search

saic-violet / bilayer-model

Mozilla Public License 2.0
246 stars 48 forks source link

installation issue #24

Open SaharHusseini opened 2 years ago

SaharHusseini commented 2 years ago

when I try conda install --file requirements.txt

Solving environment: failed with initial frozen solve. Retrying with flexible solve.

PackagesNotFoundError: The following packages are not available from current channels:

Current channels:

Is there any other way to install these packages?

usmancheema89 commented 2 years ago

try pip install -r requirements.txt maybe you will find these on pip

kjczarne commented 1 year ago

I was only able to run the demo notebook using the following environment:

dependencies:
  - python=3.10
  - pip
  - pip:
    - numba
    - six
    - face-alignment
    - tqdm
    - matplotlib
    - gdown
    - jupyterlab

Then I had to manually install a nightly version of Pytorch with CUDA 11.8 support (because I'm using an RTX 4090 at the moment):

pip install --pre torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu118

The requirements file is a mish-mash of packages that are partly only available on PyPI, partly in Anaconda channels. I've spent a few hours trying to get it to work and no dice. Even the Pytorch version listed in requirements.txt seems to be wrong as you get https://github.com/huggingface/transformers/issues/869 with the version listed. Surprisingly, nightly Pytorch 2.0 seems to work!

My suggestion for the requirements file would be to refactor it into a pip-compatible format with a minimal number of packages that are required for the project to work. Could anyone test my solution to this on a machine with an older version of CUDA, where no --pre flag for pip would be required (i.e. with Pytorch 1.13.1)? If this works with Pytorch 1.13.1 just as well, then one could just use the following environment.yml file:

dependencies:
  - python=3.10
  - pip
  - pip:
    - numba
    - six
    - face-alignment
    - tqdm
    - matplotlib
    - gdown
    - jupyterlab
    - torch==1.13.1
    - torchvision
    - torchaudio

Hope this helps! :)