Bump version to 1.0.0

README.md

@@ -38,7 +38,7 @@ and experiment with. The notations in the Triton implementation are also closer
 to what's used in our paper.
 
 
-## Beta release (0.2).
+## Installation and features
 
 Requirements:
 - CUDA 11.4 and above.

setup.py

@@ -162,7 +162,7 @@ ext_modules.append(
 
 setup(
     name="flash_attn",
-    version="0.2.8",
+    version="1.0.0",
     packages=find_packages(
         exclude=("build", "csrc", "include", "tests", "dist", "docs", "benchmarks", "flash_attn.egg-info",)
     ),

training/Dockerfile

@@ -85,11 +85,11 @@ RUN pip install transformers==4.25.1 datasets==2.8.0 pytorch-lightning==1.8.6 tr
 RUN pip install git+https://github.com/mlcommons/logging.git@2.1.0
 
 # Install FlashAttention
-RUN pip install flash-attn==0.2.8
+RUN pip install flash-attn==1.0.0
 
 # Install CUDA extensions for cross-entropy, fused dense, layer norm
 RUN git clone https://github.com/HazyResearch/flash-attention \
-    && cd flash-attention && git checkout v0.2.8 \
+    && cd flash-attention && git checkout v1.0.0 \
     && cd csrc/fused_softmax && pip install . && cd ../../ \
     && cd csrc/rotary && pip install . && cd ../../ \
     && cd csrc/xentropy && pip install . && cd ../../ \