SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

[All models] Extend config.output_attentions with output_attentions function arguments (#4538)

Browse Source 
* DOC: Replace instances of ``config.output_attentions`` with function argument ``output_attentions``

* DOC: Apply Black Formatting

* Fix errors where output_attentions was undefined

* Remove output_attentions in classes per review

* Fix regressions on tests having `output_attention`

* Fix further regressions in tests relating to `output_attentions`

Ensure proper propagation of `output_attentions` as a function parameter
to all model subclasses

* Fix more regressions in `test_output_attentions`

* Fix issues with BertEncoder

* Rename related variables to `output_attentions`

* fix pytorch tests

* fix bert and gpt2 tf

* Fix most TF tests for `test_output_attentions`

* Fix linter errors and more TF tests

* fix conflicts

* DOC: Apply Black Formatting

* Fix errors where output_attentions was undefined

* Remove output_attentions in classes per review

* Fix regressions on tests having `output_attention`

* fix conflicts

* fix conflicts

* fix conflicts

* fix conflicts

* fix pytorch tests

* fix conflicts

* fix conflicts

* Fix linter errors and more TF tests

* fix tf tests

* make style

* fix isort

* improve output_attentions

* improve tensorflow

Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
This commit is contained in:
Bharat Raghunathan
2020-06-10 03:09:06 +05:30
committed by GitHub
parent f90bc44d9a
commit 6e603cb789
38 changed files with 1108 additions and 549 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
tests/test_modeling_tf_common.py
View File

@@ -314,12 +314,11 @@ class TFModelTesterMixin:
)
for model_class in self.all_model_classes:
config.output_attentions = True
inputs_dict["output_attentions"] = True
config.output_hidden_states = False
model = model_class(config)
outputs = model(inputs_dict)
attentions = [t.numpy() for t in outputs[-1]]
self.assertEqual(model.config.output_attentions, True)
self.assertEqual(model.config.output_hidden_states, False)
self.assertEqual(len(attentions), self.model_tester.num_hidden_layers)
self.assertListEqual(
@@ -331,7 +330,6 @@ class TFModelTesterMixin:
if self.is_encoder_decoder:
self.assertEqual(out_len % 2, 0)
decoder_attentions = outputs[(out_len // 2) - 1]
self.assertEqual(model.config.output_attentions, True)
self.assertEqual(model.config.output_hidden_states, False)
self.assertEqual(len(decoder_attentions), self.model_tester.num_hidden_layers)
self.assertListEqual(
@@ -339,13 +337,25 @@ class TFModelTesterMixin:
[self.model_tester.num_attention_heads, decoder_seq_length, decoder_key_length],
)
# Check attention is always last and order is fine
# Check that output attentions can also be changed via the config
del inputs_dict["output_attentions"]
config.output_attentions = True
model = model_class(config)
outputs = model(inputs_dict)
attentions = [t.numpy() for t in outputs[-1]]
self.assertEqual(model.config.output_hidden_states, False)
self.assertEqual(len(attentions), self.model_tester.num_hidden_layers)
self.assertListEqual(
list(attentions[0].shape[-3:]),
[self.model_tester.num_attention_heads, encoder_seq_length, encoder_key_length],
)
# Check attention is always last and order is fine
inputs_dict["output_attentions"] = True
config.output_hidden_states = True
model = model_class(config)
outputs = model(inputs_dict)
self.assertEqual(out_len + (2 if self.is_encoder_decoder else 1), len(outputs))
self.assertEqual(model.config.output_attentions, True)
self.assertEqual(model.config.output_hidden_states, True)
attentions = [t.numpy() for t in outputs[-1]]
@@ -360,11 +370,9 @@ class TFModelTesterMixin:
for model_class in self.all_model_classes:
config.output_hidden_states = True
config.output_attentions = False
model = model_class(config)
outputs = model(inputs_dict)
hidden_states = [t.numpy() for t in outputs[-1]]
self.assertEqual(model.config.output_attentions, False)
self.assertEqual(model.config.output_hidden_states, True)
self.assertEqual(len(hidden_states), self.model_tester.num_hidden_layers + 1)
self.assertListEqual(