transformers

1
# coding=utf-8
2
# Copyright 2023 HuggingFace Inc.
3
#
4
# Licensed under the Apache License, Version 2.0 (the "License");
5
# you may not use this file except in compliance with the License.
6
# You may obtain a copy of the License at
7
#
8
#     http://www.apache.org/licenses/LICENSE-2.0
9
#
10
# Unless required by applicable law or agreed to in writing, software
11
# distributed under the License is distributed on an "AS IS" BASIS,
12
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13
# See the License for the specific language governing permissions and
14
# limitations under the License.
15

16

17
import itertools
18
import random
19
import unittest
20

21
import numpy as np
22
from datasets import load_dataset
23

24
from transformers import ClapFeatureExtractor
25
from transformers.testing_utils import require_torch, require_torchaudio
26
from transformers.trainer_utils import set_seed
27
from transformers.utils.import_utils import is_torch_available
28

29
from ...test_sequence_feature_extraction_common import SequenceFeatureExtractionTestMixin
30

31

32
if is_torch_available():
33
    import torch
34

35
global_rng = random.Random()
36

37

38
# Copied from tests.models.whisper.test_feature_extraction_whisper.floats_list
39
def floats_list(shape, scale=1.0, rng=None, name=None):
40
    """Creates a random float32 tensor"""
41
    if rng is None:
42
        rng = global_rng
43

44
    values = []
45
    for batch_idx in range(shape[0]):
46
        values.append([])
47
        for _ in range(shape[1]):
48
            values[-1].append(rng.random() * scale)
49

50
    return values
51

52

53
@require_torch
54
@require_torchaudio
55
# Copied from tests.models.whisper.test_feature_extraction_whisper.WhisperFeatureExtractionTester with Whisper->Clap
56
class ClapFeatureExtractionTester(unittest.TestCase):
57
    def __init__(
58
        self,
59
        parent,
60
        batch_size=7,
61
        min_seq_length=400,
62
        max_seq_length=2000,
63
        feature_size=10,
64
        hop_length=160,
65
        chunk_length=8,
66
        padding_value=0.0,
67
        sampling_rate=4_000,
68
        return_attention_mask=False,
69
        do_normalize=True,
70
    ):
71
        self.parent = parent
72
        self.batch_size = batch_size
73
        self.min_seq_length = min_seq_length
74
        self.max_seq_length = max_seq_length
75
        self.seq_length_diff = (self.max_seq_length - self.min_seq_length) // (self.batch_size - 1)
76
        self.padding_value = padding_value
77
        self.sampling_rate = sampling_rate
78
        self.return_attention_mask = return_attention_mask
79
        self.do_normalize = do_normalize
80
        self.feature_size = feature_size
81
        self.chunk_length = chunk_length
82
        self.hop_length = hop_length
83

84
    def prepare_feat_extract_dict(self):
85
        return {
86
            "feature_size": self.feature_size,
87
            "hop_length": self.hop_length,
88
            "chunk_length": self.chunk_length,
89
            "padding_value": self.padding_value,
90
            "sampling_rate": self.sampling_rate,
91
            "return_attention_mask": self.return_attention_mask,
92
            "do_normalize": self.do_normalize,
93
        }
94

95
    def prepare_inputs_for_common(self, equal_length=False, numpify=False):
96
        def _flatten(list_of_lists):
97
            return list(itertools.chain(*list_of_lists))
98

99
        if equal_length:
100
            speech_inputs = [floats_list((self.max_seq_length, self.feature_size)) for _ in range(self.batch_size)]
101
        else:
102
            # make sure that inputs increase in size
103
            speech_inputs = [
104
                floats_list((x, self.feature_size))
105
                for x in range(self.min_seq_length, self.max_seq_length, self.seq_length_diff)
106
            ]
107
        if numpify:
108
            speech_inputs = [np.asarray(x) for x in speech_inputs]
109
        return speech_inputs
110

111

112
@require_torch
113
@require_torchaudio
114
class ClapFeatureExtractionTest(SequenceFeatureExtractionTestMixin, unittest.TestCase):
115
    feature_extraction_class = ClapFeatureExtractor
116

