Block torchscript pytest because of seg fault issue#793
Conversation
|
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. |
yeonsily
requested review from
bhargaveede,
ssarkar2 and
vivekgoe
as code owners
| ) | ||
| model_kwargs["bucket_size"] = generation_config.bucket_size if generation_config.static_shapes else -1 | ||
| model_kwargs["bucket_internal"] = generation_config.bucket_internal | ||
| if generation_config.bucket_internal: |
There was a problem hiding this comment.
DO we need this if check ? This is already defined in optimum-habana/optimum/habana/transformers/generation/configuration_utils.py. And we don't do same for other generation config flags such as attn_softmax_bf16
There was a problem hiding this comment.
PR781 causes 2 issues in bert pytest BertModelTest.test_greedy_generate 1. ValueError: The following model_kwargs are not used by the model: ['use_hpu_graphs'] (note: typos in the generate arguments will also show up in this list)
2 bucket_internal = model_kwargs["bucket_internal"]
[2024-03-13T18:54:28.818Z] E KeyError: 'bucket_internal'
5 participants
What does this PR do?
Fixes # (issue)
Before submitting