[Request] Add ipex support

[fablevi]

Hi guys!

Great model, can you add support for ipex(intel) and rocm(amd)

[wchengad]

hi, @fablevi , thank you for your interest in our work. We are a small research interest group, and adapting models to different architectures is not our expertise. We warmly welcome contributors from the community with valuable pull requests.

[Jasonzzt]

Hi @wchengad @fablevi
I've supported OmniLottie on intel xpu B60, but the generation output is not steady and even wierd compared to output of Huggingface demo.
Also I got the same result on 4090.

[fablevi]

@Jasonzzt thanks for update

So to be clear. On B60 and 4090 the output is kinda wierd?

[Jasonzzt]

@Jasonzzt thanks for update

So to be clear. On B60 and 4090 the output is kinda wierd?

Yes!

[wchengad]

hi @Jasonzzt , If you got steady output aligned with Hugging Face or the demos on N cards, feel free to submit your changes as a pull request. We’d be happy to review it and merge it into the main branch.

[Jasonzzt]

Hi @wchengad
PR #5 has been submitted

hi @Jasonzzt , If you got steady output aligned with Hugging Face or the demos on N cards, feel free to submit your changes as a pull request. We’d be happy to review it and merge it into the main branch.

[wchengad]

@Jasonzzt Thanks for the contribution!