search

datamol-io / molfeat

molfeat - the hub for all your molecular featurizers
https://molfeat.datamol.io
Apache License 2.0
169 stars 16 forks source link

GPT2-Zinc480M-87M and Roberta-Zinc480M-102M not retrievable #79

Closed arunraja-hub closed 8 months ago

arunraja-hub commented 9 months ago

Is there an existing issue for this?

Bug description

I have been trying to use GPT2-Zinc480M-87M and Roberta-Zinc480M-102M and I am getting the following error. The code is as https://molfeat.datamol.io/featurizers/GPT2-Zinc480M-87M The error message:

 molfeat.store.modelstore.ModelStoreError: Can't retrieve model GPT2-Zinc480M-87M from the store !

How to reproduce the bug

https://molfeat.datamol.io/featurizers/GPT2-Zinc480M-87M

Error messages and logs

 molfeat.store.modelstore.ModelStoreError: Can't retrieve model GPT2-Zinc480M-87M from the store !

Environment

Current environment ``` #- Molfeat version (e.g., 0.1.0): #- PyTorch Version (e.g., 1.10.0): #- RDKit version (e.g., 2022.09.5): #- scikit-learn version (e.g., 1.2.1): #- OS (e.g., Linux): #- How you installed Molfeat (`conda`, `pip`, source): ```

Additional context

No response

maclandrol commented 9 months ago

Hello @arunraja-hub , I have attempted several time to reproduce this bug without success.

Let me know if which steps you took.

dyan-yen commented 9 months ago

Hi, I had the same issue earlier today too. In my case, if I downgrade the package fsspec v2023.9.2 to v2023.6.0, the error would be gone and everything works for me again.

(I am using molfeat v0.9.4)

maclandrol commented 9 months ago

Thanks @dyan-yen, I think it might be a bug recently introduced in fsspec. I believe they don't auto-create folder on file opening on write mode anymore. Should be easy to fix.