Refactor model eos43at

[GemmaTuron]

@HellenNamulinda

the tests are failing, please check, there seems to be a dependency issue

[HellenNamulinda]

Hi @GemmaTuron, The packages installed successfully, as I added in the comment here, Even torch is installed; https://github.com/ersilia-os/eos43at/actions/runs/5606270423/job/15192485169?pr=3#step:8:756, but not seen by the Deep Draph Library(dgl). yet it works well locally.

Detailed error:
🚨🚨🚨 Something went wrong with Ersilia 🚨🚨🚨
Model API eos43at:run did not produce an outputDGL does not detect a valid backend option. Which backend would you like to work with?

Backend choice (pytorch, mxnet or tensorflow): Traceback (most recent call last):
Error message:
  File "/home/runner/eos/repository/eos43at/20230720024131_7E6A30/eos43at/artifacts/framework/predict.py", line 14, in <module>

    from molgrad.net import MPNNPredictor
  File "/home/runner/eos/repository/eos43at/20230720024131_7E6A30/eos43at/artifacts/framework/molgrad/net.py", line 4, in <module>
    from dgl.nn.pytorch import Set2Set
  File "/usr/share/miniconda/envs/eos43at/lib/python3.7/site-packages/dgl/__init__.py", line 8, in <module>
    from .backend import load_backend, backend_name
  File "/usr/share/miniconda/envs/eos43at/lib/python3.7/site-packages/dgl/backend/__init__.py", line 74, in <module>
    load_backend(get_preferred_backend())
  File "/usr/share/miniconda/envs/eos43at/lib/python3.7/site-packages/dgl/backend/__init__.py", line 69, in get_preferred_backend
    backend_name = input("Backend choice (pytorch, mxnet or tensorflow): ").lower()
EOFError: EOF when reading a line

Also, @febielin changes failed on Model Test on PR, though at a different step.

Let me look more into this.

[GemmaTuron]

Hi @HellenNamulinda,

Thanks for the details. It works locally but not on the GitHub runner? Perhaps we could check if in codespaces or the likes it works

[HellenNamulinda]

Yes it works locally. I will try in codespaces.

[HellenNamulinda]

Hi @GemmaTuron, I used Colab to clone my fork and used --repo_path to fetch. The model worked well in Colab. eos43at_colab_fetch_repo.log

For codespaces, despite all packages installing, the model couldn't fetch successfully, giving an error 'NoneType' object is not iterable. eos43at_codespaces_fetch.log. This is an error with codespaces that @febielin and @ZakiaYahya reported last week.

00:53:11 | DEBUG    | InputShapeSingle shape: Single
00:53:11 | DEBUG    | API eos43at:run initialized at URL http://127.0.0.1:36605
00:53:11 | DEBUG    | Schema not yet available
00:53:11 | INFO     | No empty output available
00:53:11 | DEBUG    | Meta: None
00:53:11 | DEBUG    | Posting to run
00:53:11 | DEBUG    | Batch size 100
[]
🚨🚨🚨 Something went wrong with Ersilia 🚨🚨🚨

Error message:

'NoneType' object is not iterable
If this error message is not helpful, open an issue at:
 - https://github.com/ersilia-os/ersilia
Or feel free to reach out to us at:
 - hello[at]ersilia.io

If you haven't, try to run your command in verbose mode (-v in the CLI)
 - You will find the console log file in: /root/eos/current.log

So, I doubt the problem is with the refactored code since it works well using Colab. I would have suggested re-running the action. But since, we can't re-run the workflows now, I will say we first pause it. I will continue examining it though.

[GemmaTuron]

Hi @HellenNamulinda

In the log I see more than once the following line:

Collecting torch==1.9.0
Terminated

It appears that Torch is not being compiled in codespaces. In local, it works fetching from repo_path or only from run.sh? Seeing this error and the error on the model testing, I think there is a dependencies problem only, the model is fine

[HellenNamulinda]

Hi @GemmaTuron, Locally, the model works using both run.sh and within ersilia using --repo_path. 43at_local_fetch.log

For Colab, it also works using --repo_path

Could it be due to ram and storage? Maybe we discuss more about it during the meeting today.

[GemmaTuron]

Let's check the workflow once we can run it