Fix num. of minimal calls to the Hub with peft for pipeline #26385
Conversation
| pretrained_model_name_or_path = None | ||
| if isinstance(config, str): | ||
| pretrained_model_name_or_path = config | ||
| elif config is None and isinstance(model, str): | ||
| pretrained_model_name_or_path = model |
There was a problem hiding this comment.
try to get the repo. in different case before going to the next block of getting config, similar to the block in line 758
| f"{model} is not a valid model_id." | ||
| ) | ||
| task = get_task(model, use_auth_token) | ||
| task = get_task(model, token) |
There was a problem hiding this comment.
use_auth_token is already deprecated.
|
The documentation is not available anymore as the PR was closed or merged. |
We do have tests: 4 tests named |
| elif config is None and isinstance(model, str): | ||
| # Check for an adapter file in the model path if PEFT is available | ||
| if is_peft_available(): | ||
| subfolder = hub_kwargs.get("subfolder", None) |
There was a problem hiding this comment.
subfolder was never in hub_kwargs here
|
Merge now. |
5 participants
What does this PR do?
This is a continuation of #25715 where @sgugger fix
test_cached_model_has_minimum_calls_to_headbut nottest_cached_pipeline_has_minimum_calls_to_head(as I didn't mention this to him).This PR mostly copies the logic from #25715.
The current failing test and the error are
tests/pipelines/test_pipelines_common.py::CustomPipelineTest::test_cached_pipeline_has_minimum_calls_to_head (line 905) AssertionError: 2 != 1