Chirag Jain
50896ec574
Make nvcc threads configurable via environment variable (#885)
|
před 9 měsíci | |
|---|---|---|
| .. | ||
| README.md | před 2 roky | |
| fused_dense.cpp | před 1 rokem | |
| fused_dense_cuda.cu | před 1 rokem | |
| setup.py | před 9 měsíci | |
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 .