Shengsheng Huang

Thanks for reporting. We'll look into it.

We are fixing this in PR: https://github.com/intel-analytics/BigDL/pull/10051

please resolve conflicts.

> It can work when I set cpu_embedding=True for both HF-transformer API and Pytorch API. Does `cpu_embedding=True` has negative impact for Arc? If not, maybe we can add this option...

Sure. In general the fix looks good to me. I've started the PR validation tasks and let's wait for its results.

### about the `ValueError: Unrecognized configuration class` This error is not quite relevant to bigdl's support. The error message indicates the `AutoModelForCausalLM` does not support loading the speech t5 model....

SpeechT5 model can be successfully loaded using bigdl using `AutoModelForSpeechSeq2Seq`, instead of `AutoModelForCasualLM`. Below code works in our test (using transformers version 4.31.0 and bigdl version 2.5.0b20240124). ```python from bigdl.llm.transformers...

> > > According to the backtrace, it seems like issue with finding the GPU. `sycl-ls` shows a discrete GPU. I am using oneAPI 2023.2 and kernel 5.19-0.41 and is...

Could you provide the os, kernel and python version?

> @rnwang04 GPU finetuning suddenly stopped working and gave Seg Fault. > > are you running it inside vscode?