s : 58: predicate.specialize_transform 0.41% : 0.000007s : 34: predicate.split_environ_get_set_with_tuple_value 0.36% : 0.000006s : 35: predicate.stack_unstack_eliminate 0.17% : 0.000003s : 17: predicate.switch_call_monad_eliminater 2.11% : 0.000036s : 209: predicate.switch_defer_inline 2.28% : 0.000039s : 243: predicate.switch_layer_defer_inline 8.27% : 0.000141s : 646: predicate.switch_simplify 1.30% : 0.000022s : 123: predicate.tile_eliminate 1.06% : 0.000018s : 123: predicate.transpose_eliminate 1.63% : 0.000028s : 157: predicate.tuple_list_convert_item_index_to_positive 1.89% : 0.000032s : 157: predicate.tuple_list_get_item_const_eliminator 1.52% : 0.000026s : 157: predicate.tuple_list_get_item_depend_reorder 2.94% : 0.000050s : 228: predicate.tuple_list_get_item_eliminator 1.61% : 0.000027s : 157: predicate.tuple_list_get_set_item_eliminator 2.23% : 0.000038s : 215: predicate.tuple_list_set_item_eliminator 1.67% : 0.000029s : 170: predicate.tuple_to_list_eliminator_ 2.46% : 0.000042s : 295: predicate.updatestate_pure_node_eliminater 3.03% : 0.000052s : 353: predicate.updatestate_useless_node_eliminater 0.16% : 0.000003s : 17: predicate.value_based_eliminate 0.33% : 0.000006s : 35: predicate.virtual_dataset_eliminate 0.31% : 0.000005s : 35: predicate.virtual_output_eliminate 0.15% : 0.000003s : 17: predicate.virtual_view_grad_eliminate 0.22% : 0.000004s : 17: predicate.zero_like_fill_zero ------[func_graph_cloner_run.] 0.014992 135 58.31% : 0.008742s : 55: func_graph_cloner_run.FuncGraphClonerGraph 41.69% : 0.006250s : 80: func_graph_cloner_run.FuncGraphSpecializer ------[meta_graph.] 0.000000 0 ------[manager.] 0.000000 0 ------[pynative] 0.000000 0 ------[others.] 0.400483 209 0.00% : 0.000004s : 1: ForceFp32Comm 1.34% : 0.005359s : 1: add_attr 1.33% : 0.005340s : 1: add_attr_with_inline 0.00% : 0.000005s : 1: add_comm_op_reuse_tag 0.04% : 0.000175s : 1: add_recomputation 0.00% : 0.000005s : 1: assign_add_opt 0.20% : 0.000788s : 1: auto_monad 0.02% : 0.000071s : 1: auto_monad_reorder 0.00% : 0.000005s : 1: begin_end_overlap_inline 0.00% : 0.000007s : 1: bias_add_comm_swap 0.16% : 0.000653s : 1: bootstrap 0.01% : 0.000046s : 1: cconv 0.00% : 0.000004s : 1: comm_op_add_attrs 0.01% : 0.000045s : 1: control_data_broadcast_order 0.01% : 0.000031s : 1: convert_after_rewriter 0.02% : 0.000081s : 1: cse_after_recomputation 0.00% : 0.000006s : 1: dataset_repeat_opt 0.00% : 0.000006s : 1: detach_backward 0.01% : 0.000028s : 1: environ_conv 0.06% : 0.000229s : 1: event_method 0.00% : 0.000006s : 1: full_micro_interleaved_order_control 0.00% : 0.000006s : 1: get_jit_bprop_graph 0.01% : 0.000058s : 1: graph_reusing 0.00% : 0.000005s : 1: grouped_pairwise_exchange_alltoall 0.00% : 0.000005s : 1: handle_group_info 0.00% : 0.000007s : 1: inline 0.00% : 0.000006s : 1: insert-virtual-dataset 0.00% : 0.000004s : 1: interleave_parallel_branches 0.00% : 0.000005s : 1: interleave_split_concat_branches 0.00% : 0.000006s : 1: label_fine_grained_interleaved_index 0.00% : 0.000010s : 1: label_micro_interleaved_index 0.19% : 0.000742s : 1: loop_unroll 0.00% : 0.000006s : 1: merge_cast_opt 0.00% : 0.000006s : 1: micro_interleaved_order_control 0.21% : 0.000843s : 1: mutable_eliminate 0.00% : 0.000016s : 1: offloading_packed_experts 0.01% : 0.000044s : 1: opt.transform.loop_unroll_optimizer 0.01% : 0.000052s : 1: opt.transform.mutable_eliminate 3.37% : 0.013497s : 95: opt.transform.opt_a 0.04% : 0.000145s : 1: opt.transform.opt_after_cconv 0.02% : 0.000080s : 1: opt.transform.opt_after_jit_grad 0.14% : 0.000563s : 28: opt.transform.opt_b 0.06% : 0.000248s : 2: opt.transform.opt_trans_graph 0.05% : 0.000195s : 4: opt.transform.symbol_engine_opt 7.51% : 0.030060s : 1: opt_a 0.09% : 0.000346s : 1: opt_after_cconv 0.20% : 0.000798s : 1: opt_after_jit_grad 0.23% : 0.000903s : 1: opt_b 9.17% : 0.036725s : 1: optimize 0.01% : 0.000050s : 1: optimize_parallel_all_gather_comm 0.01% : 0.000021s : 1: order_py_execute_after_rewriter 0.01% : 0.000056s : 1: overlap_grad_flash_sp 0.00% : 0.000004s : 1: overlap_grad_matmul_and_grad_allreduce 0.00% : 0.000014s : 1: overlap_grad_ring_attention 0.00% : 0.000006s : 1: overlap_opt_shard_grad_in_pipeline 0.00% : 0.000005s : 1: overlap_opt_shard_in_pipeline 0.00% : 0.000007s : 1: overlap_param_gather 0.00% : 0.000005s : 1: overlap_recompute_allgather_and_fa_grad 0.00% : 0.000016s : 1: overlap_recompute_and_grad_model_parallel 0.00% : 0.000008s : 1: overlap_recompute_comm 0.00% : 0.000009s : 1: parallel-infer-symbol 0.00% : 0.000004s : 1: parallel-infer-symbol-second 0.00% : 0.000006s : 1: partial_unused_args_eliminate 0.00% : 0.000005s : 1: pipeline_parallel_scheduler 0.00% : 0.000005s : 1: pipeline_split 0.06% : 0.000254s : 1: pre_auto_parallel 0.07% : 0.000262s : 1: py_interpret_to_execute 0.01% : 0.000037s : 1: py_interpret_to_execute_after_opt_a 0.00% : 0.000004s : 1: remove_cast_before_assign_add 0.03% : 0.000112s : 1: remove_dup_value 1.50% : 0.006015s : 1: renormalize.infer 1.57% : 0.006277s : 1: renormalize.specialize 0.00% : 0.000008s : 1: reorder_send_recv_between_fp_bp 0.00% : 0.000009s : 1: rewriter_after_jit_bprop_graph 0.23% : 0.000907s : 1: rewriter_after_opt_a 0.20% : 0.000807s : 1: rewriter_before_opt_a 0.00% : 0.000006s : 1: slice_cell_reuse_recomputed_activation 0.00% : 0.000005s : 1: slice_recompute_activation 0.00% : 0.000005s : 1: split_layernorm_comm 0.00% : 0.000006s : 1: split_matmul_comm_elemetwise 0.01% : 0.000024s : 1: swap_dp_allreduce_reducescatter 0.14% : 0.000563s : 1: symbol_engine_optimizer 0.07% : 0.000289s : 1: tuple_transform 71.51% : 0.286392s : 1: type_inference . [hook] pytest_runtest_teardown:test_matmul_qkv_bf16_1[True-mstype0-k_n_shape2-1024] tests/st/infer/ops/test_internal_ops/test_matmul_split.py::test_matmul_qkv_bf16_1[True-mstype0-k_n_shape2-1024],max_mem:260.0M [WARNING] ME(168319:281472866918192,MainProcess):2026-01-29-17:45:21.935.778 [mindspore/context.py:1334] For 'context.set_context', the parameter 'device_target' will be deprecated and removed in a future version. Please use the api mindspore.set_device() instead. TotalTime = 0.330411, [21] [bootstrap]: 0.00081594 [type_inference]: 0.288403 [event_method]: 0.00021233 [auto_monad]: 0.00074808 [graph_reusing]: 5.133e-05 [inline]: 2.76e-06 [add_attr]: 0.00520652, [1] [add_attr_with_inline]: 0.00518823, [1] [Cycle 1]: 0.000311, [2] [tag_attr]: 0.00020681 [meta_addattr_fg_expand]: 5.428e-05 [parallel-infer-symbol]: 4.08001e-06 [pre_auto_parallel]: 0.00024457 [insert-virtual-dataset]: 2.69001e-06 [parallel-infer-symbol-second]: 1.37e-06 [dataset_repeat_opt]: 2.47001e-06 [pipeline_split]: 1.66998e-06 [optimize]: 0.0336848, [53] [py_interpret_to_execute]: 0.00024903 [rewriter_before_opt_a]: 0.00078341 [opt_a]: 0.0275699, [2] [Cycle 1]: 0.0253974, [45] [expand_dump_flag]: 1.793e-05 [switch_simplify]: 0.00093667 [loop_unroll]: 0.00033054 [a_1]: 0.0094294 [with_stream_mark]: 5.383e-05 [recompute_prepare]: 3.722e-05 [updatestate_depend_eliminate]: 2.054e-05 [updatestate_assign_eliminate]: 1.518e-05 [updatestate_loads_eliminate]: 1.855e-05 [parameter_eliminate]: 2.36998e-06 [a_2]: 0.00077583 [accelerated_algorithm]: 6.472e-05 [shard]: 2.36e-06 [meta_shard_fg_expand]: 1.958e-05 [shard_inline]: 2.315e-05 [merge_send_recv]: 1.976e-05 [auto_parallel]: 2.079e-05 [parallel]: 3.473e-05 [flash_sp]: 1.538e-05 [merge_comm]: 1.427e-05 [allreduce_fusion]: 1.317e-05 [matmul_add_comm_reduction]: 2.225e-05 [allreduce_slice_to_reducescatter]: 7.50006e-07 [virtual_shard_identity]: 2.791e-05 [virtual_dataset]: 2.277e-05 [get_grad_eliminate_]: 2.046e-05 [virtual_output]: 3.342e-05 [merge_forward]: 1.353e-05 [cell_reuse_recompute_pass]: 1.89999e-06 [offload_activation]: 2.419e-05 [cell_reuse_handle_not_recompute_node_pass]: 4.521e-05 [merge_recompute_call_nodes]: 1.49998e-06 [before_grad]: 4.254e-05 [set_forward_comm_id_for_comm_node_pass]: 1.404e-05 [meta_fg_expand]: 1.654e-05 [flash_sp_send_recv_attached]: 5.94999e-06 [receive_attached]: 2.96001e-06 [after_resolve]: 2.877e-05 [a_after_grad]: 3.702e-05 [renormalize]: 0.0114527 [add_forward_monad_depend]: 2.096e-05 [auto_monad_grad]: 2.73e-06 [auto_monad_eliminator]: 7.694e-05 [cse]: 0.00102501 [a_3]: 0.0001759 [Cycle 2]: 0.00215244, [45] [expand_dump_flag]: 2.86999e-06 [switch_simplify]: 2.581e-05 [loop_unroll]: 2.147e-05 [a_1]: 0.000655 [with_stream_mark]: 3.294e-05 [recompute_prepare]: 2.248e-05 [updatestate_depend_eliminate]: 1.656e-05 [updatestate_assign_eliminate]: 1.141e-05 [updatestate_loads_eliminate]: 1.401e-05 [parameter_eliminate]: 2.45002e-06 [a_2]: 0.00035082 [accelerated_algorithm]: 3.345e-05 [shard]: 2.22999e-06 [meta_shard_fg_expand]: 6.82002e-06 [shard_inline]: 2.104e-05 [merge_send_recv]: 2.157e-05 [auto_parallel]: 2.198e-05 [parallel]: 1.065e-05 [flash_sp]: 5.35999e-06 [merge_comm]: 1.315e-05 [allreduce_fusion]: 1.207e-05 [matmul_add_comm_reduction]: 2.318e-05 [allreduce_slice_to_reducescatter]: 9.40025e-07 [virtual_shard_identity]: 2.285e-05 [virtual_dataset]: 2.091e-05 [get_grad_eliminate_]: 2.082e-05 [virtual_output]: 1.944e-05 [merge_forward]: 1.234e-05 [cell_reuse_recompute_pass]: 3.42997e-06 [offload_activation]: 2.512e-05 [cell_reuse_handle_not_recompute_node_pass]: 3.886e-05 [merge_recompute_call_nodes]: 1.60999e-06 [before_grad]: 3.635e-05 [set_forward_comm_id_for_comm_node_pass]: 1.328e-05 [meta_fg_expand]: 9.72999e-06 [flash_sp_send_recv_attached]: 2.34001e-06 [receive_attached]: 2.78003e-06 [after_resolve]: 3.235e-05 [a_after_grad]: 3.299e-05 [renormalize]: 8.00064e-08 [add_forward_monad_depend]: 3.68999e-06 [auto_monad_grad]: 2.66e-06 [auto_monad_eliminator]: 5.285e-05 [cse]: 7.679e-05 [a_3]: 0.00014397 [py_interpret_to_execute_after_opt_a]: 3.449e-05 [slice_cell_reuse_recomputed_activation]: 2.76e-06 [rewriter_after_opt_a]: 0.00090366 [convert_after_rewriter]: 2.617e-05 [order_py_execute_after_rewriter]: 1.484e-05 [mutable_eliminate]: 0.00081263 [opt_b]: 0.00076184, [1] [Cycle 1]: 0.00075262, [7] [b_1]: 0.00055659 [b_2]: 2.488e-05 [updatestate_depend_eliminate]: 1.824e-05 [updatestate_assign_eliminate]: 1.146e-05 [updatestate_loads_eliminate]: 1.377e-05 [renormalize]: 8.39995e-07 [cse]: 7.79e-05 [optimize_parallel_all_gather_comm]: 3.783e-05 [overlap_param_gather]: 2.20002e-06 [cconv]: 3.746e-05 [loop_unroll]: 0.00054883 [opt_after_cconv]: 0.00030485, [1] [Cycle 1]: 0.0002966, [7] [c_1]: 0.00013057 [parameter_eliminate]: 4.97e-06 [updatestate_depend_eliminate]: 1.769e-05 [updatestate_assign_eliminate]: 1.165e-05 [updatestate_loads_eliminate]: 1.387e-05 [cse]: 7.731e-05 [renormalize]: 3.29979e-07 [remove_dup_value]: 9.322e-05 [tuple_transform]: 0.00025297, [1] [Cycle 1]: 0.00024684, [4] [d_1]: 0.00019471 [none_parameter_eliminate]: 2.41998e-06 [renormalize]: 3.30008e-07 [switch_simplify]: 2.29e-05 [partial_unused_args_eliminate]: 1.75001e-06 [add_recomputation]: 0.00014592 [cse_after_recomputation]: 7.18e-05, [1] [Cycle 1]: 6.53e-05, [1] [cse]: 5.683e-05 [environ_conv]: 2.089e-05 [swap_dp_allreduce_reducescatter]: 1.918e-05 [bias_add_comm_swap]: 2.81e-06 [label_micro_interleaved_index]: 5.22e-06 [label_fine_grained_interleaved_index]: 3.02002e-06 [merge_cast_opt]: 1.57001e-06 [slice_recompute_activation]: 2.48e-06 [micro_interleaved_order_control]: 2.94999e-06 [assign_add_opt]: 1.31002e-06 [ForceFp32Comm]: 8.80013e-07 [remove_cast_before_assign_add]: 1.42e-06 [full_micro_interleaved_order_control]: 2.27999e-06 [reorder_send_recv_between_fp_bp]: 2.77002e-06 [comm_op_add_attrs]: 1.26002e-06 [add_comm_op_reuse_tag]: 1.04e-06 [interleave_split_concat_branches]: 1.28002e-06 [interleave_parallel_branches]: 1.10001e-06 [overlap_opt_shard_in_pipeline]: 1.24e-06 [overlap_opt_shard_grad_in_pipeline]: 2.12001e-06 [control_data_broadcast_order]: 3.768e-05 [grouped_pairwise_exchange_alltoall]: 1.56998e-06 [offloading_packed_experts]: 1.087e-05 [overlap_recompute_and_grad_model_parallel]: 1.085e-05 [overlap_grad_matmul_and_grad_allreduce]: 1.24e-06 [overlap_recompute_allgather_and_fa_grad]: 1.47001e-06 [overlap_recompute_comm]: 2.36998e-06 [overlap_grad_ring_attention]: 9.52999e-06 [overlap_grad_flash_sp]: 4.894e-05 [begin_end_overlap_inline]: 5.59987e-07 [split_matmul_comm_elemetwise]: 2.49999e-06 [split_layernorm_comm]: 1.76003e-06 [handle_group_info]: 1.02998e-06 [symbol_engine_optimizer]: 0.00050828, [1] [Cycle 1]: 0.0005024, [6] [build]: 0.00028929 [elim_shapecalc]: 3.216e-05 [elim_not_effective]: 4.892e-05 [opt_reshape]: 3.256e-05 [fold_const_symbol]: 5.36e-05 [renormalize]: 3.10014e-07 [detach_backward]: 2.76e-06 [pipeline_parallel_scheduler]: 1.91e-06 [auto_monad_reorder]: 6.024e-05 [get_jit_bprop_graph]: 1.65001e-06 [rewriter_after_jit_bprop_graph]: 3.91999e-06 [opt_after_jit_grad]: 0.00060067 [validate]: 9.413e-05 Sums bootstrap : 0.000816s : 0.25% type_inference : 0.288403s : 89.01% event_method : 0.000212s : 0.07% auto_monad : 0.000748s : 0.23% graph_reusing : 0.000051s : 0.02% inline : 0.000003s : 0.00% add_attr.add_attr_with_inline.tag_attr : 0.000207s : 0.06% add_attr.add_attr_with_inline.meta_addattr_fg_expand : 0.000054s : 0.02% parallel-infer-symbol : 0.000004s : 0.00% pre_auto_parallel : 0.000245s : 0.08% insert-virtual-dataset : 0.000003s : 0.00% parallel-infer-symbol-second : 0.000001s : 0.00% dataset_repeat_opt : 0.000002s : 0.00% pipeline_split : 0.000002s : 0.00% optimize.py_interpret_to_execute : 0.000249s : 0.08% optimize.rewriter_before_opt_a : 0.000783s : 0.24% optimize.opt_a.expand_dump_flag : 0.000021s : 0.01% optimize.opt_a.switch_simplify : 0.000962s : 0.30% optimize.opt_a.loop_unroll : 0.000352s : 0.11% optimize.opt_a.a_1 : 0.010084s : 3.11% optimize.opt_a.with_stream_mark : 0.000087s : 0.03% optimize.opt_a.recompute_prepare : 0.000060s : 0.02% optimize.opt_a.updatestate_depend_eliminate : 0.000037s : 0.01% optimize.opt_a.updatestate_assign_eliminate : 0.000027s : 0.01% optimize.opt_a.updatestate_loads_eliminate : 0.000033s : 0.01% optimize.opt_a.parameter_eliminate : 0.000005s : 0.00% optimize.opt_a.a_2 : 0.001127s : 0.35% optimize.opt_a.accelerated_algorithm : 0.000098s : 0.03% optimize.opt_a.shard : 0.000005s : 0.00% optimize.opt_a.meta_shard_fg_expand : 0.000026s : 0.01% optimize.opt_a.shard_inline : 0.000044s : 0.01% optimize.opt_a.merge_send_recv : 0.000041s : 0.01% optimize.opt_a.auto_parallel : 0.000043s : 0.01% optimize.opt_a.parallel : 0.000045s : 0.01% optimize.opt_a.flash_sp : 0.000021s : 0.01% optimize.opt_a.merge_comm : 0.000027s : 0.01% optimize.opt_a.allreduce_fusion : 0.000025s : 0.01% optimize.opt_a.matmul_add_comm_reduction : 0.000045s : 0.01% optimize.opt_a.allreduce_slice_to_reducescatter : 0.000002s : 0.00% optimize.opt_a.virtual_shard_identity : 0.000051s : 0.02% optimize.opt_a.virtual_dataset : 0.000044s : 0.01% optimize.opt_a.get_grad_eliminate_ : 0.000041s : 0.01% optimize.opt_a.virtual_output : 0.000053s : 0.02% optimize.opt_a.merge_forward : 0.000026s : 0.01% optimize.opt_a.cell_reuse_recompute_pass : 0.000005s : 0.00% optimize.opt_a.offload_activation : 0.000049s : 0.02% optimize.opt_a.cell_reuse_handle_not_recompute_node_pass : 0.000084s : 0.03% optimize.opt_a.merge_recompute_call_nodes : 0.000003s : 0.00% optimize.opt_a.before_grad : 0.000079s : 0.02% optimize.opt_a.set_forward_comm_id_for_comm_node_pass : 0.000027s : 0.01% optimize.opt_a.meta_fg_expand : 0.000026s : 0.01% optimize.opt_a.flash_sp_send_recv_attached : 0.000008s : 0.00% optimize.opt_a.receive_attached : 0.000006s : 0.00% optimize.opt_a.after_resolve : 0.000061s : 0.02% optimize.opt_a.a_after_grad : 0.000070s : 0.02% optimize.opt_a.renormalize : 0.011453s : 3.53% optimize.opt_a.add_forward_monad_depend : 0.000025s : 0.01% optimize.opt_a.auto_monad_grad : 0.000005s : 0.00% optimize.opt_a.auto_monad_eliminator : 0.000130s : 0.04% optimize.opt_a.cse : 0.001102s : 0.34% optimize.opt_a.a_3 : 0.000320s : 0.10% optimize.py_interpret_to_execute_after_opt_a : 0.000034s : 0.01% optimize.slice_cell_reuse_recomputed_activation : 0.000003s : 0.00% optimize.rewriter_after_opt_a : 0.000904s : 0.28% optimize.convert_after_rewriter : 0.000026s : 0.01% optimize.order_py_execute_after_rewriter : 0.000015s : 0.00% optimize.mutable_eliminate : 0.000813s : 0.25% optimize.opt_b.b_1 : 0.000557s : 0.17% optimize.opt_b.b_2 : 0.000025s : 0.01% optimize.opt_b.updatestate_depend_eliminate : 0.000018s : 0.01% optimize.opt_b.updatestate_assign_eliminate : 0.000011s : 0.00% optimize.opt_b.updatestate_loads_eliminate : 0.000014s : 0.00% optimize.opt_b.renormalize : 0.000001s : 0.00% optimize.opt_b.cse : 0.000078s : 0.02% optimize.optimize_parallel_all_gather_comm : 0.000038s : 0.01% optimize.overlap_param_gather : 0.000002s : 0.00% optimize.cconv : 0.000037s : 0.01% optimize.loop_unroll : 0.000549s : 0.17% optimize.opt_after_cconv.c_1 : 0.000131s : 0.04% optimize.opt_after_cconv.parameter_eliminate : 0.000005s : 