117
    # Copied from tests.models.whisper.test_feature_extraction_whisper.WhisperFeatureExtractionTest.setUp with Whisper->Clap
118
    def setUp(self):
119
        self.feat_extract_tester = ClapFeatureExtractionTester(self)
120

121
    def test_call(self):
122
        # Tests that all call wrap to encode_plus and batch_encode_plus
123
        feature_extractor = self.feature_extraction_class(**self.feat_extract_tester.prepare_feat_extract_dict())
124
        # create three inputs of length 800, 1000, and 1200
125
        speech_inputs = [floats_list((1, x))[0] for x in range(800, 1400, 200)]
126
        np_speech_inputs = [np.asarray(speech_input) for speech_input in speech_inputs]
127

128
        # Test feature size
129
        input_features = feature_extractor(np_speech_inputs, padding="max_length", return_tensors="np").input_features
130
        self.assertTrue(input_features.ndim == 4)
131

132
        # Test not batched input
133
        encoded_sequences_1 = feature_extractor(speech_inputs[0], return_tensors="np").input_features
134
        encoded_sequences_2 = feature_extractor(np_speech_inputs[0], return_tensors="np").input_features
135
        self.assertTrue(np.allclose(encoded_sequences_1, encoded_sequences_2, atol=1e-3))
136

137
        # Test batched
138
        encoded_sequences_1 = feature_extractor(speech_inputs, return_tensors="np").input_features
139
        encoded_sequences_2 = feature_extractor(np_speech_inputs, return_tensors="np").input_features
140
        for enc_seq_1, enc_seq_2 in zip(encoded_sequences_1, encoded_sequences_2):
141
            self.assertTrue(np.allclose(enc_seq_1, enc_seq_2, atol=1e-3))
142

143
        # Test 2-D numpy arrays are batched.
144
        speech_inputs = [floats_list((1, x))[0] for x in (800, 800, 800)]
145
        np_speech_inputs = np.asarray(speech_inputs)
146
        encoded_sequences_1 = feature_extractor(speech_inputs, return_tensors="np").input_features
147
        encoded_sequences_2 = feature_extractor(np_speech_inputs, return_tensors="np").input_features
148
        for enc_seq_1, enc_seq_2 in zip(encoded_sequences_1, encoded_sequences_2):
149
            self.assertTrue(np.allclose(enc_seq_1, enc_seq_2, atol=1e-3))
150

151
    # Copied from tests.models.whisper.test_feature_extraction_whisper.WhisperFeatureExtractionTest.test_double_precision_pad
152
    def test_double_precision_pad(self):
153
        import torch
154

155
        feature_extractor = self.feature_extraction_class(**self.feat_extract_tester.prepare_feat_extract_dict())
156
        np_speech_inputs = np.random.rand(100, 32).astype(np.float64)
157
        py_speech_inputs = np_speech_inputs.tolist()
158

159
        for inputs in [py_speech_inputs, np_speech_inputs]:
160
            np_processed = feature_extractor.pad([{"input_features": inputs}], return_tensors="np")
161
            self.assertTrue(np_processed.input_features.dtype == np.float32)
162
            pt_processed = feature_extractor.pad([{"input_features": inputs}], return_tensors="pt")
163
            self.assertTrue(pt_processed.input_features.dtype == torch.float32)
164

165
    # Copied from tests.models.whisper.test_feature_extraction_whisper.WhisperFeatureExtractionTest._load_datasamples
166
    def _load_datasamples(self, num_samples):
167
        ds = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation")
168
        # automatic decoding with librispeech
169
        speech_samples = ds.sort("id").select(range(num_samples))[:num_samples]["audio"]
170

