You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Now that we're transitioning between models that use ocf_datapipes and ones using different versions of ocf-data-sampler, it would be useful to save information on the versions of tools used to train it. I think it should go into the model card on HF under 'Software' (we currently have links to repos there, which default to ocf_datapipes regardless). I'd suggest to make the change somewhere in here in the model's save_pretrained method, but open to suggestions.
Ideally, it would automatically save the versions used, but this might require changes to have we save checkpoints, so everything is stored correctly. If this is not possible, maybe it can prompt the user to input the libraries and versions on save.
The text was updated successfully, but these errors were encountered:
Detailed Description
Now that we're transitioning between models that use ocf_datapipes and ones using different versions of ocf-data-sampler, it would be useful to save information on the versions of tools used to train it. I think it should go into the model card on HF under 'Software' (we currently have links to repos there, which default to ocf_datapipes regardless). I'd suggest to make the change somewhere in here in the model's save_pretrained method, but open to suggestions.
Ideally, it would automatically save the versions used, but this might require changes to have we save checkpoints, so everything is stored correctly. If this is not possible, maybe it can prompt the user to input the libraries and versions on save.
The text was updated successfully, but these errors were encountered: