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

[Unispeech] Fix slow tests (#15818)

Browse Source 
* remove soundfile old way of loading audio

* Adapt slow test
This commit is contained in:
Patrick von Platen
2022-02-24 19:08:54 +01:00
committed by GitHub
parent 35ecf99cc4
commit ca57b45071
2 changed files with 16 additions and 32 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
tests/unispeech/test_modeling_unispeech.py
View File

@@ -538,21 +538,13 @@ class UniSpeechRobustModelTest(ModelTesterMixin, unittest.TestCase):
@slow
class UniSpeechModelIntegrationTest(unittest.TestCase):
def _load_datasamples(self, num_samples):
import soundfile as sf
ds = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation")
# automatic decoding with librispeech
speech_samples = ds.sort("id").filter(
lambda x: x["id"] in [f"1272-141231-000{i}" for i in range(num_samples)]
)[:num_samples]["audio"]
ids = [f"1272-141231-000{i}" for i in range(num_samples)]
# map files to raw
def map_to_array(batch):
speech, _ = sf.read(batch["file"])
batch["speech"] = speech
return batch
ds = load_dataset("patrickvonplaten/librispeech_asr_dummy", "clean", split="validation")
ds = ds.filter(lambda x: x["id"] in ids).sort("id").map(map_to_array)
return ds["speech"][:num_samples]
return [x["array"] for x in speech_samples]
def _load_superb(self, task, num_samples):

28
tests/unispeech_sat/test_modeling_unispeech_sat.py
View File

@@ -800,21 +800,13 @@ class UniSpeechSatRobustModelTest(ModelTesterMixin, unittest.TestCase):
@slow
class UniSpeechSatModelIntegrationTest(unittest.TestCase):
def _load_datasamples(self, num_samples):
import soundfile as sf
ds = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation")
# automatic decoding with librispeech
speech_samples = ds.sort("id").filter(
lambda x: x["id"] in [f"1272-141231-000{i}" for i in range(num_samples)]
)[:num_samples]["audio"]
ids = [f"1272-141231-000{i}" for i in range(num_samples)]
# map files to raw
def map_to_array(batch):
speech, _ = sf.read(batch["file"])
batch["speech"] = speech
return batch
ds = load_dataset("patrickvonplaten/librispeech_asr_dummy", "clean", split="validation")
ds = ds.filter(lambda x: x["id"] in ids).sort("id").map(map_to_array)
return ds["speech"][:num_samples]
return [x["array"] for x in speech_samples]
def _load_superb(self, task, num_samples):
ds = load_dataset("anton-l/superb_dummy", task, split="test")
@@ -865,10 +857,10 @@ class UniSpeechSatModelIntegrationTest(unittest.TestCase):
# fmt: off
expected_hidden_states_slice = torch.tensor(
[[[-0.1172, -0.0797],
[-0.0012, 0.0213]],
[[-0.1225, -0.1277],
[-0.0668, -0.0585]]],
[[[-0.1192, -0.0825],
[-0.0012, 0.0235]],
[[-0.1240, -0.1332],
[-0.0658, -0.0565]]],
device=torch_device,
)
# fmt: on