Diff Coverage

Diff: origin/master...HEAD, staged and unstaged changes

Source File Diff Coverage (%) Missing Lines
hyper_parallel/components/modules/expert_sort.py 100%  
hyper_parallel/components/modules/moe.py 66.7% 520
hyper_parallel/components/quantization/modules/grouped_experts.py 100%  
hyper_parallel/components/quantization/modules/hifloat8_grouped_linear.py 100%  
hyper_parallel/components/modules/moe.py
516
517
518
519
520
521
522
523
524

    Returns:
        Scalar loss tensor. Returns 0.0 for empty input.
    """
    import torch.distributed as dist  # pylint: disable=C0415,forbidden-backend-import

    num_tokens, top_k = top_scores.shape

    if num_tokens == 0: