ERROR [12/13] RUN pip install flash-attn --no-build-isolation

[promaprogga]

# Use an official Python runtime as a parent image
FROM python:3.10-slim

# Set the working directory in the container
WORKDIR /app

# Copy the current directory contents into the container
COPY . /app

# Install system dependencies
RUN apt-get update && apt-get install -y \
    git \
    && rm -rf /var/lib/apt/lists/*

# Install Python dependencies
RUN pip install --upgrade pip

# Install torch and torchvision first
RUN pip install torch==2.0.1 torchvision==0.15.2

# Install remaining Python libraries
RUN pip install transformers==4.37.2 flask packaging

RUN pip install flash-attn --no-build-isolation

getting this error when I tried to build the dockerfile in linux

=> ERROR [12/13] RUN pip install flash-attn --no-build-isolation                                                  2.8s
------
 > [12/13] RUN pip install flash-attn --no-build-isolation:
0.526 Collecting flash-attn
0.833   Downloading flash_attn-2.6.3.tar.gz (2.6 MB)
1.369      ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 2.6/2.6 MB 7.3 MB/s eta 0:00:00
1.643   Preparing metadata (setup.py): started
2.507   Preparing metadata (setup.py): finished with status 'error'
2.510   error: subprocess-exited-with-error
2.510
2.510   × python setup.py egg_info did not run successfully.
2.510   │ exit code: 1
2.510   ╰─> [20 lines of output]
2.510       fatal: not a git repository (or any of the parent directories): .git
2.510       /tmp/pip-install-arl6lbbl/flash-attn_d2269f4c4a7f49df8a1d94c7a99ab377/setup.py:95: UserWarning: flash_attn was requested, but nvcc was not found.  Are you sure your environment has nvcc available?  If you're installing within a container from https://hub.docker.com/r/pytorch/pytorch, only images whose names contain 'devel' will provide nvcc.
2.510         warnings.warn(
2.510       Traceback (most recent call last):
2.510         File "<string>", line 2, in <module>
2.510         File "<pip-setuptools-caller>", line 34, in <module>
2.510         File "/tmp/pip-install-arl6lbbl/flash-attn_d2269f4c4a7f49df8a1d94c7a99ab377/setup.py", line 179, in <module>
2.510           CUDAExtension(
2.510         File "/usr/local/lib/python3.10/site-packages/torch/utils/cpp_extension.py", line 1048, in CUDAExtension
2.510           library_dirs += library_paths(cuda=True)
2.510         File "/usr/local/lib/python3.10/site-packages/torch/utils/cpp_extension.py", line 1179, in library_paths
2.510           if (not os.path.exists(_join_cuda_home(lib_dir)) and
2.510         File "/usr/local/lib/python3.10/site-packages/torch/utils/cpp_extension.py", line 2223, in _join_cuda_home
2.510           raise EnvironmentError('CUDA_HOME environment variable is not set. '
2.510       OSError: CUDA_HOME environment variable is not set. Please set it to your CUDA install root.
2.510
2.510
2.510       torch.__version__  = 2.0.1+cu117
2.510
2.510
2.510       [end of output]
2.510
2.510   note: This error originates from a subprocess, and is likely not a problem with pip.
2.524 error: metadata-generation-failed
2.524
2.524 × Encountered error while generating package metadata.
2.524 ╰─> See above for output.
2.524
2.524 note: This is an issue with the package mentioned above, not pip.
2.524 hint: See above for details.
------
Dockerfile:36
--------------------
  34 |     RUN pip install ninja
  35 |
  36 | >>> RUN pip install flash-attn --no-build-isolation
  37 |     # Expose the Flask port
  38 |     EXPOSE 5000
--------------------
ERROR: failed to solve: process "/bin/sh -c pip install flash-attn --no-build-isolation" did not complete successfully: exit code: 1