search

Dao-AILab / flash-attention

Fast and memory-efficient exact attention
BSD 3-Clause "New" or "Revised" License
13.26k stars 1.2k forks source link

The cuda version is 11.6. Why does this error still appear during installation? #842

Open lalalabobobo opened 6 months ago

lalalabobobo commented 6 months ago

(llm) (base) hong.h@hit-airline-utah-cn5-hong-h-master-0:~$ nvcc -V nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2022 NVIDIA Corporation Built on Tue_Mar__8_18:18:20_PST_2022 Cuda compilation tools, release 11.6, V11.6.124 Build cuda_11.6.r11.6/compiler.31057947_0 (llm) (base) hong.h@hit-airline-utah-cn5-hong-h-master-0:~$ python -c "import torch; print(torch.version)"

1.13.0 (llm) (base) hong.h@hit-airline-utah-cn5-hong-h-master-0:~$ (llm) (base) hong.h@hit-airline-utah-cn5-hong-h-master-0:~$ pip show Ninja Name: ninja Version: 1.11.1.1 Summary: Ninja is a small build system with a focus on speed Home-page: http://ninja-build.org/ Author: Jean-Christophe Fillion-Robin Author-email: scikit-build@googlegroups.com License: Apache 2.0 Location: /home/hong.h/miniconda3/envs/llm/lib/python3.9/site-packages Requires: Required-by: deepspeed (llm) (base) hong.h@hit-airline-utah-cn5-hong-h-master-0:~$ pip install flash-attn --no-build-isolation Looking in indexes: https://artifactory.nioint.com/artifactory/api/pypi/dd-pypi-all-virtual/simple, https://artifactory.nioint.com/artifactory/api/pypi/dd-pypi-all-virtual/simple Collecting flash-attn Downloading https://artifactory.nioint.com/artifactory/api/pypi/dd-pypi-all-virtual/packages/packages/4e/7b/b42299ad0edf4d185c883dcf83d4f31bf577e8e86f20e70b8ea8ebc0d75d/flash_attn-2.5.4.tar.gz (2.5 MB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 2.5/2.5 MB 7.7 MB/s eta 0:00:00 Preparing metadata (setup.py) ... error error: subprocess-exited-with-error

× python setup.py egg_info did not run successfully. │ exit code: 1 ╰─> [13 lines of output] fatal: not a git repository (or any parent up to mount point /) Stopping at filesystem boundary (GIT_DISCOVERY_ACROSS_FILESYSTEM not set). Traceback (most recent call last): File "", line 2, in File "", line 34, in File "/tmp/pip-install-2__yxtvn/flash-attn_4d82813581cc4fdd941d2b9b958ddb89/setup.py", line 114, in raise RuntimeError( RuntimeError: FlashAttention is only supported on CUDA 11.6 and above. Note: make sure nvcc has a supported version by running nvcc -V.

  torch.__version__  = 1.13.0

  [end of output]

note: This error originates from a subprocess, and is likely not a problem with pip. error: metadata-generation-failed

× Encountered error while generating package metadata. ╰─> See above for output.

note: This is an issue with the package mentioned above, not pip. hint: See above for details.

Violet969 commented 6 months ago

Same problem! Are you solved it?