Cannot install apex due to Legacy Install Failure

[shicode]

When I try to install apex using the command below, I get a legacy-install-failure error

pip install -v --disable-pip-version-check --no-cache-dir --deprecated_fused_adam --global-option="--cpp_ext" --global-option="--cuda_ext" ./

Output:

`
File "", line 2, in
File "", line 34, in
File "/home/shilpa/Desktop/ASP/apex/setup.py", line 64, in
check_cuda_torch_binary_vs_bare_metal(torch.utils.cpp_extension.CUDA_HOME)
File "/home/shilpa/Desktop/ASP/apex/setup.py", line 50, in check_cuda_torch_binary_vs_bare_metal
raise RuntimeError("Cuda extensions are being compiled with a version of Cuda that does " +
RuntimeError: Cuda extensions are being compiled with a version of Cuda that does not match the version used to compile Pytorch binaries. Pytorch binaries were compiled with Cuda 11.3.
In some cases, a minor-version mismatch will not cause later errors: #323 (comment). You can try commenting out this check (at your own risk).
error: subprocess-exited-with-error

× Running setup.py install for apex did not run successfully.
│ exit code: 1
╰─> See above for output.

note: This error originates from a subprocess, and is likely not a problem with pip.
full command: /home/shilpa/anaconda3/envs/ASP/bin/python -u -c '
exec(compile('"'"''"'"''"'"'

This is -- a caller that pip uses to run setup.py

- It imports setuptools before invoking setup.py, to enable projects that directly

import from distutils.core to work with newer packaging standards.

- It provides a clear error message when setuptools is not installed.

- It sets sys.argv[0] to the underlying setup.py, when invoking setup.py so

setuptools doesn'"'"'t think the script is -c. This avoids the following warning:

manifest_maker: standard file '"'"'-c'"'"' not found".

- It generates a shim setup.py, for handling setup.cfg-only projects.

import os, sys, tokenize

try:
import setuptools
except ImportError as error:
print(
"ERROR: Can not execute setup.py since setuptools is not available in "
"the build environment.",
file=sys.stderr,
)
sys.exit(1)

file = %r
sys.argv[0] = file

if os.path.exists(file):
filename = file
with tokenize.open(file) as f:
setup_py_code = f.read()
else:
filename = ""
setup_py_code = "from setuptools import setup; setup()"

exec(compile(setup_py_code, filename, "exec"))
'"'"''"'"''"'"' % ('"'"'/home/shilpa/Desktop/ASP/apex/setup.py'"'"',), "", "exec"))' --cpp_ext --cuda_ext install --record /tmp/pip-record-bb1n75tf/install-record.txt --single-version-externally-managed --compile --install-headers /home/shilpa/anaconda3/envs/ASP/include/python3.10/apex
cwd: /home/shilpa/Desktop/ASP/apex/
Running setup.py install for apex ... error
error: legacy-install-failure

× Encountered error while trying to install package.
╰─> apex

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

`

Running cuda version 11.4 with a compatible pytorch version on Ubuntu

[crcrpar]

File "/home/shilpa/Desktop/ASP/apex/setup.py", line 64, in
check_cuda_torch_binary_vs_bare_metal(torch.utils.cpp_extension.CUDA_HOME)
File "/home/shilpa/Desktop/ASP/apex/setup.py", line 50, in check_cuda_torch_binary_vs_bare_metal
raise RuntimeError("Cuda extensions are being compiled with a version of Cuda that does " +
RuntimeError: Cuda extensions are being compiled with a version of Cuda that does not match the version used to compile Pytorch binaries. Pytorch binaries were compiled with Cuda 11.3.
In some cases, a minor-version mismatch will not cause later errors: #323 (comment). You can try commenting out this check (at your own risk).
error: subprocess-exited-with-error

Could you try following the issue and comment mentioned in the error message?

[shicode]

@crcrpar I did. It didn't fix the issue

[mayank31398]

Seeing the same issue :)

[ipostr08]

Same problem.

[ClaudiaSchulz]

Same problem

[hijkzzz]

I have solved this problem by reinstall the CUDA (nvcc toolkit) of the same version of PyTorch.

[ee2110]

Hi, I got the same issue and the error still persists if merely commented out the checking. However, I was able to solve it by reinstalling the PyTorch CUDA version that matches with requirements.

For example, I got the error:
The detected CUDA version (11.6) mismatches the version that was used to compile PyTorch (10.2). Please make sure to use the same CUDA versions.

Therefore, I reinstall the specific PyTorch version from here:
https://pytorch.org/get-started/previous-versions/

After that, I manage to install apex again although I got a bunch of warning lines.

[ptrblck]

I have solved this problem by reinstall the CUDA (nvcc toolkit) of the same version of PyTorch.

However, I was able to solve it by reinstalling the PyTorch CUDA version that matches with requirements.

That's good to hear as it's the right solution.
Closing

[BehnamSherafat]

Hi, I got the same issue and the error still persists if merely commented out the checking. However, I was able to solve it by reinstalling the PyTorch CUDA version that matches with requirements.

For example, I got the error: The detected CUDA version (11.6) mismatches the version that was used to compile PyTorch (10.2). Please make sure to use the same CUDA versions.

Therefore, I reinstall the specific PyTorch version from here: https://pytorch.org/get-started/previous-versions/

After that, I manage to install apex again although I got a bunch of warning lines.

I did the same as you but it takes so long to install Apex and as you said many warnings and weird logs!! I don't know if it installs Apex correctly or not!

[narked165]

It is pretty easy to fix this. The issue is the legacy versioning in requirements.txt
fastapi==0.59.0
xmltodict==0.12.0
pydantic==1.6.1
uvicorn==0.13.4
uvloop==0.15.2
gunicorn==20.0.4
httptools==0.1.1

I just used pip install <package name (ie. fastapi)> One at a time. I suppose, alternatively, you could remove the sempher versioning, in requirements, as well.

uvicorn rrdrest:rrd_rest --host "127.0.0.1"
--port 9000
INFO: Started server process [2520441]
INFO: Waiting for application startup.
INFO: Application startup complete.
INFO: Uvicorn running on http://127.0.0.1:9000 (Press CTRL+C to quit)