some questions about FlashAttention

[Rolin-zrl]

Building the wheel is very slow. I chose the version "flash_attn-2.6.1+cu118torch2.1cxx11abiTRUE-cp310-cp310-linux_x86_64.whl" to download and install via pip, based on the specifications torch==2.11, cuda==11.8, python==3.10 on my Linux system. there still exists errors, as shown below:
ImportError: /usr/lib/x86_64-linux-gnu/libstdc++.so.6: version `GLIBCXX_3.4.26' not found (required by /home/whuav/anaconda3/envs/MeshAnything/lib/python3.10/site-packages/flash_attn_2_cuda.cpython-310-x86_64-linux-gnu.so)

any suggestions?

[Rolin-zrl]

after upgrading pytorch, it's working. =='
pip install torch==2.2.2 torchvision==0.17.2 torchaudio==2.2.2 --index-url https://download.pytorch.org/whl/cu118