nsidn98 / InforMARL

Code for our paper: Scalable Multi-Agent Reinforcement Learning through Intelligent Information Aggregation
https://nsidn98.github.io/InforMARL/
MIT License
72 stars 22 forks source link

dependencies errors and wandb #1

Closed DH-O closed 11 months ago

DH-O commented 12 months ago

Sorry, I'm not familiar with this kind of code, so under questions can be ridiculous.

  1. The torch version in the requirement.txt and dependencies of MAPE are different from 1.8.0 and 1.11.0. So should I install 1.11.0?

  2. When I pip install libraries, torch-* things are not installed well.

  3. I'm not familiar with wandb. Do you know if the next error is usual? wandb: ERROR Error while calling W&B API: permission denied (<Response [403]>) Should I get your permission...? Or do you mean I should turn off the wandb.

Thanks for your help.

nsidn98 commented 12 months ago

Hi @DH-O no worries at all!

  1. I think we trained/tested our models on torch==1.11.0. Are you getting an error when you try to install MAPE? I have updated the requirement.txt file.
  2. Could you show the errors you are getting while installing the torch related libraries?
  3. You will need to change the --user_name flag to your wandb username. Refer this: https://docs.wandb.ai/guides/app/settings-page/user-settings. You will need to get an API key from wandb as well (it is free to use)
reidsouchultralabs commented 11 months ago

Hi @DH-O no worries at all!

  1. I think we trained/tested our models on torch==1.11.0. Are you getting an error when you try to install MAPE? I have updated the requirement.txt file.
  2. Could you show the errors you are getting while installing the torch related libraries?
  3. You will need to change the --user_name flag to your wandb username. Refer this: https://docs.wandb.ai/guides/app/settings-page/user-settings. You will need to get an API key from wandb as well (it is free to use)

I am also trying to install the repository. I have tried with python 3.7, 3.8, and 3.10 and I get a subprocess-exited-with-error when installing requirements. This happens when the install gets to torch-cluster. Could you please let me know what python version was originally used for this? Maybe that will help.

DH-O commented 11 months ago

1, 2. I solved the issues by installing torch-* things through conda install as below conda install -c esri torch-cluster [(https://anaconda.org/Esri/torch-cluster)]

The problem was that torch==1.11.0 is not compatible with torch-cluster==1.5.9 In my case, requirement.txt should be like this

absl-py
gym==0.10.5
numpy==1.19.4
pyglet==1.5.26
protobuff==3.20.0
PyOpenGL==3.1.5
PyQt5==5.9.2
requests==2.22.0
tensorboard==2.4.0
tensorboard-plugin-wit==1.7.0
tensorboardX==2.1
torch==1.8.1
torch-cluster==1.5.9
torch-geometric==2.0.4
torch-scatter==2.0.7
torch-sparse==0.6.10
torch-spline-conv==1.2.1
torchvision==0.9.1
tqdm==4.44.1
typing==3.6.4
typing-extensions==3.7.4.3
wandb==0.10.31

after that, there seem to be no library issues.

  1. Thank you after changing the user_name flag, it works well.
nsidn98 commented 11 months ago

@DH-O glad that everything works well now. I will update the requirements.txt. Thanks a lot!

nsidn98 commented 11 months ago

@reidsouchultralabs we tested using python 3.8 and 3.10. Could you try installing all torch related stuff using this:

TORCH="1.8.0"
CUDA="cu102"
pip install --no-index torch-scatter -f https://data.pyg.org/whl/torch-${TORCH}+${CUDA}.html --user
pip install --no-index torch-sparse -f https://data.pyg.org/whl/torch-${TORCH}+${CUDA}.html --user
pip install torch-geometric --user

Also, could you show the error traceback so that I can help debugging it?

nsidn98 commented 11 months ago

@reidsouchultralabs were you able to install it? Do you need help in debugging? Can I close this issue?

reidsouchultralabs commented 11 months ago

@reidsouchultralabs were you able to install it? Do you need help in debugging? Can I close this issue?

Hey. I cannot get it installed. I tried with both Python 3.8 and 3.10. With conda, I get dependency errors with the versions in requirement.txt... Do you have any suggestions?

nsidn98 commented 11 months ago

Can you show the dependency errors you are getting?

reidsouchultralabs commented 11 months ago

Can you show the dependency errors you are getting?

With python 3.10, using pip install -r requirement.txt : `Collecting numpy==1.19.4 Using cached numpy-1.19.4.zip (7.3 MB) Installing build dependencies ... done Getting requirements to build wheel ... done Preparing metadata (pyproject.toml) ... error error: subprocess-exited-with-error

× Preparing metadata (pyproject.toml) did not run successfully. │ exit code: 1 ╰─> [288 lines of output] setup.py:67: RuntimeWarning: NumPy 1.19.4 may not yet support Python 3.10.`

Removing numpy and installing separately, I get: ERROR: Could not find a version that satisfies the requirement protobuff==3.20.0 (from versions: none) ERROR: No matching distribution found for protobuff==3.20.0

similarly... ERROR: Could not find a version that satisfies the requirement torch==1.8.1 (from versions: 1.11.0, 1.12.0, 1.12.1, 1.13.0, 1.13.1, 2.0.0, 2.0.1, 2.1.0) ERROR: No matching distribution found for torch==1.8.1

nsidn98 commented 11 months ago

Did you try installing one library at a time instead of pip install -r requirements.txt? Can you give more information about your OS and if you are using conda or not?

reidsouchultralabs commented 11 months ago

Did you try installing one library at a time instead of pip install -r requirements.txt? Can you give more information about your OS and if you are using conda or not?

I am using linux through WSL. When installing with pip individually, I get the same errors. I will try one by one with conda and get back to you.

reidsouchultralabs commented 11 months ago

Did you try installing one library at a time instead of pip install -r requirements.txt? Can you give more information about your OS and if you are using conda or not?

I am using linux through WSL. When installing with pip individually, I get the same errors. I will try one by one with conda and get back to you.

@nsidn98 I was able to get everything installed. I had to use python 3.8 and install each package individually with pip.

nsidn98 commented 11 months ago

@reidsouchultralabs Great! I'll close the issue now. Reopen if you need help with it!