add gist model generation utils to library

[uSaiPrashanth]

Other modifications worth mentioning:

• Changed scripts/convert_hf_checkpoint.py to support loading of finetuned Llama-3 models from safetensors state dict
• Added finetuned configs to model.py (Finetuned models use a vocab size of one token higher than Llama 3)

[uSaiPrashanth]

Thank you for the review! I will try to integrate the changes into this PR soon

we do have gist models with gist token after instruction and input and only instruction Uploaded to huggingface.

• We can train another model with gist tokens after input as well as instruction and then compare these to see which one works the best after evaluation