0.00% optimize.opt_after_cconv.updatestate_depend_eliminate : 0.000018s : 0.01% optimize.opt_after_cconv.updatestate_assign_eliminate : 0.000012s : 0.00% optimize.opt_after_cconv.updatestate_loads_eliminate : 0.000014s : 0.00% optimize.opt_after_cconv.cse : 0.000077s : 0.02% optimize.opt_after_cconv.renormalize : 0.000000s : 0.00% optimize.remove_dup_value : 0.000093s : 0.03% optimize.tuple_transform.d_1 : 0.000195s : 0.06% optimize.tuple_transform.none_parameter_eliminate : 0.000002s : 0.00% optimize.tuple_transform.renormalize : 0.000000s : 0.00% optimize.tuple_transform.switch_simplify : 0.000023s : 0.01% optimize.partial_unused_args_eliminate : 0.000002s : 0.00% optimize.add_recomputation : 0.000146s : 0.05% optimize.cse_after_recomputation.cse : 0.000057s : 0.02% optimize.environ_conv : 0.000021s : 0.01% optimize.swap_dp_allreduce_reducescatter : 0.000019s : 0.01% optimize.bias_add_comm_swap : 0.000003s : 0.00% optimize.label_micro_interleaved_index : 0.000005s : 0.00% optimize.label_fine_grained_interleaved_index : 0.000003s : 0.00% optimize.merge_cast_opt : 0.000002s : 0.00% optimize.slice_recompute_activation : 0.000002s : 0.00% optimize.micro_interleaved_order_control : 0.000003s : 0.00% optimize.assign_add_opt : 0.000001s : 0.00% optimize.ForceFp32Comm : 0.000001s : 0.00% optimize.remove_cast_before_assign_add : 0.000001s : 0.00% optimize.full_micro_interleaved_order_control : 0.000002s : 0.00% optimize.reorder_send_recv_between_fp_bp : 0.000003s : 0.00% optimize.comm_op_add_attrs : 0.000001s : 0.00% optimize.add_comm_op_reuse_tag : 0.000001s : 0.00% optimize.interleave_split_concat_branches : 0.000001s : 0.00% optimize.interleave_parallel_branches : 0.000001s : 0.00% optimize.overlap_opt_shard_in_pipeline : 0.000001s : 0.00% optimize.overlap_opt_shard_grad_in_pipeline : 0.000002s : 0.00% optimize.control_data_broadcast_order : 0.000038s : 0.01% optimize.grouped_pairwise_exchange_alltoall : 0.000002s : 0.00% optimize.offloading_packed_experts : 0.000011s : 0.00% optimize.overlap_recompute_and_grad_model_parallel : 0.000011s : 0.00% optimize.overlap_grad_matmul_and_grad_allreduce : 0.000001s : 0.00% optimize.overlap_recompute_allgather_and_fa_grad : 0.000001s : 0.00% optimize.overlap_recompute_comm : 0.000002s : 0.00% optimize.overlap_grad_ring_attention : 0.000010s : 0.00% optimize.overlap_grad_flash_sp : 0.000049s : 0.02% optimize.begin_end_overlap_inline : 0.000001s : 0.00% optimize.split_matmul_comm_elemetwise : 0.000002s : 0.00% optimize.split_layernorm_comm : 0.000002s : 0.00% optimize.handle_group_info : 0.000001s : 0.00% optimize.symbol_engine_optimizer.build : 0.000289s : 0.09% optimize.symbol_engine_optimizer.elim_shapecalc : 0.000032s : 0.01% optimize.symbol_engine_optimizer.elim_not_effective : 0.000049s : 0.02% optimize.symbol_engine_optimizer.opt_reshape : 0.000033s : 0.01% optimize.symbol_engine_optimizer.fold_const_symbol : 0.000054s : 0.02% optimize.symbol_engine_optimizer.renormalize : 0.000000s : 0.00% detach_backward : 0.000003s : 0.00% pipeline_parallel_scheduler : 0.000002s : 0.00% auto_monad_reorder : 0.000060s : 0.02% get_jit_bprop_graph : 0.000002s : 0.00% rewriter_after_jit_bprop_graph : 0.000004s : 0.00% opt_after_jit_grad : 0.000601s : 0.19% validate : 0.000094s : 0.03% Time group info: ------[substitution.] 0.003506 569 0.51% : 0.000018s : 8: substitution.depend_value_elim 0.34% : 0.000012s : 14: substitution.elim_not_effective 0.51% : 0.000018s : 12: substitution.float_tuple_getitem_switch 0.63% : 0.000022s : 14: substitution.fold_const_symbol 0.43% : 0.000015s : 17: substitution.graph_param_transform 67.94% : 0.002382s : 71: substitution.inline 0.39% : 0.000014s : 29: substitution.j_node_and_user_rematch 1.04% : 0.000036s : 6: substitution.less_batch_normalization 1.79% : 0.000063s : 2: substitution.list_to_tuple_eliminator_ 0.17% : 0.000006s : 6: substitution.load_eliminater 0.86% : 0.000030s : 21: substitution.minmaximum_grad 0.21% : 0.000007s : 2: substitution.opt_reshape 0.55% : 0.000019s : 29: substitution.remove_not_recompute_node 0.25% : 0.000009s : 4: substitution.replace_old_param 1.87% : 0.000066s : 18: substitution.reshape_eliminate 1.48% : 0.000052s : 24: substitution.switch_simplify 4.44% : 0.000156s : 51: substitution.tuple_list_convert_item_index_to_positive 2.46% : 0.000086s : 51: substitution.tuple_list_get_item_const_eliminator 3.63% : 0.000127s : 51: substitution.tuple_list_get_item_depend_reorder 6.19% : 0.000217s : 73: substitution.tuple_list_get_item_eliminator 3.11% : 0.000109s : 51: substitution.tuple_list_get_set_item_eliminator 0.67% : 0.000023s : 6: substitution.updatestate_pure_node_eliminater 0.54% : 0.000019s : 9: substitution.updatestate_useless_node_eliminater ------[type_inference.] 0.287963 2 93.64% : 0.269658s : 1: type_inference.infer 6.36% : 0.018306s : 1: type_inference.specialize ------[replace.] 0.001277 115 3.68% : 0.000047s : 5: replace.depend_value_elim 53.79% : 0.000687s : 71: replace.inline 3.82% : 0.000049s : 2: replace.list_to_tuple_eliminator_ 30.76% : 0.000393s : 24: replace.switch_simplify 7.95% : 0.000101s : 13: replace.tuple_list_get_item_eliminator ------[match.] 0.002469 115 0.08% : 0.000002s : 5: match.depend_value_elim 94.59% : 0.002335s : 71: match.inline 2.43% : 0.000060s : 2: match.list_to_tuple_eliminator_ 1.53% : 0.000038s : 24: match.switch_simplify 1.37% : 0.000034s : 13: match.tuple_list_get_item_eliminator ------[predicate.] 0.001620 9701 1.20% : 0.000019s : 123: predicate.accumulaten_eliminater 0.33% : 0.000005s : 17: predicate.ad_related_special_op_eliminate 0.48% : 0.000008s : 58: predicate.addn_check_dump 1.19% : 0.000019s : 123: predicate.addn_zero_filter 1.11% : 0.000018s : 123: predicate.adjust_all_reduce_mul_add 2.28% : 0.000037s : 181: predicate.arithmetic_simplify 1.18% : 0.000019s : 123: predicate.cast_eliminate 0.32% : 0.000005s : 34: predicate.check_bprop_eliminate 0.49% : 0.000008s : 58: predicate.compare_switch_simplify 0.09% : 0.000001s : 17: predicate.const_output_eliminate 0.59% : 0.000009s : 61: predicate.depend_value_elim 1.27% : 0.000021s : 123: predicate.dict_get_item_const_eliminator 1.43% : 0.000023s : 123: predicate.dict_get_item_eliminator 1.13% : 0.000018s : 123: predicate.dict_set_item_eliminator 0.37% : 0.000006s : 34: predicate.dumpgradient_eliminate 0.09% : 0.000001s : 17: predicate.elim_not_effective 0.21% : 0.000003s : 17: predicate.elim_shapecalc_of_broadcastargs 1.24% : 0.000020s : 140: predicate.environ_add_const_eliminate 1.30% : 0.000021s : 140: predicate.environ_get_add_eliminate 1.26% : 0.000020s : 140: predicate.environ_get_depend_swap 1.75% : 0.000028s : 198: predicate.environ_get_eliminate 1.26% : 0.000020s : 140: predicate.environ_get_set_eliminate 2.06% : 0.000033s : 209: predicate.exchange_switch_depend_value 2.91% : 0.000047s : 209: predicate.float_depend_g_call 0.49% : 0.000008s : 58: predicate.float_environ_get_switch 0.68% : 0.000011s : 75: predicate.float_tuple_getitem_switch 0.09% : 0.000001s : 17: predicate.fold_const_symbol 0.33% : 0.000005s : 35: predicate.get_grad_eliminate 0.09% : 0.000001s : 17: predicate.graph_param_transform 0.49% : 0.000008s : 58: predicate.incorporate_call 0.46% : 0.000007s : 58: predicate.incorporate_call_switch 5.87% : 0.000095s : 459: predicate.inline 0.38% : 0.000006s : 35: predicate.inline_without_move 0.15% : 0.000002s : 35: predicate.j_node_and_user_rematch 0.46% : 0.000007s : 36: predicate.less_batch_normalization 1.65% : 0.000027s : 172: predicate.list_to_tuple_eliminator_ 2.80% : 0.000045s : 295: predicate.load_eliminater 0.35% : 0.000006s : 17: predicate.loop_unroll_after_grad 3.35% : 0.000054s : 314: predicate.loop_unroll_before_grad 1.57% : 0.000026s : 157: predicate.make_slice_get_slice_eliminator 0.50% : 0.000008s : 58: predicate.merge_addn 0.30% : 0.000005s : 34: predicate.micro_step_allgather_replace 0.33% : 0.000005s : 34: predicate.mini_step_allgather_replace 1.17% : 0.000019s : 123: predicate.minmaximum_grad 0.49% : 0.000008s : 17: predicate.mutable_eliminate 0.17% : 0.000003s : 17: predicate.opt_reshape 0.18% : 0.000003s : 17: predicate.parallel_virtual_node 3.47% : 0.000056s : 209: predicate.partial_defer_inline 1.52% : 0.000025s : 155: predicate.partial_eliminate 1.14% : 0.000018s : 123: predicate.print_const_string_wrapper 0.46% : 0.000007s : 53: predicate.reduce_all_const_elim 1.54% : 0.000025s : 123: predicate.reduce_eliminate 2.62% : 0.000042s : 295: predicate.redundant_stop_gradient_eliminater 0.16% : 0.000003s : 35: predicate.remove_not_recompute_node 1.08% : 0.000017s : 172: predicate.replace_applicator 0.20% : 0.000003s : 35: predicate.replace_old_param 0.11% : 0.000002s : 17: predicate.reset_defer_inline 1.28% : 0.000021s : 123: predicate.reshape_eliminate 0.31% : 0.000005s : 34: predicate.row_tensor_add_zeros_like 0.19% : 0.000003s : 17: predicate.row_tensor_eliminate 0.46% : 0.000007s : 34: predicate.same_eliminate 0.20% : 0.000003s : 40: predicate.set_cell_output_no_recompute 0.40% : 0.000006s : 35: predicate.shard_identity_eliminate 0.35% : 0.000006s : 34: predicate.special_op_eliminate 0.57% : 0.000009s : 58: predicate.specialize_transform 0.44% : 0.000007s : 34: predicate.split_environ_get_set_with_tuple_value 0.41% : 0.000007s : 35: predicate.stack_unstack_eliminate 0.17% : 0.000003s : 17: predicate.switch_call_monad_eliminater 2.21% : 0.000036s : 209: predicate.switch_defer_inline 2.37% : 0.000038s : 243: predicate.switch_layer_defer_inline 6.96% : 0.000113s : 646: predicate.switch_simplify 1.19% : 0.000019s : 123: predicate.tile_eliminate 1.21% : 0.000020s : 123: predicate.transpose_eliminate 1.74% : 0.000028s : 157: predicate.tuple_list_convert_item_index_to_positive 1.76% : 0.000029s : 157: predicate.tuple_list_get_item_const_eliminator 1.60% : 0.000026s : 157: predicate.tuple_list_get_item_depend_reorder 3.25% : 0.000053s : 228: predicate.tuple_list_get_item_eliminator 1.62% : 0.000026s : 157: predicate.tuple_list_get_set_item_eliminator 2.27% : 0.000037s : 215: predicate.tuple_list_set_item_eliminator 1.58% : 0.000026s : 170: predicate.tuple_to_list_eliminator_ 2.88% : 0.000047s : 295: predicate.updatestate_pure_node_eliminater 3.18% : 0.000051s : 353: predicate.updatestate_useless_node_eliminater 0.17% : 0.000003s : 17: predicate.value_based_eliminate 0.36% : 0.000006s : 35: predicate.virtual_dataset_eliminate 0.33% : 0.000005s : 35: predicate.virtual_output_eliminate 0.15% : 0.000003s : 17: predicate.virtual_view_grad_eliminate 0.22% : 0.000004s : 17: predicate.zero_like_fill_zero ------[func_graph_cloner_run.] 0.014848 135 58.67% : 0.008711s : 55: func_graph_cloner_run.FuncGraphClonerGraph 41.33% : 0.006137s : 80: func_graph_cloner_run.FuncGraphSpecializer ------[meta_graph.] 0.000000 0 ------[manager.] 0.000000 0 ------[pynative] 0.000000 0 ------[others.] 0.395093 209 0.00% : 0.000003s : 1: ForceFp32Comm 1.32% : 0.005213s : 1: add_attr 1.31% : 0.005194s : 1: add_attr_with_inline 0.00% : 0.000005s : 1: add_comm_op_reuse_tag 0.04% : 0.000152s : 1: add_recomputation 0.00% : 0.000005s : 1: assign_add_opt 0.19% : 0.000766s : 1: auto_monad 0.02% : 0.000066s : 1: auto_monad_reorder 0.00% : 0.000004s : 1: begin_end_overlap_inline 0.00% : 0.000006s : 1: bias_add_comm_swap 0.22% : 0.000884s : 1: bootstrap 0.01% : 0.000042s : 1: cconv 0.00% : 0.000004s : 1: comm_op_add_attrs 0.01% : 0.000041s : 1: control_data_broadcast_order 0.01% : 0.000032s : 1: convert_after_rewriter 0.02% : 0.000075s : 1: cse_after_recomputation 0.00% : 0.000006s : 1: dataset_repeat_opt 0.00% : 0.000006s : 1: detach_backward 0.01% : 0.000024s : 1: environ_conv 0.06% : 0.000225s : 1: event_method 0.00% : 0.000006s : 1: full_micro_interleaved_order_control 0.00% : 0.000006s : 1: get_jit_bprop_graph 0.01% : 0.000059s : 1: graph_reusing 0.00% : 0.000004s : 1: grouped_pairwise_exchange_alltoall 0.00% : 0.000005s : 1: handle_group_info 0.00% : 0.000006s : 1: inline 0.00% : 0.000006s : 1: insert-virtual-dataset 0.00% : 0.000004s : 1: interleave_parallel_branches 0.00% : 0.000004s : 1: interleave_split_concat_branches 0.00% : 0.000007s : 1: label_fine_grained_interleaved_index 0.00% : 0.000008s : 1: label_micro_interleaved_index 0.14% : 0.000560s : 1: loop_unroll 0.00% : 0.000005s : 1: merge_cast_opt 0.00% : 0.000006s : 1: micro_interleaved_order_control 0.21% : 0.000825s : 1: mutable_eliminate 0.00% : 0.000014s : 1: offloading_packed_experts 0.01% : 0.000035s : 1: opt.transform.loop_unroll_optimizer 0.01% : 0.000047s : 1: opt.transform.mutable_eliminate 3.39% : 0.013384s : 95: opt.transform.opt_a 0.03% : 0.000129s : 1: opt.transform.opt_after_cconv 0.02% : 0.000072s : 1: opt.transform.opt_after_jit_grad 0.14% : 0.000543s : 28: opt.transform.opt_b 0.05% : 0.000215s : 2: opt.transform.opt_trans_graph 0.04% : 0.000162s : 4: opt.transform.symbol_engine_opt 6.98% : 0.027575s : 1: opt_a 0.08% : 0.000309s : 1: opt_after_cconv 0.15% : 0.000612s : 1: opt_after_jit_grad 0.19% : 0.000766s : 1: opt_b 8.53% : 0.033692s : 1: optimize 0.01% : 0.000042s : 1: optimize_parallel_all_gather_comm 0.00% : 0.000019s : 1: order_py_execute_after_rewriter 0.01% : 0.000053s : 1: overlap_grad_flash_sp 0.00% : 0.000004s : 1: overlap_grad_matmul_and_grad_allreduce 0.00% : 0.000013s : 1: overlap_grad_ring_attention 0.00% : 0.000006s : 1: overlap_opt_shard_grad_in_pipeline 0.00% : 0.000004s : 1: overlap_opt_shard_in_pipeline 0.00% : 0.000007s : 1: overlap_param_gather 0.00% : 0.000004s : 1: overlap_recompute_allgather_and_fa_grad 0.00% : 0.000014s : 1: overlap_recompute_and_grad_model_parallel 0.00% : 0.000006s : 1: overlap_recompute_comm 0.00% : 0.000010s : 1: parallel-infer-symbol 0.00% : 0.000005s : 1: parallel-infer-symbol-second 0.00% : 0.000005s : 1: partial_unused_args_eliminate 0.00% : 0.000005s : 1: pipeline_parallel_scheduler 0.00% : 0.000005s : 1: pipeline_split 0.06% : 0.000254s : 1: pre_auto_parallel 0.07% : 0.000260s : 1: py_interpret_to_execute 0.01% : 0.000040s : 1: py_interpret_to_execute_after_opt_a 0.00% : 0.000004s : 1: remove_cast_before_assign_add 0.03% : 0.000100s : 1: remove_dup_value 1.47% : 0.005808s : 1: renormalize.infer 1.42% : 0.005627s : 1: renormalize.specialize 0.00% : 0.000006s : 1: reorder_send_recv_between_fp_bp 0.00% : 0.000007s : 1: rewriter_after_jit_bprop_graph 0.23% : 0.000917s : 1: rewriter_after_opt_a 0.20% : 0.000800s : 1: rewriter_before_opt_a 0.00% : 0.000006s : 1: slice_cell_reuse_recomputed_activation 0.00% : 0.000005s : 1: slice_recompute_activation 0.00% : 0.000004s : 1: split_layernorm_comm 0.00% : 0.000005s : 1: split_matmul_comm_elemetwise 0.01% : 0.000022s : 1: swap_dp_allreduce_reducescatter 0.13% : 0.000511s : 1: symbol_engine_optimizer 0.07% : 0.000258s : 1: tuple_transform 73.01% : 0.288438s : 1: type_inference . [hook] pytest_runtest_teardown:test_matmul_qkv_bf16_1[True-mstype0-k_n_shape3-256] tests/st/infer/ops/test_internal_ops/test_matmul_split.py::test_matmul_qkv_bf16_1[True-mstype0-k_n_shape3-256],max_mem:260.0M [WARNING] ME(168319:281472866918192,MainProcess):2026-01-29-17:45:38.738.201 [mindspore/context.py:1334] For 'context.set_context', the parameter 'device_target' will be deprecated and removed in a future version. Please use the api mindspore.set_device() instead. TotalTime = 0.331201, [21] [bootstrap]: 0.00068149 [type_inference]: 0.288135 [event_method]: 0.00023071 [auto_monad]: 0.00085553 [graph_reusing]: 5.403e-05 [inline]: 2.94001e-06 [add_attr]: 0.00542197, [1] [add_attr_with_inline]: 0.00540487, [1] [Cycle 1]: 0.00031148, [2] [tag_attr]: 0.00021044 [meta_addattr_fg_expand]: 5.766e-05 [parallel-infer-symbol]: 4.12998e-06 [pre_auto_parallel]: 0.00025405 [insert-virtual-dataset]: 2.54999e-06 [parallel-infer-symbol-second]: 1.12999e-06 [dataset_repeat_opt]: 2.36e-06 [pipeline_split]: 1.69e-06 [optimize]: 0.0345807, [53] [py_interpret_to_execute]: 0.00027721 [rewriter_before_opt_a]: 0.00081052 [opt_a]: 0.0286357, [2] [Cycle 1]: 0.0265293, [45] [expand_dump_flag]: 3.032e-05 [switch_simplify]: 0.00097465 [loop_unroll]: 0.00035847 [a_1]: 0.00943662 [with_stream_mark]: 5.336e-05 [recompute_prepare]: 3.653e-05 [updatestate_depend_eliminate]: 2.123e-05 [updatestate_assign_eliminate]: 1.528e-05 [updatestate_loads_eliminate]: 1.912e-05 [parameter_eliminate]: 3.04001e-06 [a_2]: 0.0007624 [accelerated_algorithm]: 6.067e-05 [shard]: 2.51998e-06 [meta_shard_fg_expand]: 2.014e-05 [shard_inline]: 2.187e-05 [merge_send_recv]: 2.026e-05 [auto_parallel]: 1.89e-05 [parallel]: 3.196e-05 [flash_sp]: 1.483e-05 [merge_comm]: 1.365e-05 [allreduce_fusion]: 1.274e-05 [matmul_add_comm_reduction]: 2.26e-05 [allreduce_slice_to_reducescatter]: 1.39998e-06 [virtual_shard_identity]: 2.53e-05 [virtual_dataset]: 2.252e-05 [get_grad_eliminate_]: 2.131e-05 [virtual_output]: 0.00014067 [merge_forward]: 1.502e-05 [cell_reuse_recompute_pass]: 2.02001e-06 [offload_activation]: 