171
        return [x["array"] for x in speech_samples]
172

173
    def test_integration_fusion_short_input(self):
174
        # fmt: off
175
        EXPECTED_INPUT_FEATURES = torch.tensor(
176
            [
177
                [
178
                    # "repeat"
179
                    [
180
                        -20.1049, -19.9764, -20.0731, -19.5055, -27.5018, -22.5761, -26.6071,
181
                        -29.0091, -26.4659, -26.4236, -28.8808, -31.9190, -32.4848, -34.1186,
182
                        -34.0340, -32.8803, -30.9895, -37.6238, -38.0347, -40.6263, -36.3496,
183
                        -42.2533, -32.9132, -27.7068, -29.3704, -30.3208, -22.5972, -27.1494,
184
                        -30.1975, -31.1005, -29.9372, -27.1917, -25.9806, -30.3489, -33.2380,
185
                        -31.9062, -36.5498, -32.8721, -30.5629, -27.4674, -22.2232, -22.5653,
186
                        -16.3868, -17.2713, -25.9738, -30.6256, -34.3766, -31.1292, -27.8950,
187
                        -27.0588, -25.6206, -23.0712, -26.6050, -28.0112, -32.6847, -34.3396,
188
                        -34.9738, -35.8463, -39.2324, -37.1188, -33.3705, -28.9230, -28.9112,
189
                        -28.6578
190
                    ],
191
                    [
192
                        -36.7233, -30.0587, -24.8431, -18.4611, -16.8149, -23.9319, -32.8580,
193
                        -34.2264, -27.4332, -26.8027, -29.2721, -33.9033, -39.3403, -35.3232,
194
                        -26.8076, -28.6460, -35.2780, -36.0738, -35.4996, -37.7631, -39.5056,
195
                        -34.7112, -36.8741, -34.1066, -32.9474, -33.6604, -27.9937, -30.9594,
196
                        -26.2928, -32.0485, -29.2151, -29.2917, -32.7308, -29.6542, -31.1454,
197
                        -37.0088, -32.3388, -37.3086, -31.1024, -27.2889, -19.6788, -21.1488,
198
                        -19.5144, -14.8889, -21.2006, -24.7488, -27.7940, -31.1058, -27.5068,
199
                        -21.5737, -22.3780, -21.5151, -26.3086, -30.9223, -33.5043, -32.0307,
200
                        -37.3806, -41.6188, -45.6650, -40.5131, -32.5023, -26.7385, -26.3709,
201
                        -26.7761
202
                    ]
203
                ],
204
                [
205
                    # "repeatpad"
206
                    [
207
                        -25.7496, -24.9339, -24.1357, -23.1271, -23.7853, -26.1264, -29.1456,
208
                        -33.2060, -37.8179, -42.4833, -41.9386, -41.2164, -42.3566, -44.2575,
209
                        -40.0217, -36.6794, -36.6974, -38.7819, -42.0880, -45.5560, -39.9368,
210
                        -36.3219, -35.5981, -36.6434, -35.1851, -33.0684, -30.0437, -30.2010,
211
                        -34.3476, -42.1373, -38.8039, -37.3355, -40.4576, -41.0485, -40.6377,
212
                        -38.2275, -42.7481, -34.6084, -34.7048, -29.5149, -26.3935, -26.8952,
213
                        -34.1336, -26.2904, -28.2571, -32.5642, -36.7240, -35.5334, -38.2451,
214
                        -34.8177, -28.9754, -25.1096, -27.9768, -32.3184, -37.0269, -40.5136,
215
                        -40.8061, -36.4948, -40.3767, -38.9671, -38.3552, -34.1250, -30.9035,
216
                        -31.6112
217
                    ],
218
                    [
219
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
220
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
221
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
222
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
223
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
224
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
225
                        -100., -100., -100., -100.
