Update transformers requirement from <2.11,>=2.9 to >=2.9,<2.12 (#4315)

* Update transformers requirement from <2.11,>=2.9 to >=2.9,<2.12

Updates the requirements on [transformers](https://github.com/huggingface/transformers) to permit the latest version.
- [Release notes](https://github.com/huggingface/transformers/releases)
- [Commits](huggingface/transformers@v2.9.0...v2.11.0)

Signed-off-by: dependabot-preview[bot] <[email protected]>

* try fix

Co-authored-by: dependabot-preview[bot] <27856297+dependabot-preview[bot]@users.noreply.github.com>
Co-authored-by: epwalsh <[email protected]>

Author: dependabot-preview[bot]

setup.py

@@ -64,7 +64,7 @@
         "scikit-learn",
         "scipy",
         "pytest",
-        "transformers>=2.9,<2.11",
+        "transformers>=2.9,<2.12",
         "jsonpickle",
         "dataclasses;python_version<'3.7'",
         "filelock>=3.0,<3.1",

tests/modules/token_embedders/pretrained_transformer_embedder_test.py

@@ -214,7 +214,7 @@ def test_unfold_long_sequences(self):
         assert (unfolded_embeddings_out == unfolded_embeddings).all()
 
     def test_encoder_decoder_model(self):
-        token_embedder = PretrainedTransformerEmbedder("bart-large", sub_module="encoder")
+        token_embedder = PretrainedTransformerEmbedder("facebook/bart-large", sub_module="encoder")
         token_ids = torch.LongTensor([[1, 2, 3], [2, 3, 4]])
         mask = torch.ones_like(token_ids).bool()
         token_embedder(token_ids, mask)