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

Fix wav2vec2 is_batched check to include 2-D numpy arrays (#23223)

Browse Source 
* Fix wav2vec2 is_batched check to include 2-D numpy arrays

* address comment

* Add tests

* oops

* oops

* Switch to np array

Co-authored-by: Sanchit Gandhi <93869735+sanchit-gandhi@users.noreply.github.com>

* Switch to np array

* condition merge

* Specify mono channel only in comment

* oops, add other comment too

* make style

* Switch list check from falsiness to empty

---------

Co-authored-by: Sanchit Gandhi <93869735+sanchit-gandhi@users.noreply.github.com>
This commit is contained in:
LWprogramming
2023-05-22 09:57:45 -07:00
committed by GitHub
parent 4ddd9de9d3
commit 5de2a6d5e5
5 changed files with 31 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
tests/models/wav2vec2/test_feature_extraction_wav2vec2.py
View File

@@ -123,6 +123,14 @@ class Wav2Vec2FeatureExtractionTest(SequenceFeatureExtractionTestMixin, unittest
for enc_seq_1, enc_seq_2 in zip(encoded_sequences_1, encoded_sequences_2):
self.assertTrue(np.allclose(enc_seq_1, enc_seq_2, atol=1e-3))
# Test 2-D numpy arrays are batched.
speech_inputs = [floats_list((1, x))[0] for x in (800, 800, 800)]
np_speech_inputs = np.asarray(speech_inputs)
encoded_sequences_1 = feat_extract(speech_inputs, return_tensors="np").input_values
encoded_sequences_2 = feat_extract(np_speech_inputs, return_tensors="np").input_values
for enc_seq_1, enc_seq_2 in zip(encoded_sequences_1, encoded_sequences_2):
self.assertTrue(np.allclose(enc_seq_1, enc_seq_2, atol=1e-3))
def test_zero_mean_unit_variance_normalization_np(self):
feat_extract = self.feature_extraction_class(**self.feat_extract_tester.prepare_feat_extract_dict())
speech_inputs = [floats_list((1, x))[0] for x in range(800, 1400, 200)]

8
tests/models/wav2vec2/test_tokenization_wav2vec2.py
View File

@@ -164,6 +164,14 @@ class Wav2Vec2TokenizerTest(unittest.TestCase):
for enc_seq_1, enc_seq_2 in zip(encoded_sequences_1, encoded_sequences_2):
self.assertTrue(np.allclose(enc_seq_1, enc_seq_2, atol=1e-3))
# Test 2-D numpy arrays are batched.
speech_inputs = [floats_list((1, x))[0] for x in (800, 800, 800)]
np_speech_inputs = np.asarray(speech_inputs)
encoded_sequences_1 = tokenizer(speech_inputs, return_tensors="np").input_values
encoded_sequences_2 = tokenizer(np_speech_inputs, return_tensors="np").input_values
for enc_seq_1, enc_seq_2 in zip(encoded_sequences_1, encoded_sequences_2):
self.assertTrue(np.allclose(enc_seq_1, enc_seq_2, atol=1e-3))
def test_padding(self, max_length=50):
def _input_values_have_equal_length(input_values):
length = len(input_values[0])