226
                    ]
227
                ],
228
                [
229
                    # None, same as "repeatpad"
230
                    [
231
                        -25.7496, -24.9339, -24.1357, -23.1271, -23.7853, -26.1264, -29.1456,
232
                        -33.2060, -37.8179, -42.4833, -41.9386, -41.2164, -42.3566, -44.2575,
233
                        -40.0217, -36.6794, -36.6974, -38.7819, -42.0880, -45.5560, -39.9368,
234
                        -36.3219, -35.5981, -36.6434, -35.1851, -33.0684, -30.0437, -30.2010,
235
                        -34.3476, -42.1373, -38.8039, -37.3355, -40.4576, -41.0485, -40.6377,
236
                        -38.2275, -42.7481, -34.6084, -34.7048, -29.5149, -26.3935, -26.8952,
237
                        -34.1336, -26.2904, -28.2571, -32.5642, -36.7240, -35.5334, -38.2451,
238
                        -34.8177, -28.9754, -25.1096, -27.9768, -32.3184, -37.0269, -40.5136,
239
                        -40.8061, -36.4948, -40.3767, -38.9671, -38.3552, -34.1250, -30.9035,
240
                        -31.6112
241
                    ],
242
                    [
243
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
244
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
245
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
246
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
247
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
248
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
249
                        -100., -100., -100., -100.
250
                    ]
251
                ],
252
                [
253
                    # "pad"
254
                    [
255
                        -58.5260, -58.1155, -57.8623, -57.5059, -57.9178, -58.7171, -59.2343,
256
                        -59.9833, -60.9764, -62.0722, -63.5723, -65.7111, -67.5153, -68.7088,
257
                        -69.8325, -70.2987, -70.1548, -70.6233, -71.5702, -72.5159, -72.3821,
258
                        -70.1817, -67.0315, -64.1387, -62.2202, -61.0717, -60.4951, -61.6005,
259
                        -63.7358, -67.1400, -67.6185, -65.5635, -64.3593, -63.7138, -63.6209,
260
                        -66.4950, -72.6284, -63.3961, -56.8334, -52.7319, -50.6310, -51.3728,
261
                        -53.5619, -51.9190, -50.9708, -52.8684, -55.8073, -58.8227, -60.6991,
262
                        -57.0547, -52.7611, -51.4388, -54.4892, -60.8950, -66.1024, -72.4352,
263
                        -67.8538, -65.1463, -68.7588, -72.3080, -68.4864, -60.4688, -57.1516,
264
                        -60.9460
265
                    ],
266
                    [
267
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
268
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
269
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
270
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
271
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
272
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
273
                        -100., -100., -100., -100.
274
                    ]
275
                ]
276
            ]
277
        )
278
        # fmt: on
279
        MEL_BIN = [[976, 977], [976, 977], [976, 977], [196, 197]]
280
        input_speech = self._load_datasamples(1)
281
        feature_extractor = ClapFeatureExtractor()
282
        for padding, EXPECTED_VALUES, idx_in_mel in zip(
283
            ["repeat", "repeatpad", None, "pad"], EXPECTED_INPUT_FEATURES, MEL_BIN
284
        ):
285
            input_features = feature_extractor(input_speech, return_tensors="pt", padding=padding).input_features
286
            self.assertEqual(input_features.shape, (1, 4, 1001, 64))
287

