Chirag Jain
50896ec574
Make nvcc threads configurable via environment variable (#885)
|
пре 9 месеци | |
|---|---|---|
| .. | ||
| README.md | пре 2 година | |
| fused_dense.cpp | пре 1 година | |
| fused_dense_cuda.cu | пре 1 година | |
| setup.py | пре 9 месеци | |
README.md
This CUDA extension implements fused matmul + bias (forward and backward), and fused matmul + bias + gelu (forward and backward), adapted from Apex's FusedDense. We make it work for bfloat16.
For best performance, you should use CUDA >= 11.8. CuBLAS versions before this doesn't have the best matmul + bias + gelu performance for bfloat16.
It has only been tested on A100s.
cd csrc/fused_dense_lib && pip install .