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

[Wav2Vec2] Fix torch srcipt (#24062)

Browse Source 
* [Wav2Vec2] Fix torch srcipt

* fix more
This commit is contained in:
Patrick von Platen
2023-06-07 13:27:07 +02:00
committed by GitHub
parent 612b2a1a6d
commit 52972e70c7
2 changed files with 5 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
src/transformers/models/wav2vec2/modeling_wav2vec2.py
View File

@@ -1178,8 +1178,7 @@ class Wav2Vec2PreTrainedModel(PreTrainedModel):
if isinstance(module, (Wav2Vec2Encoder, Wav2Vec2EncoderStableLayerNorm, Wav2Vec2FeatureEncoder)):
module.gradient_checkpointing = value
@property
def _adapters(self):
def _get_adapters(self):
if self.config.adapter_attn_dim is None:
raise ValueError(f"{self.__class__} has no adapter layers. Make sure to define `config.adapter_attn_dim`.")
@@ -1339,7 +1338,7 @@ class Wav2Vec2PreTrainedModel(PreTrainedModel):
f" directory containing a file named {filepath}."
)
adapter_weights = self._adapters
adapter_weights = self._get_adapters()
unexpected_keys = set(state_dict.keys()) - set(adapter_weights.keys())
missing_keys = set(adapter_weights.keys()) - set(state_dict.keys())

6
tests/models/wav2vec2/test_modeling_wav2vec2.py
View File

@@ -297,7 +297,7 @@ class Wav2Vec2ModelTester:
config.adapter_attn_dim = 16
model = Wav2Vec2ForCTC(config=config)
self.parent.assertIsNotNone(model._adapters)
self.parent.assertIsNotNone(model._get_adapters())
model.to(torch_device)
model.eval()
@@ -1146,7 +1146,7 @@ class Wav2Vec2RobustModelTest(ModelTesterMixin, unittest.TestCase):
model = Wav2Vec2ForCTC.from_pretrained(tempdir)
logits = get_logits(model, input_features)
adapter_weights = model._adapters
adapter_weights = model._get_adapters()
# save safe weights
safe_filepath = os.path.join(tempdir, WAV2VEC2_ADAPTER_SAFE_FILE.format("eng"))
@@ -1168,7 +1168,7 @@ class Wav2Vec2RobustModelTest(ModelTesterMixin, unittest.TestCase):
model = Wav2Vec2ForCTC.from_pretrained(tempdir)
logits = get_logits(model, input_features)
adapter_weights = model._adapters
adapter_weights = model._get_adapters()
# save pt weights
pt_filepath = os.path.join(tempdir, WAV2VEC2_ADAPTER_PT_FILE.format("eng"))