288
            self.assertTrue(torch.allclose(input_features[0, 0, idx_in_mel[0]], EXPECTED_VALUES[0], atol=1e-4))
289
            self.assertTrue(torch.allclose(input_features[0, 0, idx_in_mel[1]], EXPECTED_VALUES[1], atol=1e-4))
290

291
            self.assertTrue(torch.all(input_features[0, 0] == input_features[0, 1]))
292
            self.assertTrue(torch.all(input_features[0, 0] == input_features[0, 2]))
293
            self.assertTrue(torch.all(input_features[0, 0] == input_features[0, 3]))
294

295
    def test_integration_rand_trunc_short_input(self):
296
        # fmt: off
297
        EXPECTED_INPUT_FEATURES = torch.tensor(
298
            [
299
                [
300
                    # "repeat"
301
                    [
302
                        -35.0483, -35.7865, -38.2884, -40.0220, -42.5349, -44.9489, -43.2228,
303
                        -44.6499, -47.6253, -49.6983, -50.2127, -52.5483, -52.2223, -51.9157,
304
                        -49.4082, -51.2024, -57.0476, -56.2803, -58.1618, -60.7474, -55.0389,
305
                        -60.9514, -59.3080, -50.4419, -47.8172, -48.7570, -55.2552, -44.5036,
306
                        -44.1148, -50.8218, -51.0968, -52.9408, -51.1037, -48.9789, -47.5897,
307
                        -52.0915, -55.4216, -54.1529, -58.0149, -58.0866, -52.7798, -52.6154,
308
                        -45.9144, -46.2008, -40.7603, -41.1703, -50.2250, -55.4112, -59.4818,
309
                        -54.5795, -53.5552, -51.3668, -49.8358, -50.3186, -54.0452, -57.6030,
310
                        -61.1589, -61.6415, -63.2756, -66.5890, -62.8543, -58.0665, -56.7203,
311
                        -56.7632
312
                    ],
313
                    [
314
                        -47.1320, -37.9961, -34.0076, -36.7109, -47.9057, -48.4924, -43.8371,
315
                        -44.9728, -48.1689, -52.9141, -57.6077, -52.8520, -44.8502, -45.6764,
316
                        -51.8389, -56.4284, -54.6972, -53.4889, -55.6077, -58.7149, -60.3760,
317
                        -54.0136, -56.0730, -55.9870, -54.4017, -53.1094, -53.5640, -50.3064,
318
                        -49.9520, -49.3239, -48.1668, -53.4852, -50.4561, -50.8688, -55.1970,
319
                        -51.5538, -53.0260, -59.6933, -54.8183, -59.5895, -55.9589, -50.3761,
320
                        -44.1282, -44.1463, -43.8540, -39.1168, -45.3893, -49.5542, -53.1505,
321
                        -55.2870, -50.3921, -46.8511, -47.4444, -49.5633, -56.0034, -59.0815,
322
                        -59.0018, -63.7589, -69.5745, -71.5789, -64.0498, -56.0558, -54.3475,
323
                        -54.7004
324
                    ]
325
                ],
326
                [
327
                    # "repeatpad"
328
                    [
329
                        -40.3184, -39.7186, -39.8807, -41.6508, -45.3613, -50.4785, -57.0297,
330
                        -60.4944, -59.1642, -58.9495, -60.4661, -62.5300, -58.4759, -55.2865,
331
                        -54.8973, -56.0780, -57.5482, -59.6557, -64.3309, -65.0330, -59.4941,
332
                        -56.8552, -55.0519, -55.9817, -56.9739, -55.2827, -54.5312, -51.4141,
333
                        -50.4289, -51.9131, -57.5821, -63.9979, -59.9180, -58.9489, -62.3247,
334
                        -62.6975, -63.7948, -60.5250, -64.6107, -58.7905, -57.0229, -54.3084,
335
                        -49.8445, -50.4459, -57.0172, -50.6425, -52.5992, -57.4207, -61.6358,
336
                        -60.6540, -63.1968, -57.4360, -52.3263, -51.7695, -57.1946, -62.9610,
337
                        -66.7359, -67.0335, -63.7440, -68.1775, -66.3798, -62.8650, -59.8972,
338
                        -59.3139
339
                    ],
340
                    [
341
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
342
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
343
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
344
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
345
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
346
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
