DeepSeek announces the open source of DeepGEMM

ChainCatcher news, Deepseek announced the open-source of DeepGEMM on the third day of Open Source Week. DeepGEMM is a library designed for concise and efficient FP8 general matrix multiplication (GEMM), featuring fine-grained scaling capabilities as proposed in DeepSeek-V3. It supports GEMM for both standard and mixture of experts (MoE) groups.

The library is written in CUDA and does not require compilation during installation, compiling all kernels at runtime using a lightweight just-in-time (JIT) module. The FP8GEMM library supports V3/R1 training and inference.