Drop Pytorch 2.1

Author: tridao

.github/workflows/publish.yml

@@ -44,7 +44,7 @@ jobs:
           # manylinux docker image, but I haven't figured out how to install CUDA on manylinux.
           os: [ubuntu-20.04]
           python-version: ['3.9', '3.10', '3.11', '3.12', '3.13']
-          torch-version: ['2.1.2', '2.2.2', '2.3.1', '2.4.0', '2.5.1', '2.6.0']
+          torch-version: ['2.2.2', '2.3.1', '2.4.0', '2.5.1', '2.6.0']
           cuda-version: ['12.4.1']
           # We need separate wheels that either uses C++11 ABI (-D_GLIBCXX_USE_CXX11_ABI) or not.
           # Pytorch wheels currently don't use it, but nvcr images have Pytorch compiled with C++11 ABI.
@@ -53,12 +53,7 @@ jobs:
           cxx11_abi: ['FALSE', 'TRUE']
           exclude:
             # see https://github.com/pytorch/pytorch/blob/main/RELEASE.md#release-compatibility-matrix
-            # Pytorch < 2.2 does not support Python 3.12
-            - torch-version: '2.1.2'
-              python-version: '3.12'
             # Pytorch < 2.5 does not support Python 3.13
-            - torch-version: '2.1.2'
-              python-version: '3.13'
             - torch-version: '2.2.2'
               python-version: '3.13'
             - torch-version: '2.3.1'
@@ -122,8 +117,8 @@ jobs:
           # see https://github.com/pytorch/pytorch/blob/main/RELEASE.md#release-compatibility-matrix
           # This code is ugly, maybe there's a better way to do this.
           export TORCH_CUDA_VERSION=$(python -c "from os import environ as env; \
-            minv = {'2.1': 118, '2.2': 118, '2.3': 118, '2.4': 118, '2.5': 118, '2.6': 118}[env['MATRIX_TORCH_VERSION']]; \
-            maxv = {'2.1': 121, '2.2': 121, '2.3': 121, '2.4': 124, '2.5': 124, '2.6': 124}[env['MATRIX_TORCH_VERSION']]; \
+            minv = {'2.2': 118, '2.3': 118, '2.4': 118, '2.5': 118, '2.6': 118}[env['MATRIX_TORCH_VERSION']]; \
+            maxv = {'2.2': 121, '2.3': 121, '2.4': 124, '2.5': 124, '2.6': 124}[env['MATRIX_TORCH_VERSION']]; \
             print(minv if int(env['MATRIX_CUDA_VERSION']) < 120 else maxv)" \
           )
           if [[ ${{ matrix.torch-version }} == *"dev"* ]]; then

README.md

@@ -65,7 +65,7 @@ flash_attn_interface.flash_attn_func()
 ## Installation and features
 **Requirements:**
 - CUDA toolkit or ROCm toolkit
-- PyTorch 2.1 and above.
+- PyTorch 2.2 and above.
 - `packaging` Python package (`pip install packaging`)
 - `ninja` Python package (`pip install ninja`) *
 - Linux. Might work for Windows starting v2.3.2 (we've seen a few positive [reports](https://github.com/Dao-AILab/flash-attention/issues/595)) but Windows compilation still requires more testing. If you have ideas on how to set up prebuilt CUDA wheels for Windows, please reach out via Github issue.

flash_attn/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "2.7.4"
+__version__ = "2.7.4.post1"
 
 from flash_attn.flash_attn_interface import (
     flash_attn_func,