347
                        -100., -100., -100., -100.
348
                    ]
349
                ],
350
                [
351
                    # None, same as "repeatpad"
352
                    [
353
                        -40.3184, -39.7186, -39.8807, -41.6508, -45.3613, -50.4785, -57.0297,
354
                        -60.4944, -59.1642, -58.9495, -60.4661, -62.5300, -58.4759, -55.2865,
355
                        -54.8973, -56.0780, -57.5482, -59.6557, -64.3309, -65.0330, -59.4941,
356
                        -56.8552, -55.0519, -55.9817, -56.9739, -55.2827, -54.5312, -51.4141,
357
                        -50.4289, -51.9131, -57.5821, -63.9979, -59.9180, -58.9489, -62.3247,
358
                        -62.6975, -63.7948, -60.5250, -64.6107, -58.7905, -57.0229, -54.3084,
359
                        -49.8445, -50.4459, -57.0172, -50.6425, -52.5992, -57.4207, -61.6358,
360
                        -60.6540, -63.1968, -57.4360, -52.3263, -51.7695, -57.1946, -62.9610,
361
                        -66.7359, -67.0335, -63.7440, -68.1775, -66.3798, -62.8650, -59.8972,
362
                        -59.3139
363
                    ],
364
                    [
365
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
366
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
367
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
368
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
369
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
370
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
371
                        -100., -100., -100., -100.
372
                    ]
373
                ],
374
                [
375
                    # "pad"
376
                    [
377
                        -73.3190, -73.6349, -74.1451, -74.8539, -75.7476, -76.5438, -78.5540,
378
                        -80.1339, -81.8911, -83.7560, -85.5387, -86.7466, -88.2072, -88.6090,
379
                        -88.8243, -89.0784, -89.4364, -89.8179, -91.3146, -92.2833, -91.7221,
380
                        -90.9440, -88.1315, -86.2425, -84.2281, -82.4893, -81.5993, -81.1328,
381
                        -81.5759, -83.1068, -85.6525, -88.9520, -88.9187, -87.2703, -86.3052,
382
                        -85.7188, -85.8802, -87.9996, -95.0464, -88.0133, -80.8561, -76.5597,
383
                        -74.2816, -74.8109, -77.3615, -76.0719, -75.3426, -77.6428, -80.9663,
384
                        -84.5275, -84.9907, -80.5205, -77.2851, -78.6259, -84.7740, -91.4535,
385
                        -98.1894, -94.3872, -92.3735, -97.6807, -98.1501, -91.4344, -85.2842,
386
                        -88.4338
387
                    ],
388
                    [
389
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
390
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
391
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
392
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
393
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
394
                        -100., -100., -100., -100., -100., -100., -100., -100., -100., -100.,
395
                        -100., -100., -100., -100.
396
                    ]
397
                ]
398
            ]
399
        )
400
        # fmt: on
401
        MEL_BIN = [[976, 977], [976, 977], [976, 977], [196, 197]]
402
        input_speech = self._load_datasamples(1)
403
        feature_extractor = ClapFeatureExtractor()
404
        for padding, EXPECTED_VALUES, idx_in_mel in zip(
405
            ["repeat", "repeatpad", None, "pad"], EXPECTED_INPUT_FEATURES, MEL_BIN
406
        ):
407
            input_features = feature_extractor(
408
                input_speech, return_tensors="pt", truncation="rand_trunc", padding=padding
409
            ).input_features
410
            self.assertEqual(input_features.shape, (1, 1, 1001, 64))
411
            self.assertTrue(torch.allclose(input_features[0, 0, idx_in_mel[0]], EXPECTED_VALUES[0], atol=1e-4))
412
            self.assertTrue(torch.allclose(input_features[0, 0, idx_in_mel[1]], EXPECTED_VALUES[1], atol=1e-4))
413

