GPTQ enablement

[mikekgfb]

GPTQ is currently not available in llama-fast

@HDCharles can you please bring the GPTQ support from gpt-fast over to llama-fast?

x-ref: pytorch-labs/gpt-fast#148

Thanks so much!

@supriyar please re-assign if appropriate

[supriyar]

@jerryzh168 assigning this to you based on our chat. Let's use the torchao APIs that we used for TorchTune in llama-fast as well?

[mikekgfb]

Resolved with #100