Add Ascend NPU support for nf4 quant

[statelesshz]

What does this PR do?

This PR adds Ascend NPU support for nf4 quant/dequant and allows QLoRA fine-tuning for LLMs using transformers, peft, and trl.

You may notice that the nf4 quantization method is currently implemented in PyTorch. This interim measure is due to the fact that the high-performance version implemented with AscendC is still in progress 😞 . Meanwhile, we've received feedback from many in the Ascend NPU community expressing their keen interest in using QLoRA to fine-tune LLMs as soon as possible, so there is this PR.

Related PR: huggingface/transformers#31512

Collaborators

@SlightwindSec @Ginray @MatrixPlayer

cc @Titus-von-Koeller @matthewdouglas

[statelesshz]

Refer to this blog, I did a E2E test on the llama2-7b-hf with QLoRA fine-tuning in my env with NPU device, it works 🤗.

Here is the script I used.

[baymax591]

Thanks a lot for sharing this PR and the video demo! Thanks to the demo, I was able to successfully run NF4 quant/dequant on the NPU with ease. The detailed explanation in the video really helped me understand the process and key steps. Looking forward to more updates in the future—great work!

[baymax591]

I hope this PR can be merged soon, as it provides valuable improvements. Looking forward to seeing it merged!
cc @Titus-von-Koeller

[SunMarc]

Nice work and thanks for the demo ! Can you have a look @matthewdouglas ?

[github-actions]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[matthewdouglas]

I will be able to look in more detail next week, but at first glance it looks nice. Thanks @statelesshz !