Fixed transformers version conflict and fixed UT error #483
Conversation
Signed-off-by: Cheng, Penghui <[email protected]>
Signed-off-by: Cheng, Penghui <[email protected]>
Signed-off-by: Cheng, Penghui <[email protected]>
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. |
|
https://github.com/huggingface/optimum-intel/actions/runs/7097503652/job/19317724224?pr=483#step:5:747 This case failed due to HF connect issue, @echarlaix could you help to check? |
|
Could you rebase @PenghuiCheng ? (recent fixes were integrated for the failing tests) |
Signed-off-by: Cheng, Penghui <[email protected]>
Signed-off-by: Cheng, Penghui <[email protected]>
| @@ -77,7 +77,7 @@ class OptimizationTest(INCTestMixin): | |||
| ) | |||
|
|
|||
| TEXT_GENERATION_SUPPORTED_ARCHITECTURES = ( | |||
| "hf-internal-testing/tiny-random-BloomForCausalLM", | |||
| "hf-tiny-model-private/tiny-random-BloomForCausalLM", | |||
There was a problem hiding this comment.
why this modification ?
There was a problem hiding this comment.
there is error in test_quantize_text_generate_model UT with "hf-internal-testing/tiny-random-BloomForCausalLM" :
2023-12-19 11:14:01 [ERROR] Unexpected exception RuntimeError('INDICES element is out of DATA bounds, id=6168 axis_dim=1024') happened during tunin
g.
Traceback (most recent call last):
File "/home/penghuic/inc_pr/neural_compressor/quantization.py", line 234, in fit
strategy.traverse()
File "/home/penghuic/inc_pr/neural_compressor/strategy/auto.py", line 140, in traverse
super().traverse()
File "/home/penghuic/inc_pr/neural_compressor/strategy/strategy.py", line 504, in traverse
q_model = self.adaptor.quantize(copy.deepcopy(tune_cfg), self.model, self.calib_dataloader, self.q_func)
File "/home/penghuic/inc_pr/neural_compressor/utils/utility.py", line 304, in fi
res = func(*args, **kwargs)
File "/home/penghuic/inc_pr/neural_compressor/adaptor/pytorch.py", line 3630, in quantize
q_func(q_model._model)
File "/home/penghuic/optimum-intel-new/tests/neural_compressor/test_optimization.py", line 368, in calibration_fn
tmp_model.generate(**tokens, max_new_tokens=32, do_sample=False)
File "/home/penghuic/anaconda3/envs/py3.10/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
There was a problem hiding this comment.
thanks for your answer @PenghuiCheng, could you let me know in which case this issue is appearing and if yes what is causing it? I don't see it in the current CI, is this appearing when installing neural-compressor from source ?
There was a problem hiding this comment.
any news on this @PenghuiCheng ?
There was a problem hiding this comment.
Hi, Ella, Yes, This issue does not exist, it seems that my model has not been fully downloaded. I will close it. thanks!
Signed-off-by: Cheng, Penghui <[email protected]>
|
hi, Ella, the pre-ci tests on the OpenVino section. It is not the cause of this PR. Could you review this PR? |
Yes the failing tests are unrelated to the PR and can be ignored |
What does this PR do?
Fixed transformers version conflict and fixed UT error