Cannot load finetuned model due to changes in *_RW.py files

#6
by marcovirgolin - opened

Hi there.

I finetuned the falcon model prior to the changes to the configuration and modelling files (from *_RW.py to *_falcon.py).
Now, when I try to load my local model, I get:
Could not locate the configuration_RW.py inside tiiuae/falcon-rw-1b.

I tried passing the revision field, pointing to the commit hash that preceeds the changes, with:
model = AutoModelForCausalLM.from_pretrained( local_path, trust_remote_code=True, revision="7fb349a0a7b09213458a45a2861342c7f2d2d3fc")
But I still get that error. I see, via debugging, that revision==None as it goes through the transformers library code.

Can you help me with this? I'd be great if backward compatibility is not lost.

Sign up or log in to comment