2.674e-05 [cell_reuse_handle_not_recompute_node_pass]: 4.918e-05 [merge_recompute_call_nodes]: 2.42001e-06 [before_grad]: 4.255e-05 [set_forward_comm_id_for_comm_node_pass]: 1.582e-05 [meta_fg_expand]: 1.872e-05 [flash_sp_send_recv_attached]: 1.045e-05 [receive_attached]: 2.31998e-06 [after_resolve]: 2.97e-05 [a_after_grad]: 3.703e-05 [renormalize]: 0.0119453 [add_forward_monad_depend]: 1.945e-05 [auto_monad_grad]: 3.09999e-06 [auto_monad_eliminator]: 7.526e-05 [cse]: 0.0014741 [a_3]: 0.00018119 [Cycle 2]: 0.00208668, [45] [expand_dump_flag]: 2.74999e-06 [switch_simplify]: 2.82e-05 [loop_unroll]: 2.294e-05 [a_1]: 0.00066479 [with_stream_mark]: 2.839e-05 [recompute_prepare]: 2.112e-05 [updatestate_depend_eliminate]: 1.559e-05 [updatestate_assign_eliminate]: 1.134e-05 [updatestate_loads_eliminate]: 1.383e-05 [parameter_eliminate]: 1.96e-06 [a_2]: 0.00030581 [accelerated_algorithm]: 2.971e-05 [shard]: 2.34999e-06 [meta_shard_fg_expand]: 7.45e-06 [shard_inline]: 2.097e-05 [merge_send_recv]: 1.877e-05 [auto_parallel]: 1.941e-05 [parallel]: 1.055e-05 [flash_sp]: 4.53999e-06 [merge_comm]: 1.289e-05 [allreduce_fusion]: 1.241e-05 [matmul_add_comm_reduction]: 2.063e-05 [allreduce_slice_to_reducescatter]: 6.79982e-07 [virtual_shard_identity]: 2.236e-05 [virtual_dataset]: 1.959e-05 [get_grad_eliminate_]: 2.025e-05 [virtual_output]: 2.029e-05 [merge_forward]: 1.253e-05 [cell_reuse_recompute_pass]: 3.53e-06 [offload_activation]: 2.392e-05 [cell_reuse_handle_not_recompute_node_pass]: 4.073e-05 [merge_recompute_call_nodes]: 1.71998e-06 [before_grad]: 3.775e-05 [set_forward_comm_id_for_comm_node_pass]: 1.321e-05 [meta_fg_expand]: 9.42999e-06 [flash_sp_send_recv_attached]: 1.99e-06 [receive_attached]: 2.52001e-06 [after_resolve]: 2.761e-05 [a_after_grad]: 4.748e-05 [renormalize]: 1.19995e-07 [add_forward_monad_depend]: 2.64001e-06 [auto_monad_grad]: 2.93e-06 [auto_monad_eliminator]: 4.914e-05 [cse]: 7.824e-05 [a_3]: 0.00014205 [py_interpret_to_execute_after_opt_a]: 2.973e-05 [slice_cell_reuse_recomputed_activation]: 2.07001e-06 [rewriter_after_opt_a]: 0.00084717 [convert_after_rewriter]: 2.28e-05 [order_py_execute_after_rewriter]: 1.48e-05 [mutable_eliminate]: 0.0007266 [opt_b]: 0.00071346, [1] [Cycle 1]: 0.00070457, [7] [b_1]: 0.00052499 [b_2]: 2.368e-05 [updatestate_depend_eliminate]: 1.596e-05 [updatestate_assign_eliminate]: 1.123e-05 [updatestate_loads_eliminate]: 1.287e-05 [renormalize]: 6.59988e-07 [cse]: 7.409e-05 [optimize_parallel_all_gather_comm]: 3.643e-05 [overlap_param_gather]: 2.04999e-06 [cconv]: 3.027e-05 [loop_unroll]: 0.00055405 [opt_after_cconv]: 0.00029382, [1] [Cycle 1]: 0.00028569, [7] [c_1]: 0.00012862 [parameter_eliminate]: 3.11001e-06 [updatestate_depend_eliminate]: 1.823e-05 [updatestate_assign_eliminate]: 1.114e-05 [updatestate_loads_eliminate]: 1.368e-05 [cse]: 7.129e-05 [renormalize]: 2.19996e-07 [remove_dup_value]: 9.149e-05 [tuple_transform]: 0.00029205, [1] [Cycle 1]: 0.00028612, [4] [d_1]: 0.00023675 [none_parameter_eliminate]: 2.37999e-06 [renormalize]: 2.00002e-07 [switch_simplify]: 2.289e-05 [partial_unused_args_eliminate]: 1.92001e-06 [add_recomputation]: 0.00015054 [cse_after_recomputation]: 6.979e-05, [1] [Cycle 1]: 6.322e-05, [1] [cse]: 5.666e-05 [environ_conv]: 2.118e-05 [swap_dp_allreduce_reducescatter]: 1.868e-05 [bias_add_comm_swap]: 3.46001e-06 [label_micro_interleaved_index]: 5.35001e-06 [label_fine_grained_interleaved_index]: 3.39001e-06 [merge_cast_opt]: 1.89e-06 [slice_recompute_activation]: 2.14999e-06 [micro_interleaved_order_control]: 2.71999e-06 [assign_add_opt]: 1.37999e-06 [ForceFp32Comm]: 9.09989e-07 [remove_cast_before_assign_add]: 1.10999e-06 [full_micro_interleaved_order_control]: 2.48998e-06 [reorder_send_recv_between_fp_bp]: 3.14999e-06 [comm_op_add_attrs]: 1.05999e-06 [add_comm_op_reuse_tag]: 1.08001e-06 [interleave_split_concat_branches]: 1.32999e-06 [interleave_parallel_branches]: 1.05001e-06 [overlap_opt_shard_in_pipeline]: 1.34e-06 [overlap_opt_shard_grad_in_pipeline]: 1.77999e-06 [control_data_broadcast_order]: 3.694e-05 [grouped_pairwise_exchange_alltoall]: 2.02001e-06 [offloading_packed_experts]: 1.137e-05 [overlap_recompute_and_grad_model_parallel]: 1.025e-05 [overlap_grad_matmul_and_grad_allreduce]: 1.44998e-06 [overlap_recompute_allgather_and_fa_grad]: 1.47001e-06 [overlap_recompute_comm]: 2.64999e-06 [overlap_grad_ring_attention]: 9.50001e-06 [overlap_grad_flash_sp]: 5.041e-05 [begin_end_overlap_inline]: 5.90022e-07 [split_matmul_comm_elemetwise]: 2.46998e-06 [split_layernorm_comm]: 1.72001e-06 [handle_group_info]: 1.09003e-06 [symbol_engine_optimizer]: 0.0004751, [1] [Cycle 1]: 0.00046969, [6] [build]: 0.00026276 [elim_shapecalc]: 3.119e-05 [elim_not_effective]: 4.994e-05 [opt_reshape]: 3.201e-05 [fold_const_symbol]: 5.388e-05 [renormalize]: 2.50002e-07 [detach_backward]: 2.79999e-06 [pipeline_parallel_scheduler]: 1.39e-06 [auto_monad_reorder]: 5.891e-05 [get_jit_bprop_graph]: 1.90001e-06 [rewriter_after_jit_bprop_graph]: 4.45999e-06 [opt_after_jit_grad]: 0.00058815 [validate]: 9.77e-05 Sums bootstrap : 0.000681s : 0.21% type_inference : 0.288135s : 88.74% event_method : 0.000231s : 0.07% auto_monad : 0.000856s : 0.26% graph_reusing : 0.000054s : 0.02% inline : 0.000003s : 0.00% add_attr.add_attr_with_inline.tag_attr : 0.000210s : 0.06% add_attr.add_attr_with_inline.meta_addattr_fg_expand : 0.000058s : 0.02% parallel-infer-symbol : 0.000004s : 0.00% pre_auto_parallel : 0.000254s : 0.08% insert-virtual-dataset : 0.000003s : 0.00% parallel-infer-symbol-second : 0.000001s : 0.00% dataset_repeat_opt : 0.000002s : 0.00% pipeline_split : 0.000002s : 0.00% optimize.py_interpret_to_execute : 0.000277s : 0.09% optimize.rewriter_before_opt_a : 0.000811s : 0.25% optimize.opt_a.expand_dump_flag : 0.000033s : 0.01% optimize.opt_a.switch_simplify : 0.001003s : 0.31% optimize.opt_a.loop_unroll : 0.000381s : 0.12% optimize.opt_a.a_1 : 0.010101s : 3.11% optimize.opt_a.with_stream_mark : 0.000082s : 0.03% optimize.opt_a.recompute_prepare : 0.000058s : 0.02% optimize.opt_a.updatestate_depend_eliminate : 0.000037s : 0.01% optimize.opt_a.updatestate_assign_eliminate : 0.000027s : 0.01% optimize.opt_a.updatestate_loads_eliminate : 0.000033s : 0.01% optimize.opt_a.parameter_eliminate : 0.000005s : 0.00% optimize.opt_a.a_2 : 0.001068s : 0.33% optimize.opt_a.accelerated_algorithm : 0.000090s : 0.03% optimize.opt_a.shard : 0.000005s : 0.00% optimize.opt_a.meta_shard_fg_expand : 0.000028s : 0.01% optimize.opt_a.shard_inline : 0.000043s : 0.01% optimize.opt_a.merge_send_recv : 0.000039s : 0.01% optimize.opt_a.auto_parallel : 0.000038s : 0.01% optimize.opt_a.parallel : 0.000043s : 0.01% optimize.opt_a.flash_sp : 0.000019s : 0.01% optimize.opt_a.merge_comm : 0.000027s : 0.01% optimize.opt_a.allreduce_fusion : 0.000025s : 0.01% optimize.opt_a.matmul_add_comm_reduction : 0.000043s : 0.01% optimize.opt_a.allreduce_slice_to_reducescatter : 0.000002s : 0.00% optimize.opt_a.virtual_shard_identity : 0.000048s : 0.01% optimize.opt_a.virtual_dataset : 0.000042s : 0.01% optimize.opt_a.get_grad_eliminate_ : 0.000042s : 0.01% optimize.opt_a.virtual_output : 0.000161s : 0.05% optimize.opt_a.merge_forward : 0.000028s : 0.01% optimize.opt_a.cell_reuse_recompute_pass : 0.000006s : 0.00% optimize.opt_a.offload_activation : 0.000051s : 0.02% optimize.opt_a.cell_reuse_handle_not_recompute_node_pass : 0.000090s : 0.03% optimize.opt_a.merge_recompute_call_nodes : 0.000004s : 0.00% optimize.opt_a.before_grad : 0.000080s : 0.02% optimize.opt_a.set_forward_comm_id_for_comm_node_pass : 0.000029s : 0.01% optimize.opt_a.meta_fg_expand : 0.000028s : 0.01% optimize.opt_a.flash_sp_send_recv_attached : 0.000012s : 0.00% optimize.opt_a.receive_attached : 0.000005s : 0.00% optimize.opt_a.after_resolve : 0.000057s : 0.02% optimize.opt_a.a_after_grad : 0.000085s : 0.03% optimize.opt_a.renormalize : 0.011945s : 3.68% optimize.opt_a.add_forward_monad_depend : 0.000022s : 0.01% optimize.opt_a.auto_monad_grad : 0.000006s : 0.00% optimize.opt_a.auto_monad_eliminator : 0.000124s : 0.04% optimize.opt_a.cse : 0.001552s : 0.48% optimize.opt_a.a_3 : 0.000323s : 0.10% optimize.py_interpret_to_execute_after_opt_a : 0.000030s : 0.01% optimize.slice_cell_reuse_recomputed_activation : 0.000002s : 0.00% optimize.rewriter_after_opt_a : 0.000847s : 0.26% optimize.convert_after_rewriter : 0.000023s : 0.01% optimize.order_py_execute_after_rewriter : 0.000015s : 0.00% optimize.mutable_eliminate : 0.000727s : 0.22% optimize.opt_b.b_1 : 0.000525s : 0.16% optimize.opt_b.b_2 : 0.000024s : 0.01% optimize.opt_b.updatestate_depend_eliminate : 0.000016s : 0.00% optimize.opt_b.updatestate_assign_eliminate : 0.000011s : 0.00% optimize.opt_b.updatestate_loads_eliminate : 0.000013s : 0.00% optimize.opt_b.renormalize : 0.000001s : 0.00% optimize.opt_b.cse : 0.000074s : 0.02% optimize.optimize_parallel_all_gather_comm : 0.000036s : 0.01% optimize.overlap_param_gather : 0.000002s : 0.00% optimize.cconv : 0.000030s : 0.01% optimize.loop_unroll : 0.000554s : 0.17% optimize.opt_after_cconv.c_1 : 0.000129s : 0.04% optimize.opt_after_cconv.parameter_eliminate : 0.000003s : 0.00% optimize.opt_after_cconv.updatestate_depend_eliminate : 0.000018s : 0.01% optimize.opt_after_cconv.updatestate_assign_eliminate : 0.000011s : 0.00% optimize.opt_after_cconv.updatestate_loads_eliminate : 0.000014s : 0.00% optimize.opt_after_cconv.cse : 0.000071s : 0.02% optimize.opt_after_cconv.renormalize : 0.000000s : 0.00% optimize.remove_dup_value : 0.000091s : 0.03% optimize.tuple_transform.d_1 : 0.000237s : 0.07% optimize.tuple_transform.none_parameter_eliminate : 0.000002s : 0.00% optimize.tuple_transform.renormalize : 0.000000s : 0.00% optimize.tuple_transform.switch_simplify : 0.000023s : 0.01% optimize.partial_unused_args_eliminate : 0.000002s : 0.00% optimize.add_recomputation : 0.000151s : 0.05% optimize.cse_after_recomputation.cse : 0.000057s : 0.02% optimize.environ_conv : 0.000021s : 0.01% optimize.swap_dp_allreduce_reducescatter : 0.000019s : 0.01% optimize.bias_add_comm_swap : 0.000003s : 0.00% optimize.label_micro_interleaved_index : 0.000005s : 0.00% optimize.label_fine_grained_interleaved_index : 0.000003s : 0.00% optimize.merge_cast_opt : 0.000002s : 0.00% optimize.slice_recompute_activation : 0.000002s : 0.00% optimize.micro_interleaved_order_control : 0.000003s : 0.00% optimize.assign_add_opt : 0.000001s : 0.00% optimize.ForceFp32Comm : 0.000001s : 0.00% optimize.remove_cast_before_assign_add : 0.000001s : 0.00% optimize.full_micro_interleaved_order_control : 0.000002s : 0.00% optimize.reorder_send_recv_between_fp_bp : 0.000003s : 0.00% optimize.comm_op_add_attrs : 0.000001s : 0.00% optimize.add_comm_op_reuse_tag : 0.000001s : 0.00% optimize.interleave_split_concat_branches : 0.000001s : 0.00% optimize.interleave_parallel_branches : 0.000001s : 0.00% optimize.overlap_opt_shard_in_pipeline : 0.000001s : 0.00% optimize.overlap_opt_shard_grad_in_pipeline : 0.000002s : 0.00% optimize.control_data_broadcast_order : 0.000037s : 0.01% optimize.grouped_pairwise_exchange_alltoall : 0.000002s : 0.00% optimize.offloading_packed_experts : 0.000011s : 0.00% optimize.overlap_recompute_and_grad_model_parallel : 0.000010s : 0.00% optimize.overlap_grad_matmul_and_grad_allreduce : 0.000001s : 0.00% optimize.overlap_recompute_allgather_and_fa_grad : 0.000001s : 0.00% optimize.overlap_recompute_comm : 0.000003s : 0.00% optimize.overlap_grad_ring_attention : 0.000010s : 0.00% optimize.overlap_grad_flash_sp : 0.000050s : 0.02% optimize.begin_end_overlap_inline : 0.000001s : 0.00% optimize.split_matmul_comm_elemetwise : 0.000002s : 0.00% optimize.split_layernorm_comm : 0.000002s : 0.00% optimize.handle_group_info : 0.000001s : 0.00% optimize.symbol_engine_optimizer.build : 0.000263s : 0.08% optimize.symbol_engine_optimizer.elim_shapecalc : 0.000031s : 0.01% optimize.symbol_engine_optimizer.elim_not_effective : 0.000050s : 0.02% optimize.symbol_engine_optimizer.opt_reshape : 0.000032s : 0.01% optimize.symbol_engine_optimizer.fold_const_symbol : 0.000054s : 0.02% optimize.symbol_engine_optimizer.renormalize : 0.000000s : 0.00% detach_backward : 0.000003s : 0.00% pipeline_parallel_scheduler : 0.000001s : 0.00% auto_monad_reorder : 0.000059s : 0.02% get_jit_bprop_graph : 0.000002s : 0.00% rewriter_after_jit_bprop_graph : 0.000004s : 0.00% opt_after_jit_grad : 0.000588s : 0.18% validate : 0.000098s : 0.03% Time group info: ------[substitution.] 0.003553 569 0.47% : 0.000017s : 8: substitution.depend_value_elim 0.35% : 0.000013s : 14: substitution.elim_not_effective 0.49% : 0.000017s : 12: substitution.float_tuple_getitem_switch 0.64% : 0.000023s : 14: substitution.fold_const_symbol 0.46% : 0.000016s : 17: substitution.graph_param_transform 70.36% : 0.002500s : 71: substitution.inline 0.37% : 0.000013s : 29: substitution.j_node_and_user_rematch 0.97% : 0.000035s : 6: substitution.less_batch_normalization 1.73% : 0.000062s : 2: substitution.list_to_tuple_eliminator_ 0.17% : 0.000006s : 6: substitution.load_eliminater 0.81% : 0.000029s : 21: substitution.minmaximum_grad 0.22% : 0.000008s : 2: substitution.opt_reshape 0.56% : 0.000020s : 29: substitution.remove_not_recompute_node 0.22% : 0.000008s : 4: substitution.replace_old_param 1.80% : 0.000064s : 18: substitution.reshape_eliminate 1.53% : 0.000055s : 24: substitution.switch_simplify 4.20% : 0.000149s : 51: substitution.tuple_list_convert_item_index_to_positive 2.34% : 0.000083s : 51: substitution.tuple_list_get_item_const_eliminator 2.92% : 0.000104s : 51: substitution.tuple_list_get_item_depend_reorder 5.39% : 0.000191s : 73: substitution.tuple_list_get_item_eliminator 2.89% : 0.000103s : 51: substitution.tuple_list_get_set_item_eliminator 0.55% : 0.000019s : 6: substitution.updatestate_pure_node_eliminater 0.55% : 0.000020s : 9: substitution.updatestate_useless_node_eliminater ------[type_inference.] 0.287726 2 93.15% : 0.268022s : 1: type_inference.infer 6.85% : 0.019704s : 1: type_inference.specialize ------[replace.] 0.001145 115 3.23% : 0.000037s : 5: replace.depend_value_elim 55.85% : 0.000639s : 71: replace.inline 4.05% : 0.000046s : 2: replace.list_to_tuple_eliminator_ 28.08% : 0.000321s : 24: replace.switch_simplify 8.80% : 0.000101s : 13: replace.tuple_list_get_item_eliminator ------[match.] 0.002585 115 0.08% : 0.000002s : 5: match.depend_value_elim 94.98% : 0.002455s : 71: match.inline 2.25% : 0.000058s : 2: match.list_to_tuple_eliminator_ 1.54% : 0.000040s : 24: match.switch_simplify 1.15% : 0.000030s : 13: match.tuple_list_get_item_eliminator ------[predicate.] 0.001722 9701 1.06% : 0.000018s : 123: predicate.accumulaten_eliminater 0.34% : 0.000006s : 17: predicate.ad_related_special_op_eliminate 0.46% : 0.000008s : 58: predicate.addn_check_dump 1.13% : 0.000020s : 123: predicate.addn_zero_filter 1.05% : 0.000018s : 123: predicate.adjust_all_reduce_mul_add 2.08% : 0.000036s : 181: predicate.arithmetic_simplify 1.12% : 0.000019s : 123: predicate.cast_eliminate 0.30% : 0.000005s : 34: predicate.check_bprop_eliminate 0.45% : 0.000008s : 58: predicate.compare_switch_simplify 0.08% : 0.000001s : 17: predicate.const_output_eliminate 0.58% : 0.000010s : 61: predicate.depend_value_elim 1.18% : 0.000020s : 123: predicate.dict_get_item_const_eliminator 1.32% : 0.000023s : 123: predicate.dict_get_item_eliminator 1.05% : 0.000018s : 123: predicate.dict_set_item_eliminator 0.33% : 0.000006s : 34: predicate.dumpgradient_eliminate 0.08% : 0.000001s : 17: predicate.elim_not_effective 0.17% : 0.000003s : 17: predicate.elim_shapecalc_of_broadcastargs 1.20% : 0.000021s : 140: predicate.environ_add_const_eliminate 1.14% : 0.000020s : 140: predicate.environ_get_add_eliminate 1.24% : 0.000021s : 140: predicate.environ_get_depend_swap 1.65% : 0.000028s : 198: predicate.environ_get_eliminate 1.23% : 0.000021s : 140: predicate.environ_get_set_eliminate 1.78% : 0.000031s : 209: predicate.exchange_switch_depend_value 2.47% : 0.000043s : 209: predicate.float_depend_g_call 0.47% : 0.000008s : 58: predicate.float_environ_get_switch 0.71% : 0.000012s : 75: predicate.float_tuple_getitem_switch 0.07% : 0.000001s : 17: predicate.fold_const_symbol 0.31% : 0.000005s : 35: predicate.get_grad_eliminate 0.09% : 0.000002s : 17: predicate.graph_param_transform 0.47% : 0.000008s : 58: predicate.incorporate_call 0.43% : 0.000007s : 58: predicate.incorporate_call_switch 5.30% : 0.000091s : 459: predicate.inline 0.41% : 0.000007s : 35: predicate.inline_without_move 0.14% : 0.000002s : 35: predicate.j_node_and_user_rematch 0.43% : 0.000007s : 36: predicate.less_batch_normalization 1.61% : 0.000028s : 172: predicate.list_to_tuple_eliminator_ 2.61% : 0.000045s : 295: predicate.load_eliminater 0.36% : 0.000006s : 17: predicate.loop_unroll_after_grad 3.53% : 0.000061s : 314: predicate.loop_unroll_before_grad 