Open violet17 opened 9 months ago
We have reproduced the issue and found a way to optimize it. The corresponding PR will be merged today.
We have reproduced the issue and found a way to optimize it. The corresponding PR will be merged today.
Thanks for quick response.
Already fixed by https://github.com/intel-analytics/BigDL/pull/10203, could you please have a try? @violet17
Already fixed by #10203, could you please have a try? @violet17
Hi ,still got error:
Exception in thread Thread-7 (generate):
Traceback (most recent call last):
File "C:\Users\test\Documents\rag\llmsd_env\lib\threading.py", line 1016, in _bootstrap_inner
self.run()
File "C:\Users\test\Documents\rag\llmsd_env\lib\threading.py", line 953, in run
self._target(*self._args, **self._kwargs)
File "C:\Users\test\Documents\rag\llmsd_env\lib\site-packages\torch\utils\_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "C:\Users\test\Documents\rag\llmsd_env\lib\site-packages\transformers\generation\utils.py", line 1588, in generate
return self.sample(
File "C:\Users\test\Documents\rag\llmsd_env\lib\site-packages\transformers\generation\utils.py", line 2678, in sample
next_tokens = torch.multinomial(probs, num_samples=1).squeeze(1)
RuntimeError: Native API failed. Native API returns: -999 (Unknown PI error) -999 (Unknown PI error)
I got some output text, and VRAM is up to 15.1GB when inferencing. After a while, got this error.
MTL 32GB mem baichuan13B got error.
The input prompt and input token can be found following:
version: