Question about the modeling_code
So after you code the entire modeling code and the config code do you then push the code to the original Transformers repo or what do we do because I'm confused ??
If you're asking what you should do after using the merge method I outlined in my howto.md file, in order to use the model - I have it already routed to grab the modeling, tokenizer and config files from this repo. If you're asking what you should do after training this model, there shouldn't be any required changes on your end. I will eventually do a pull request to the main transformers repo for Gemmoe, but I want to be sure I have the modeling_files dialed in exactly as I want them to be - so for now having trust_remote_code: True is needed.
If either of those answers didn't answer your question whatsoever, I must have misunderstood!
I see,I get it now thanks a lot