1.46% : 0.000025s : 157: predicate.make_slice_get_slice_eliminator 0.47% : 0.000008s : 58: predicate.merge_addn 0.31% : 0.000005s : 34: predicate.micro_step_allgather_replace 0.28% : 0.000005s : 34: predicate.mini_step_allgather_replace 1.05% : 0.000018s : 123: predicate.minmaximum_grad 0.33% : 0.000006s : 17: predicate.mutable_eliminate 0.16% : 0.000003s : 17: predicate.opt_reshape 0.17% : 0.000003s : 17: predicate.parallel_virtual_node 3.23% : 0.000056s : 209: predicate.partial_defer_inline 1.42% : 0.000024s : 155: predicate.partial_eliminate 1.13% : 0.000019s : 123: predicate.print_const_string_wrapper 0.46% : 0.000008s : 53: predicate.reduce_all_const_elim 1.44% : 0.000025s : 123: predicate.reduce_eliminate 2.50% : 0.000043s : 295: predicate.redundant_stop_gradient_eliminater 0.15% : 0.000003s : 35: predicate.remove_not_recompute_node 1.06% : 0.000018s : 172: predicate.replace_applicator 0.16% : 0.000003s : 35: predicate.replace_old_param 0.09% : 0.000002s : 17: predicate.reset_defer_inline 1.17% : 0.000020s : 123: predicate.reshape_eliminate 0.30% : 0.000005s : 34: predicate.row_tensor_add_zeros_like 0.19% : 0.000003s : 17: predicate.row_tensor_eliminate 0.46% : 0.000008s : 34: predicate.same_eliminate 0.18% : 0.000003s : 40: predicate.set_cell_output_no_recompute 0.34% : 0.000006s : 35: predicate.shard_identity_eliminate 0.32% : 0.000006s : 34: predicate.special_op_eliminate 0.53% : 0.000009s : 58: predicate.specialize_transform 0.38% : 0.000007s : 34: predicate.split_environ_get_set_with_tuple_value 0.36% : 0.000006s : 35: predicate.stack_unstack_eliminate 0.17% : 0.000003s : 17: predicate.switch_call_monad_eliminater 2.05% : 0.000035s : 209: predicate.switch_defer_inline 2.28% : 0.000039s : 243: predicate.switch_layer_defer_inline 6.84% : 0.000118s : 646: predicate.switch_simplify 1.17% : 0.000020s : 123: predicate.tile_eliminate 1.07% : 0.000018s : 123: predicate.transpose_eliminate 1.62% : 0.000028s : 157: predicate.tuple_list_convert_item_index_to_positive 1.59% : 0.000027s : 157: predicate.tuple_list_get_item_const_eliminator 1.47% : 0.000025s : 157: predicate.tuple_list_get_item_depend_reorder 2.81% : 0.000048s : 228: predicate.tuple_list_get_item_eliminator 1.55% : 0.000027s : 157: predicate.tuple_list_get_set_item_eliminator 2.15% : 0.000037s : 215: predicate.tuple_list_set_item_eliminator 1.50% : 0.000026s : 170: predicate.tuple_to_list_eliminator_ 2.36% : 0.000041s : 295: predicate.updatestate_pure_node_eliminater 3.00% : 0.000052s : 353: predicate.updatestate_useless_node_eliminater 0.17% : 0.000003s : 17: predicate.value_based_eliminate 0.31% : 0.000005s : 35: predicate.virtual_dataset_eliminate 6.97% : 0.000120s : 35: predicate.virtual_output_eliminate 0.14% : 0.000002s : 17: predicate.virtual_view_grad_eliminate 0.19% : 0.000003s : 17: predicate.zero_like_fill_zero ------[func_graph_cloner_run.] 0.015838 135 63.72% : 0.010092s : 55: func_graph_cloner_run.FuncGraphClonerGraph 36.28% : 0.005746s : 80: func_graph_cloner_run.FuncGraphSpecializer ------[meta_graph.] 0.000000 0 ------[manager.] 0.000000 0 ------[pynative] 0.000000 0 ------[others.] 0.397649 209 0.00% : 0.000003s : 1: ForceFp32Comm 1.37% : 0.005430s : 1: add_attr 1.36% : 0.005411s : 1: add_attr_with_inline 0.00% : 0.000005s : 1: add_comm_op_reuse_tag 0.04% : 0.000156s : 1: add_recomputation 0.00% : 0.000005s : 1: assign_add_opt 0.22% : 0.000873s : 1: auto_monad 0.02% : 0.000064s : 1: auto_monad_reorder 0.00% : 0.000004s : 1: begin_end_overlap_inline 0.00% : 0.000006s : 1: bias_add_comm_swap 0.18% : 0.000712s : 1: bootstrap 0.01% : 0.000034s : 1: cconv 0.00% : 0.000004s : 1: comm_op_add_attrs 0.01% : 0.000040s : 1: control_data_broadcast_order 0.01% : 0.000028s : 1: convert_after_rewriter 0.02% : 0.000073s : 1: cse_after_recomputation 0.00% : 0.000005s : 1: dataset_repeat_opt 0.00% : 0.000006s : 1: detach_backward 0.01% : 0.000025s : 1: environ_conv 0.06% : 0.000244s : 1: event_method 0.00% : 0.000007s : 1: full_micro_interleaved_order_control 0.00% : 0.000006s : 1: get_jit_bprop_graph 0.02% : 0.000061s : 1: graph_reusing 0.00% : 0.000005s : 1: grouped_pairwise_exchange_alltoall 0.00% : 0.000005s : 1: handle_group_info 0.00% : 0.000007s : 1: inline 0.00% : 0.000006s : 1: insert-virtual-dataset 0.00% : 0.000004s : 1: interleave_parallel_branches 0.00% : 0.000004s : 1: interleave_split_concat_branches 0.00% : 0.000006s : 1: label_fine_grained_interleaved_index 0.00% : 0.000008s : 1: label_micro_interleaved_index 0.14% : 0.000565s : 1: loop_unroll 0.00% : 0.000006s : 1: merge_cast_opt 0.00% : 0.000006s : 1: micro_interleaved_order_control 0.19% : 0.000737s : 1: mutable_eliminate 0.00% : 0.000014s : 1: offloading_packed_experts 0.01% : 0.000038s : 1: opt.transform.loop_unroll_optimizer 0.01% : 0.000038s : 1: opt.transform.mutable_eliminate 3.40% : 0.013537s : 95: opt.transform.opt_a 0.03% : 0.000127s : 1: opt.transform.opt_after_cconv 0.02% : 0.000073s : 1: opt.transform.opt_after_jit_grad 0.13% : 0.000514s : 28: opt.transform.opt_b 0.06% : 0.000255s : 2: opt.transform.opt_trans_graph 0.04% : 0.000162s : 4: opt.transform.symbol_engine_opt 7.20% : 0.028640s : 1: opt_a 0.07% : 0.000298s : 1: opt_after_cconv 0.15% : 0.000599s : 1: opt_after_jit_grad 0.18% : 0.000717s : 1: opt_b 8.70% : 0.034587s : 1: optimize 0.01% : 0.000041s : 1: optimize_parallel_all_gather_comm 0.00% : 0.000019s : 1: order_py_execute_after_rewriter 0.01% : 0.000054s : 1: overlap_grad_flash_sp 0.00% : 0.000004s : 1: overlap_grad_matmul_and_grad_allreduce 0.00% : 0.000012s : 1: overlap_grad_ring_attention 0.00% : 0.000006s : 1: overlap_opt_shard_grad_in_pipeline 0.00% : 0.000004s : 1: overlap_opt_shard_in_pipeline 0.00% : 0.000006s : 1: overlap_param_gather 0.00% : 0.000004s : 1: overlap_recompute_allgather_and_fa_grad 0.00% : 0.000013s : 1: overlap_recompute_and_grad_model_parallel 0.00% : 0.000007s : 1: overlap_recompute_comm 0.00% : 0.000009s : 1: parallel-infer-symbol 0.00% : 0.000004s : 1: parallel-infer-symbol-second 0.00% : 0.000005s : 1: partial_unused_args_eliminate 0.00% : 0.000004s : 1: pipeline_parallel_scheduler 0.00% : 0.000004s : 1: pipeline_split 0.07% : 0.000263s : 1: pre_auto_parallel 0.07% : 0.000289s : 1: py_interpret_to_execute 0.01% : 0.000034s : 1: py_interpret_to_execute_after_opt_a 0.00% : 0.000004s : 1: remove_cast_before_assign_add 0.02% : 0.000097s : 1: remove_dup_value 1.65% : 0.006560s : 1: renormalize.infer 1.35% : 0.005369s : 1: renormalize.specialize 0.00% : 0.000007s : 1: reorder_send_recv_between_fp_bp 0.00% : 0.000008s : 1: rewriter_after_jit_bprop_graph 0.22% : 0.000855s : 1: rewriter_after_opt_a 0.21% : 0.000823s : 1: rewriter_before_opt_a 0.00% : 0.000005s : 1: slice_cell_reuse_recomputed_activation 0.00% : 0.000005s : 1: slice_recompute_activation 0.00% : 0.000004s : 1: split_layernorm_comm 0.00% : 0.000005s : 1: split_matmul_comm_elemetwise 0.01% : 0.000022s : 1: swap_dp_allreduce_reducescatter 0.12% : 0.000479s : 1: symbol_engine_optimizer 0.07% : 0.000297s : 1: tuple_transform 72.47% : 0.288163s : 1: type_inference . [hook] pytest_runtest_teardown:test_matmul_qkv_bf16_1[True-mstype0-k_n_shape3-1024] tests/st/infer/ops/test_internal_ops/test_matmul_split.py::test_matmul_qkv_bf16_1[True-mstype0-k_n_shape3-1024],max_mem:252.0M =============================== warnings summary =============================== ../../../../../../../../../../usr/local/Ascend/cann-8.5.0/python/site-packages/tbe/dsl/classifier/transdata/transdata_classifier.py:222 /usr/local/Ascend/cann-8.5.0/python/site-packages/tbe/dsl/classifier/transdata/transdata_classifier.py:222: DeprecationWarning: invalid escape sequence \B """ ../../../../../../../../../../usr/local/Ascend/cann-8.5.0/python/site-packages/tbe/dsl/unify_schedule/vector/transdata/common/graph/transdata_graph_info.py:143 /usr/local/Ascend/cann-8.5.0/python/site-packages/tbe/dsl/unify_schedule/vector/transdata/common/graph/transdata_graph_info.py:143: DeprecationWarning: invalid escape sequence \c """ ../../../../../../../../../../usr/local/Ascend/cann-8.5.0/python/site-packages/tbe/dsl/unify_schedule/vector/transdata/common/graph/transdata_graph_info.py:170 /usr/local/Ascend/cann-8.5.0/python/site-packages/tbe/dsl/unify_schedule/vector/transdata/common/graph/transdata_graph_info.py:170: DeprecationWarning: invalid escape sequence \c """ ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/numpy/core/getlimits.py:549 