[FEATURE] Qwen-7B Request

[152334H]

Qwen-7B is a model that (allegedly) outperforms Llama-2-13B on important benchmarks like MMLU, HumanEval, GSM8K, etc. It has a chat model as well, and is available for commercial use for people with <100Mil monthly active users.

It is architecturally fairly similar to llama, but unfortunately has custom modelling code, and therefore does not immediately work with this repo.

[CheshireAI]

I'd also be thrilled with this. I don't know if it helps but they have it implemented in these projects:
https://github.com/cckuailong/SuperAdapters
https://github.com/yangjianxin1/Firefly
https://github.com/ssbuild/qwen_finetuning

[qwopqwop200]

#240