Copied from for test files
#26713
Copied from for test files
#26713
Conversation
|
well, I need to check why fails. |
|
The documentation is not available anymore as the PR was closed or merged. |
| # Detail: the `Copied from` statement is originally designed to work with the last part of `TRANSFORMERS_PATH`, | ||
| # (which is `transformers`). The same should be applied for `MODEL_TEST_PATH`. However, its last part is `models` | ||
| # (to only check and search in it) which is a bit confusing. So we keep the copied statement staring with | ||
| # `tests.models.` and change it to `tests` here. | ||
| if base_path == MODEL_TEST_PATH: | ||
| base_path = "tests" |
There was a problem hiding this comment.
This is to make # Copied from tests.models. work (instead of # Copied from models.) and still keep MODEL_TEST_PATH = "tests/models"
|
Will have to perform some fixes about copied statements before merge. |
There was a problem hiding this comment.
The changes to the check_copies.py file look good to me. I'm curious if we couldn't rework slightly our test files so that close-to-exact copies can be copied as entire files rather than individual methods. I think this would make contributions quite simpler
(if not, it's ok as long as users call add-new-model-like)
|
|
||
| # Copied from transformers.tests.roberta.test_modeling_roberta.py with Roberta->Longformer | ||
| @require_tokenizers | ||
| class LongformerTokenizationTest(TokenizerTesterMixin, unittest.TestCase): |
There was a problem hiding this comment.
Would be simpler for contributors to have the full class copied instead of adding manual copied from statements to so many tests. It wasn't possible in this case?
|
|
||
| # Copied from transformers.tests.mistral.test_modelling_mistral.MistralModelTest with Llama->Mistral | ||
| class MistralModelTester: |
There was a problem hiding this comment.
Here too it would be really great to keep the class-wide comment
|
If there is a single place, say |
|
Ok! |
What does this PR do?
Copied fromfor test files.Run
make fix-copieswill showI will need to update
test_tokenization_longformer.pybefore merge.