ml-jku / clamp Goto Github PK

Dear Mr. Philipp Seidl,

I appreciate your keen work.
I am encountering an error while trying to run the encode_compound.py script for the PubChem23 dataset. I am not sure if the error is due to the downloaded dataset or if there is a bug in the script. Can you please help me look at the attached file?

Thank you for your time and assistance.
Best regards,
Linh

I run the repo code in the readme, and the error is as follow:
RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory
I located the origin code, then found it is about the checkpoint file. I re-downloaded this file, but it still didn't work. How can I solve this problem?
clamp-main/clamp/models/pretrained.py", line 32, in init
cp = torch.load(self.checkpoint, map_location=device)

Hi, thank you sharing your great work.
I would like to build the clamp environment using docker.
It would be very helpful if you have Dockerfile for that.

Sincerely.

model = clamp.CLAMP(device='cpu') causes error:
AttributeError: module 'clamp' has no attribute 'CLAMP'

Hi,

When running the code, it seems to fail on the import clamp command:

File ~/DD_tools/clamp/clamp/models/models.py:1
----> 1 from clamp.deeptox.models.perceptron import NetworkLayerNorm, MultilayerPerceptron
      3 from typing import List, Tuple
      5 import torch

ModuleNotFoundError: No module named 'clamp.deeptox'

Is this import correct, this is also the same in the clamp/models/scaled.py and clamp/models/gnn.py scripts.

Thanks

Hi Phillip,

Do you have a PyTorch v2 training script with multi-GPU support? If so, would you be able to share it with me?

As far as I know, your Arxiv paper states that the total compute runtime was around 170 days and 800 runs (without linear probing). I am wondering why you used only one GPU to train the huge model.

Have a nice weekend.
Linh

Hi, thank you for sharing your great work!
I am interested in the concept of your paper and would like to try pretraining as written in your paper.
How can I pretrain using this repository?
Another question is about the computational resource.
In your paper, it took total 170 days and 800 times. Does the pretraining require the same computational time?
Is it possible to pretrain using a single GPU?

Thank you in advance:)