414
    def test_integration_fusion_long_input(self):
415
        # fmt: off
416
        EXPECTED_INPUT_FEATURES = torch.tensor(
417
            [
418
                [
419
                    -11.1830, -10.1894, -8.6051, -4.8578, -1.3268, -8.4606, -14.5453,
420
                     -9.2017, 0.5781, 16.2129, 14.8289, 3.6326, -3.8794, -6.5544,
421
                     -2.4408, 1.9531, 6.0967, 1.7590, -7.6730, -6.1571, 2.0052,
422
                     16.6694, 20.6447, 21.2145, 13.4972, 15.9043, 16.8987, 4.1766,
423
                     11.9428, 21.2372, 12.3016, 4.8604, 6.7241, 1.8543, 4.9235,
424
                      5.3188, -0.9897, -1.2416, -6.5864, 2.9529, 2.9274, 6.4753,
425
                     10.2300, 11.2127, 3.4042, -1.0055, -6.0475, -6.7524, -3.9801,
426
                     -1.4434, 0.4740, -0.1584, -4.5457, -8.5746, -8.8428, -13.1475,
427
                     -9.6079, -8.5798, -4.1143, -3.7966, -7.1651, -6.1517, -8.0258,
428
                    -12.1486
429
                ],
430
                [
431
                    -10.2017, -7.9924, -5.9517, -3.9372, -1.9735, -4.3130, 16.1647,
432
                     25.0592, 23.5532, 14.4974, -7.0778, -10.2262, 6.4782, 20.3454,
433
                     19.4269, 1.7976, -16.5070, 4.9380, 12.3390, 6.9285, -13.6325,
434
                     -8.5298, 1.0839, -5.9629, -8.4812, 3.1331, -2.0963, -16.6046,
435
                    -14.0070, -17.5707, -13.2080, -17.2168, -17.7770, -12.1111, -18.6184,
436
                    -17.1897, -13.9801, -12.0426, -23.5400, -25.6823, -23.5813, -18.7847,
437
                    -20.5473, -25.6458, -19.7585, -27.6007, -28.9276, -24.8948, -25.4458,
438
                    -22.2807, -19.6613, -19.2669, -15.7813, -19.6821, -24.3439, -22.2598,
439
                    -28.2631, -30.1017, -32.7646, -33.6525, -27.5639, -22.0548, -27.8054,
440
                    -29.6947
441
                ],
442
                [
443
                    -9.2078, -7.2963, -6.2095, -7.9959, -2.9280, -11.1843, -6.1490,
444
                    5.0733, 19.2957, 21.4578, 14.6803, -3.3153, -6.3334, -2.3542,
445
                    6.9509, 15.2965, 14.6620, 5.2075, -0.0873, 1.1919, 18.1986,
446
                    20.8470, 10.8035, 2.2516, 7.6905, 7.7427, -1.2543, -5.0018,
447
                    0.9809, -2.1584, -5.4580, -5.4760, -11.8888, -9.0605, -8.4638,
448
                    -9.9897, -0.0540, -5.1629, 0.0483, -4.1504, -4.8140, -7.8236,
449
                    -9.0622, -10.1742, -8.9597, -11.5380, -16.5603, -17.1858, -17.5032,
450
                    -20.9326, -23.9543, -25.2602, -25.3429, -27.4536, -26.8859, -22.7852,
451
                    -25.8288, -24.8399, -23.8893, -24.2096, -26.5415, -23.7281, -25.6851,
452
                    -22.3629
453
                ],
454
                [
455
                      1.3448, 2.9883, 4.0366, -0.8019, -10.4191, -10.0883, -4.3812,
456
                      0.8136, 2.1579, 0.0832, 1.0949, -0.9759, -5.5319, -4.6009,
457
                     -6.5452, -14.9155, -20.1584, -9.3611, -2.4271, 1.4031, 4.9910,
458
                      8.6916, 8.6785, 10.1973, 9.9029, 5.3840, 7.5336, 5.2803,
459
                      2.8144, -0.3138, 2.2216, 5.7328, 7.5574, 7.7402, 1.0681,
460
                      3.1049, 7.0742, 6.5588, 7.3712, 5.7881, 8.6874, 8.7725,
461
                      2.8133, -4.5809, -6.1317, -5.1719, -5.0192, -9.0977, -10.9391,
462
                     -6.0769, 1.6016, -0.8965, -7.2252, -7.8632, -11.4468, -11.7446,
463
                    -10.7447, -7.0601, -2.7748, -4.1798, -2.8433, -3.1352, 0.8097,
464
                      6.4212
465
                ]
466
            ]
467
        )
468
        # fmt: on
469
        MEL_BIN = 963
470
        input_speech = torch.cat([torch.tensor(x) for x in self._load_datasamples(5)])
471
        feature_extractor = ClapFeatureExtractor()
472
        for padding, EXPECTED_VALUES, block_idx in zip(
473
            ["repeat", "repeatpad", None, "pad"], EXPECTED_INPUT_FEATURES, [1, 2, 0, 3]
474
        ):
475
            set_seed(987654321)
476
            input_features = feature_extractor(input_speech, return_tensors="pt", padding=padding).input_features
477
            self.assertEqual(input_features.shape, (1, 4, 1001, 64))
478
            self.assertTrue(torch.allclose(input_features[0, block_idx, MEL_BIN], EXPECTED_VALUES, atol=1e-3))
479

