DeepSeek/DeepGEMM
1
0
Fork 0
mirror of https://github.com/deepseek-ai/DeepGEMM synced 2025-04-06 22:14:28 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge pull request #63 from fzyzcjy/patch-2

Browse Source 
Super tiny fix typo
This commit is contained in:
Liang 2025-03-14 10:27:48 +08:00 committed by GitHub
commit 4377c4dc57
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
deep_gemm/jit_kernels/m_grouped_gemm.py
View File

@ -51,7 +51,7 @@ def m_grouped_gemm_fp8_fp8_bf16_nt_contiguous(lhs: Tuple[torch.Tensor, torch.Ten
the second element is an FP32 128x128 scaling tensor for RHS of shape `[num_groups, ⌈n / 128⌉, ⌈k / 128⌉]`.
out: the BF16 output tensor of shape `[m_sum, n]`, representing the result.
m_indices: a tensor of shape `[m_sum]` with type `torch.int`.
`m_indices[i]` records the group which the j-th row of the LHS belong to,
`m_indices[i]` records the group which the i-th row of the LHS belong to,
which means that the i-th row of the LHS matrix will be multiplied with `rhs[m_indices[i]]`.
Values of `m_indices` in every-m-alignment-block must also be the same.
"""