Install ipex main branch conda env will meet an error

[hzjane]

Describe the issue

I follow this readme to install conda env, and try to build docker or just install from conda. And I always meet an error. This is my steps:

git clone https://github.com/intel/intel-extension-for-pytorch.git
conda create -n llm python=3.10 -y
conda activate llm
cd ./intel-extension-for-pytorch/examples/cpu/inference/python/llm
export PATH=/root/anaconda3/bin/:$PATH
bash ./tools/env_setup.sh

I think it may be caused by the version conflict. This is the error output:

status code:  0
reformatted /root/wangjian/project/test/intel-extension-for-pytorch/intel_extension_for_pytorch/frontend.py
reformatted /root/wangjian/project/test/intel-extension-for-pytorch/intel_extension_for_pytorch/transformers/optimize.py
reformatted /root/wangjian/project/test/intel-extension-for-pytorch/intel_extension_for_pytorch/transformers/models/reference/modules/attentions.py

All done! ✨ 🍰 ✨
3 files reformatted, 125 files left unchanged.

status code:  1
All done! ✨ 🍰 ✨
3 files left unchanged.

status code:  0
reformatted /root/wangjian/project/test/intel-extension-for-pytorch/tests/cpu/test_paged_attention.py

All done! ✨ 🍰 ✨
1 file reformatted, 90 files left unchanged.

status code:  1
ERROR: flake8 found format errors!
/root/wangjian/project/test/intel-extension-for-pytorch/setup.py:236: DeprecationWarning: distutils Version classes are deprecated. Use packaging.version instead.
  return LooseVersion(line.strip().split(" ")[2])
/root/wangjian/project/test/intel-extension-for-pytorch/setup.py:242: DeprecationWarning: distutils Version classes are deprecated. Use packaging.version instead.
  _cmake_min_version = LooseVersion("3.13.0")
-- The C compiler identification is GNU 12.3.0
-- The CXX compiler identification is GNU 12.3.0
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working C compiler: /root/anaconda3/envs/wangjian-test/bin/gcc - skipped
-- Detecting C compile features
-- Detecting C compile features - done
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: /root/anaconda3/envs/wangjian-test/bin/g++ - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
CMake Error at CMakeLists.txt:34 (find_package):
  Could not find a configuration file for package "Torch" that is compatible
  with requested version "2.3".

  The following configuration files were considered but not accepted:

    /root/anaconda3/envs/wangjian-test/lib/python3.10/site-packages/torch/share/cmake/Torch/TorchConfig.cmake, version: 2.2.0



-- Configuring incomplete, errors occurred!
Traceback (most recent call last):
  File "/root/wangjian/project/test/intel-extension-for-pytorch/setup.py", line 1197, in <module>
    setup(
  File "/root/anaconda3/envs/wangjian-test/lib/python3.10/site-packages/setuptools/__init__.py", line 103, in setup
    return distutils.core.setup(**attrs)
  File "/root/anaconda3/envs/wangjian-test/lib/python3.10/site-packages/setuptools/_distutils/core.py", line 185, in setup
    return run_commands(dist)
  File "/root/anaconda3/envs/wangjian-test/lib/python3.10/site-packages/setuptools/_distutils/core.py", line 201, in run_commands
    dist.run_commands()
  File "/root/anaconda3/envs/wangjian-test/lib/python3.10/site-packages/setuptools/_distutils/dist.py", line 969, in run_commands
    self.run_command(cmd)
  File "/root/anaconda3/envs/wangjian-test/lib/python3.10/site-packages/setuptools/dist.py", line 989, in run_command
    super().run_command(command)
  File "/root/anaconda3/envs/wangjian-test/lib/python3.10/site-packages/setuptools/_distutils/dist.py", line 988, in run_command
    cmd_obj.run()
  File "/root/anaconda3/envs/wangjian-test/lib/python3.10/site-packages/wheel/bdist_wheel.py", line 364, in run
    self.run_command("build")
  File "/root/anaconda3/envs/wangjian-test/lib/python3.10/site-packages/setuptools/_distutils/cmd.py", line 318, in run_command
    self.distribution.run_command(command)
  File "/root/anaconda3/envs/wangjian-test/lib/python3.10/site-packages/setuptools/dist.py", line 989, in run_command
    super().run_command(command)
  File "/root/anaconda3/envs/wangjian-test/lib/python3.10/site-packages/setuptools/_distutils/dist.py", line 988, in run_command
    cmd_obj.run()
  File "/root/anaconda3/envs/wangjian-test/lib/python3.10/site-packages/setuptools/_distutils/command/build.py", line 131, in run
    self.run_command(cmd_name)
  File "/root/anaconda3/envs/wangjian-test/lib/python3.10/site-packages/setuptools/_distutils/cmd.py", line 318, in run_command
    self.distribution.run_command(command)
  File "/root/anaconda3/envs/wangjian-test/lib/python3.10/site-packages/setuptools/dist.py", line 989, in run_command
    super().run_command(command)
  File "/root/anaconda3/envs/wangjian-test/lib/python3.10/site-packages/setuptools/_distutils/dist.py", line 988, in run_command
    cmd_obj.run()
  File "/root/wangjian/project/test/intel-extension-for-pytorch/setup.py", line 1166, in run
    self.run_command("build_clib")
  File "/root/anaconda3/envs/wangjian-test/lib/python3.10/site-packages/setuptools/_distutils/cmd.py", line 318, in run_command
    self.distribution.run_command(command)
  File "/root/anaconda3/envs/wangjian-test/lib/python3.10/site-packages/setuptools/dist.py", line 989, in run_command
    super().run_command(command)
  File "/root/anaconda3/envs/wangjian-test/lib/python3.10/site-packages/setuptools/_distutils/dist.py", line 988, in run_command
    cmd_obj.run()
  File "/root/wangjian/project/test/intel-extension-for-pytorch/setup.py", line 785, in run
    _gen_build_cfg_from_cmake(
  File "/root/wangjian/project/test/intel-extension-for-pytorch/setup.py", line 617, in _gen_build_cfg_from_cmake
    check_call(
  File "/root/anaconda3/envs/wangjian-test/lib/python3.10/subprocess.py", line 369, in check_call
    raise CalledProcessError(retcode, cmd)
subprocess.CalledProcessError: Command '['cmake', '/root/wangjian/project/test/intel-extension-for-pytorch', '-DBUILD_MODULE_TYPE=CPU', '-DBUILD_WITH_XPU=OFF', '-DCMAKE_BUILD_TYPE=Release', '-DCMAKE_INSTALL_LIBDIR=lib', '-DCMAKE_INSTALL_PREFIX=/root/wangjian/project/test/intel-extension-for-pytorch/build/Release/packages/intel_extension_for_pytorch', '-DCMAKE_PREFIX_PATH=/root/anaconda3/envs/wangjian-test/lib/python3.10/site-packages/torch/share/cmake;/root/anaconda3/envs/wangjian-ipex;/root/anaconda3/envs/wangjian-ipex/x86_64-conda-linux-gnu/sysroot/usr', '-DCMAKE_PROJECT_VERSION=2.3.0', '-DIPEX_PROJ_NAME=intel_extension_for_pytorch', '-DLIBIPEX_GITREV=5ed3a2413', '-DLIBIPEX_VERSION=2.3.0+git5ed3a24', '-DPYTHON_EXECUTABLE=/root/anaconda3/envs/wangjian-test/bin/python', '-DPYTHON_INCLUDE_DIR=/root/anaconda3/envs/wangjian-test/include/python3.10', '-DPYTHON_PLATFORM_INFO=Linux-5.19.17-tdx.v3.6.mvp23.el8.x86_64-x86_64-with-glibc2.28']' returned non-zero exit status 1.

