Issues with the gptfast setup scripts

[vayuda]

convert_hf_checkpoint.py attempts to import ModelConfig from models.py, but this file is not in the right place for this import statement to work

This is a result of just copying from the gptfast repo most likely
In that repo, model.py is located in the parent directory but for us its in torchao/_models/llama

The fix is simple enough but I'm using this issue as a placeholder while I continue testing this

[HDCharles]

Yeah, you're right, I guess i had gpt-fast in my path so it didn't error when i tested this.