480
    def test_integration_rand_trunc_long_input(self):
481
        # fmt: off
482
        EXPECTED_INPUT_FEATURES = torch.tensor(
483
            [
484
                [
485
                    -35.4022, -32.7555, -31.2004, -32.7764, -42.5770, -41.6339, -43.1630,
486
                    -44.5080, -44.3029, -48.9628, -39.5022, -39.2105, -43.1350, -43.2195,
487
                    -48.4894, -52.2344, -57.6891, -52.2228, -45.5155, -44.2893, -43.4697,
488
                    -46.6702, -43.7490, -40.4819, -42.7275, -46.3434, -46.8412, -41.2003,
489
                    -43.1681, -46.2948, -46.1925, -47.8333, -45.6812, -44.9182, -41.7786,
490
                    -43.3809, -44.3199, -42.8814, -45.4771, -46.7114, -46.9746, -42.7090,
491
                    -41.6057, -38.3965, -40.1980, -41.0263, -34.1256, -28.3289, -29.0201,
492
                    -30.4453, -29.5561, -30.1734, -25.9406, -19.0897, -15.8452, -20.1351,
493
                    -23.6515, -23.1194, -17.1845, -19.4399, -23.6527, -22.8768, -20.7279,
494
                    -22.7864
495
                ],
496
                [
497
                    -35.7719, -27.2566, -23.6964, -27.5521, 0.2510, 7.4391, 1.3917,
498
                    -13.3417, -28.1758, -17.0856, -5.7723, -0.8000, -7.8832, -15.5548,
499
                    -30.5935, -24.7571, -13.7009, -10.3432, -21.2464, -24.8118, -19.4080,
500
                    -14.9779, -11.7991, -18.4485, -20.1982, -17.3652, -20.6328, -28.2967,
501
                    -25.7819, -21.8962, -28.5083, -29.5719, -30.2120, -35.7033, -31.8218,
502
                    -34.0408, -37.7744, -33.9653, -31.3009, -30.9063, -28.6153, -32.2202,
503
                    -28.5456, -28.8579, -32.5170, -37.9152, -43.0052, -46.4849, -44.0786,
504
                    -39.1933, -33.2757, -31.6313, -42.6386, -52.3679, -53.5785, -55.6444,
505
                    -47.0050, -47.6459, -56.6361, -60.6781, -61.5244, -55.8272, -60.4832,
506
                    -58.1897
507
                ],
508
                [
509
                    -38.2686, -36.6285, -32.5835, -35.1693, -37.7938, -37.4035, -35.3132,
510
                    -35.6083, -36.3609, -40.9472, -36.7846, -36.1544, -38.9076, -39.3618,
511
                    -35.4953, -34.2809, -39.9466, -39.7433, -34.8347, -37.5674, -41.5689,
512
                    -38.9161, -34.3947, -30.2924, -30.4841, -34.5831, -28.9261, -24.8849,
513
                    -31.2324, -27.1622, -27.2107, -25.9385, -30.1691, -30.9223, -23.9495,
514
                    -25.6047, -26.7119, -28.5523, -27.7481, -32.8427, -35.4650, -31.0399,
515
                    -31.2073, -30.5163, -22.9819, -20.8892, -19.2510, -24.7905, -28.9426,
516
                    -28.1998, -26.7386, -25.0140, -27.9223, -32.9913, -33.1864, -34.9742,
517
                    -38.5995, -39.6990, -29.3203, -22.4697, -25.6415, -33.5608, -33.0945,
518
                    -27.1716
519
                ],
520
                [
521
                    -33.2015, -28.7741, -21.9457, -23.4888, -32.1072, -8.6307, 3.2724,
522
                      5.9157, -0.9221, -30.1814, -31.0015, -27.4508, -27.0477, -9.5342,
523
                      0.3221, 0.6511, -7.1596, -25.9707, -32.8924, -32.2300, -13.8974,
524
                     -0.4895, 0.9168, -10.7663, -27.1176, -35.0829, -11.6859, -4.8855,
525
                    -11.8898, -26.6167, -5.6192, -3.8443, -19.7947, -14.4101, -8.6236,
526
                    -21.2458, -21.0801, -17.9136, -24.4663, -18.6333, -24.8085, -15.5854,
527
                    -15.4344, -11.5046, -22.3625, -27.3387, -32.4353, -30.9670, -31.3789,
528
                    -35.4044, -34.4591, -25.2433, -28.0773, -33.8736, -33.0224, -33.3155,
529
                    -38.5302, -39.2741, -36.6395, -34.7729, -32.4483, -42.4001, -49.2857,
530
                    -39.1682
531
                ]
532
            ]
533
        )
534
        # fmt: on
535
        MEL_BIN = 963
536
        SEEDS = [987654321, 1234, 666, 5555]
537
        input_speech = torch.cat([torch.tensor(x) for x in self._load_datasamples(5)])
538
        feature_extractor = ClapFeatureExtractor()
539
        for padding, EXPECTED_VALUES, seed in zip(
540
            ["repeat", "repeatpad", None, "pad"], EXPECTED_INPUT_FEATURES, SEEDS
541
        ):
542
            set_seed(seed)
543
            input_features = feature_extractor(
544
                input_speech, return_tensors="pt", truncation="rand_trunc", padding=padding
545
            ).input_features
546
            self.assertEqual(input_features.shape, (1, 1, 1001, 64))
547
            self.assertTrue(torch.allclose(input_features[0, 0, MEL_BIN], EXPECTED_VALUES, atol=1e-4))
548