I try using branch v2.1.0 and it can successfully run the llama model. I want to use main branch to run baichuan and chatglm model. Please help me, thank you!

[jgong5]

What pytorch version are you using? Have you tried pytorch mainline?

[hzjane]

What pytorch version are you using? Have you tried pytorch mainline?

I run this bash to install related libs: bash ./tools/env_setup.sh . The pytorch version is controlled by the compile_bundle.sh, and the main branch will install version 2.2.0.dev20231213+cpu. Building Intel Extension for PyTorch. Version: 2.3.0+git5ed3a24 will meet the error.

[jgong5]

I run this bash to install related libs: bash ./tools/env_setup.sh . The pytorch version is controlled by the compile_bundle.sh, and the main branch will install version 2.2.0.dev20231213+cpu. Building Intel Extension for PyTorch. Version: 2.3.0+git5ed3a24 will meet the error.

Thanks. I guess that's the problem we should fix from the mainline. cc @jingxu10

On the other hand, we usually recommend users to use a more stable release version of IPEX instead of the mainline directly. May I know any consideration why you want to use the mainline?

[hzjane]

Thanks. I guess that's the problem we should fix from the mainline. cc @jingxu10

On the other hand, we usually recommend users to use a more stable release version of IPEX instead of the mainline directly. May I know any consideration why you want to use the mainline?

Because I want to run Baichuan and chatglm and mistral model which is supported in main branch.

[jgong5]

Because I want to run Baichuan and chatglm and mistral model which is supported in main branch.

Perhaps, you can try to hack the following line to work around the problem. I haven't tried locally though. https://github.com/intel/intel-extension-for-pytorch/blob/5ed3a2413db5f0a5e53bcca0b3e84a814d87bb50/CMakeLists.txt#L33 Change it to: set(Torch_COMP_VERION "2.2")

[hzjane]

Perhaps, you can try to hack the following line to work around the problem. I haven't tried locally though.

I try it and don't meet the error again. But it wll meet a new error after setting up the ENV, the pytorch version does not match the ipex version:

======================================================
Note: Set environment variable "export LD_PRELOAD=/root/anaconda3/envs/wangjian-test/lib/libstdc++.so.6.0.32" to avoid the "version `GLIBCXX_N.N.NN' not found" error.
======================================================
torch_cxx11_abi:     False
torch_version:       2.2.0.dev20231213+cpu
ERROR! Intel® Extension for PyTorch* needs to work with PyTorch/libtorch 2.3.*, but PyTorch/libtorch 2.2.0 is found. Please switch to the matching version and run again.

The install torch version:

pip list | grep torch
intel-extension-for-pytorch              2.3.0+git5ed3a24
torch                                    2.2.0.dev20231213+cpu

[ZailiWang]

Hi, this was a version mismatch issue now it's fixed, please try again with latest master, thanks.