Unable to find installation candidates for torch (2.0.1+cpu)

[maxwkut]

Great course so far!

I'm having some trouble running make install in the streaming_pipeline folder. Any suggestions?

Main dependencies are all installed.

[iusztinpaul]

Thanks! Happy to hear that.

What OS are you using?

[maxwkut]

I'm using macOS Sonoma 14. Not sure if this is a mac specific problem though - in the video tutorial, it looked like @Paulescu ran the code on a mac as well and didn't have any issues.

[iusztinpaul]

Yeah. You are right.

The streaming pipeline (out of the 3 modules) should work on macos. I tested my code only on Ubuntu. @Paulescu, do you know about this issue? Have you encountered it?

[Paulescu]

@maxwkut Have you tried removing the pip install torch... from the makefile and running poetry add torch ?

[iusztinpaul]

@maxwkut Have you managed to make it work?

[maxwkut]

@Paulescu @iusztinpaul I took out the torch-cpu dependency from pyproject.toml and then added the regular torch package using poetry add torch. Not sure if torch is a valid substitute for torch-cpu, but that at least allowed me to proceed to the next step: make run_real_time. Here I am now getting an error that there is an incompatible architecture (have arm64, but binary was compiled for x86_64)

[iusztinpaul]

Try to remove the torch completely from pyproject.toml by running poetry remove torch and then install it in your Poetry virtual environment using pip: poetry run pip install torch==2.0.1

Let me know if this works.

If that doesn't work, try to install it directly from the wheel, depending on your platform, something like this:

poetry add "https://download.pytorch.org/whl/cu118/torch-2.0.0%2Bcu118-cp310-cp310-linux_x86_64.whl" --platform linux
poetry add "https://download.pytorch.org/whl/cu118/torch-2.0.0%2Bcu118-cp310-cp310-linux_x86_64.whl" --platform darwin

Let me know if that works for you.

[maxwkut]

As you can see, I ran poetry run pip install torch == 2.0.1 after removing it.

It's saying that streaming-pipeline 0.1.0 requires torch<3.0.0,>=2.1.2?

I also tried running make run_real_time afterwards and the incompatible architecture error still persists

[iusztinpaul]

Have you tried the second option?