Inicio Explorar Ayuda
Registro Iniciar sesión
david
/
flash-attention
espejo de https://github.com/Dao-AILab/flash-attention
1
0
Fork 0
Archivos Incidencias 0 Wiki
Árbol: 6b1d059eda
Ramas Etiquetas
flash-attention
/
csrc
/
xentropy
sclarkson 1feb711f46 Fix compilation with clang on ARM64 (#1285) hace 2 meses
..
README.md 5400fdc4ac [CE] Implement CrossEntropyLoss in Triton hace 1 año
interface.cpp dff68c2b22 Add smoothing for CrossEntropyParallel, rename to CrossEntropyLoss hace 2 años
setup.py 50896ec574 Make nvcc threads configurable via environment variable (#885) hace 11 meses
xentropy_kernel.cu 1feb711f46 Fix compilation with clang on ARM64 (#1285) hace 2 meses

README.md

This CUDA extension implements optimized cross-entropy loss, adapted from Apex's Xentropy. We make it work for bfloat16 and support in-place backward to save memory.

It has only been tested on A100s.

cd csrc/xentropy && pip install .

As of 2023-09-15, this extension is no longer used in the FlashAttention repo. We've instead switched to a Triton-based implementation. See the CrossEntropyLoss module for more details.