biondizzle/DeepGEMM
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Make various updates and fixes: (#164)

Browse Source 
- Add BF16 support for SM90 and SM100
- Refactor Python APIs
- Other fixes and code refactoring
This commit is contained in:
Ray Wang
2025-08-15 18:32:35 +08:00
committed by GitHub
parent 3254b758e2
commit f85ec649d7
34 changed files with 2293 additions and 495 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
README.md
View File

@@ -105,7 +105,7 @@ We also provide a K-axis-grouped API for MoE weight backward (with M and N must
During the inference decoding phase, when CUDA graph is enabled and the CPU is unaware of the number of tokens each expert receives, we support masked grouped GEMMs. By providing a mask tensor, the kernel computes only the valid portions.
Use `fp8_m_grouped_gemm_nt_masked` for this purpose and consult the relevant documentation. An example usage is to use the output of low-latency kernels from [DeepEP](https://github.com/deepseek-ai/DeepEP) as input.
Use `m_grouped_fp8_gemm_nt_masked` for this purpose and consult the relevant documentation. An example usage is to use the output of low-latency kernels from [DeepEP](https://github.com/deepseek-ai/DeepEP) as input.
#### Utilities