Add ninja to pyproject.toml build-system, bump to v1.0.5

pyproject.toml

@@ -1,3 +1,3 @@
 [build-system]
-requires = ["packaging", "setuptools", "torch", "wheel"]
+requires = ["ninja", "packaging", "setuptools", "torch", "wheel"]
 build-backend = "setuptools.build_meta"

setup.py

@@ -162,7 +162,7 @@ def append_nvcc_threads(nvcc_extra_args):
 
 setup(
     name="flash_attn",
-    version="1.0.4",
+    version="1.0.5",
     packages=find_packages(
         exclude=("build", "csrc", "include", "tests", "dist", "docs", "benchmarks", "flash_attn.egg-info",)
     ),
@@ -184,5 +184,6 @@ def append_nvcc_threads(nvcc_extra_args):
         "torch",
         "einops",
         "packaging",
+        "ninja",
     ],
 )

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/[email protected]
 
 # Install FlashAttention
-RUN pip install flash-attn==1.0.4
+RUN pip install flash-attn==1.0.5
 
 # Install CUDA extensions for cross-entropy, fused dense, layer norm
 RUN git clone https://github.com/HazyResearch/flash-attention \
-    && cd flash-attention && git checkout v1.0.4 \
+    && cd flash-attention && git checkout v1.0.5 \
     && cd csrc/fused_softmax && pip install . && cd ../../ \
     && cd csrc/rotary && pip install . && cd ../../ \
     && cd csrc/xentropy && pip install . && cd ../../ \