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/numpy/core/getlimits.py:549: UserWarning: The value of the smallest subnormal for type is zero. setattr(self, word, getattr(machar, word).flat[0]) ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/numpy/core/getlimits.py:89 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/numpy/core/getlimits.py:89: UserWarning: The value of the smallest subnormal for type is zero. return self._float_to_str(self.smallest_subnormal) ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/numpy/core/getlimits.py:549 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/numpy/core/getlimits.py:549: UserWarning: The value of the smallest subnormal for type is zero. setattr(self, word, getattr(machar, word).flat[0]) ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/numpy/core/getlimits.py:89 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/numpy/core/getlimits.py:89: UserWarning: The value of the smallest subnormal for type is zero. return self._float_to_str(self.smallest_subnormal) ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/batchnorm_fold2.py:57 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/batchnorm_fold2.py:57: DeprecationWarning: te_fusion.fusion_manager.fusion_manager.register is deprecated,please replace it with tbe.common.register.register_op_compute @fusion_manager.register("batchnorm_fold2") ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/batchnorm_fold2_grad.py:56 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/batchnorm_fold2_grad.py:56: DeprecationWarning: te_fusion.fusion_manager.fusion_manager.register is deprecated,please replace it with tbe.common.register.register_op_compute @fusion_manager.register("batchnorm_fold2_grad") ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/batchnorm_fold2_grad_reduce.py:48 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/batchnorm_fold2_grad_reduce.py:48: DeprecationWarning: te_fusion.fusion_manager.fusion_manager.register is deprecated,please replace it with tbe.common.register.register_op_compute @fusion_manager.register("batchnorm_fold2_grad_reduce") ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/correction_mul.py:51 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/correction_mul.py:51: DeprecationWarning: te_fusion.fusion_manager.fusion_manager.register is deprecated,please replace it with tbe.common.register.register_op_compute @fusion_manager.register("correction_mul") ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/correction_mul_grad.py:51 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/correction_mul_grad.py:51: DeprecationWarning: te_fusion.fusion_manager.fusion_manager.register is deprecated,please replace it with tbe.common.register.register_op_compute @fusion_manager.register("correction_mul_grad") ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/correction_mul_grad.py:143 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/correction_mul_grad.py:143: DeprecationWarning: te_fusion.fusion_manager.fusion_manager.register is deprecated,please replace it with tbe.common.register.register_op_compute @fusion_manager.register("correction_mul_grad_reduce") ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/fake_learned_scale_quant_perlayer.py:50 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/fake_learned_scale_quant_perlayer.py:50: DeprecationWarning: te_fusion.fusion_manager.fusion_manager.register is deprecated,please replace it with tbe.common.register.register_op_compute @fusion_manager.register("fake_learned_scale_quant_perlayer") ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/fake_learned_scale_quant_perlayer_grad.py:92 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/fake_learned_scale_quant_perlayer_grad.py:92: DeprecationWarning: te_fusion.fusion_manager.fusion_manager.register is deprecated,please replace it with tbe.common.register.register_op_compute @fusion_manager.register("fake_learned_scale_quant_perlayer_grad_d") ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/fake_learned_scale_quant_perlayer_grad_reduce.py:49 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/fake_learned_scale_quant_perlayer_grad_reduce.py:49: DeprecationWarning: te_fusion.fusion_manager.fusion_manager.register is deprecated,please replace it with tbe.common.register.register_op_compute @fusion_manager.register("fake_learned_scale_quant_perlayer_grad_d_reduce") ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/fake_learned_scale_quant_perchannel.py:50 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/fake_learned_scale_quant_perchannel.py:50: DeprecationWarning: te_fusion.fusion_manager.fusion_manager.register is deprecated,please replace it with tbe.common.register.register_op_compute @fusion_manager.register("fake_learned_scale_quant_perchannel") ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/fake_learned_scale_quant_perchannel_grad.py:91 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/fake_learned_scale_quant_perchannel_grad.py:91: DeprecationWarning: te_fusion.fusion_manager.fusion_manager.register is deprecated,please replace it with tbe.common.register.register_op_compute @fusion_manager.register("fake_learned_scale_quant_perchannel_grad_d") ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/fake_learned_scale_quant_perchannel_grad_reduce.py:48 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/fake_learned_scale_quant_perchannel_grad_reduce.py:48: DeprecationWarning: te_fusion.fusion_manager.fusion_manager.register is deprecated,please replace it with tbe.common.register.register_op_compute @fusion_manager.register("fake_learned_scale_quant_perchannel_grad_d_reduce") ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/fake_quant_perchannel.py:52 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/fake_quant_perchannel.py:52: DeprecationWarning: te_fusion.fusion_manager.fusion_manager.register is deprecated,please replace it with tbe.common.register.register_op_compute @fusion_manager.register("fake_quant_perchannel") ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/fake_quant_perchannel_grad.py:81 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/fake_quant_perchannel_grad.py:81: DeprecationWarning: te_fusion.fusion_manager.fusion_manager.register is deprecated,please replace it with tbe.common.register.register_op_compute @fusion_manager.register("fake_quant_perchannel_grad") ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/fake_quant_perlayer.py:54 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/fake_quant_perlayer.py:54: DeprecationWarning: te_fusion.fusion_manager.fusion_manager.register is deprecated,please replace it with tbe.common.register.register_op_compute @fusion_manager.register("fake_quant_per_layer") ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/fake_quant_perlayer_grad.py:81 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/fake_quant_perlayer_grad.py:81: DeprecationWarning: te_fusion.fusion_manager.fusion_manager.register is deprecated,please replace it with tbe.common.register.register_op_compute @fusion_manager.register("fake_quant_per_layer_grad") ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/minmax_update_perchannel.py:50 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/minmax_update_perchannel.py:50: DeprecationWarning: te_fusion.fusion_manager.fusion_manager.register is deprecated,please replace it with tbe.common.register.register_op_compute @fusion_manager.register("minmax_update_perchannel") ../../../../../../../../anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/minmax_update_perlayer.py:50 /home/jenkins/anaconda3/envs/ci39/lib/python3.9/site-packages/mindspore/ops/_op_impl/_custom_op/minmax_update_perlayer.py:50: DeprecationWarning: te_fusion.fusion_manager.fusion_manager.register is deprecated,please replace it with tbe.common.register.register_op_compute @fusion_manager.register("minmax_update_perlayer") test_matmul_split.py::test_matmul_qkv_bf16_1[False-mstype0-k_n_shape3-1024] /home/jenkins/mindspore/testcases/testcases/tests/st/infer/ops/test_internal_ops/test_matmul_split.py:84: RuntimeWarning: divide by zero encountered in divide err_cnt = np.sum(np.abs(out_flatten - expect_flatten) / -- Docs: https://docs.pytest.org/en/stable/warnings.html ================= 16 passed, 26 warnings in 539.20s (0:08:59) ==================