Hugging Face's logo

nrshoudi
/
wav2vec-arabic-V2-10

Automatic Speech Recognition
Transformers
Safetensors
wav2vec2
Generated from Trainer
Inference Endpoints
Model card Files Community
nrshoudi commited on
Commit
651697d
•
1 Parent(s): f0ed31d

End of training

Browse files
Files changed (6) hide show
  1. README.md +81 -0
  2. config.json +109 -0
  3. model.safetensors +3 -0
  4. preprocessor_config.json +10 -0
  5. trainer_state.json +348 -0
  6. training_args.bin +3 -0
README.md ADDED
@@ -0,0 +1,81 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ base_model: facebook/wav2vec2-xls-r-300m
4
+ tags:
5
+ - generated_from_trainer
6
+ metrics:
7
+ - wer
8
+ model-index:
9
+ - name: wav2vec-arabic-V2-10
10
+ results: []
11
+ ---
12
+
13
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
14
+ should probably proofread and complete it, then remove this comment. -->
15
+
16
+ # wav2vec-arabic-V2-10
17
+
18
+ This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on an unknown dataset.
19
+ It achieves the following results on the evaluation set:
20
+ - Loss: 0.3686
21
+ - Wer: 0.0763
22
+ - Per: 0.0613
23
+
24
+ ## Model description
25
+
26
+ More information needed
27
+
28
+ ## Intended uses & limitations
29
+
30
+ More information needed
31
+
32
+ ## Training and evaluation data
33
+
34
+ More information needed
35
+
36
+ ## Training procedure
37
+
38
+ ### Training hyperparameters
39
+
40
+ The following hyperparameters were used during training:
41
+ - learning_rate: 0.0001
42
+ - train_batch_size: 2
43
+ - eval_batch_size: 2
44
+ - seed: 42
45
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
46
+ - lr_scheduler_type: linear
47
+ - lr_scheduler_warmup_ratio: 0.1
48
+ - num_epochs: 20
49
+
50
+ ### Training results
51
+
52
+ | Training Loss | Epoch | Step | Validation Loss | Wer | Per |
53
+ |:-------------:|:-----:|:----:|:---------------:|:------:|:------:|
54
+ | 22.0067 | 1.0 | 164 | 5.5136 | 1.0 | 1.0 |
55
+ | 3.7377 | 2.0 | 328 | 3.2749 | 1.0 | 1.0 |
56
+ | 3.2647 | 3.0 | 492 | 3.3163 | 1.0 | 1.0 |
57
+ | 3.2513 | 4.0 | 656 | 3.2319 | 1.0 | 1.0 |
58
+ | 3.2314 | 5.0 | 820 | 3.2902 | 1.0 | 1.0 |
59
+ | 3.1754 | 6.0 | 984 | 3.0334 | 1.0 | 1.0 |
60
+ | 2.8137 | 7.0 | 1148 | 2.5538 | 0.9930 | 0.9959 |
61
+ | 2.1146 | 8.0 | 1312 | 1.6577 | 0.6542 | 0.6458 |
62
+ | 1.2679 | 9.0 | 1476 | 0.8287 | 0.2089 | 0.1803 |
63
+ | 0.6981 | 10.0 | 1640 | 0.5801 | 0.1423 | 0.1229 |
64
+ | 0.4737 | 11.0 | 1804 | 0.4867 | 0.1219 | 0.1017 |
65
+ | 0.3733 | 12.0 | 1968 | 0.4482 | 0.1098 | 0.0929 |
66
+ | 0.2913 | 13.0 | 2132 | 0.4176 | 0.1092 | 0.0933 |
67
+ | 0.2358 | 14.0 | 2296 | 0.4200 | 0.1084 | 0.0915 |
68
+ | 0.2196 | 15.0 | 2460 | 0.3999 | 0.0987 | 0.0824 |
69
+ | 0.1845 | 16.0 | 2624 | 0.3993 | 0.1006 | 0.0826 |
70
+ | 0.1831 | 17.0 | 2788 | 0.3812 | 0.0810 | 0.0653 |
71
+ | 0.1629 | 18.0 | 2952 | 0.3676 | 0.0785 | 0.0636 |
72
+ | 0.1578 | 19.0 | 3116 | 0.3679 | 0.0806 | 0.0658 |
73
+ | 0.1317 | 20.0 | 3280 | 0.3686 | 0.0763 | 0.0613 |
74
+
75
+
76
+ ### Framework versions
77
+
78
+ - Transformers 4.35.2
79
+ - Pytorch 2.1.0+cu121
80
+ - Datasets 2.16.1
81
+ - Tokenizers 0.15.0
config.json ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "facebook/wav2vec2-xls-r-300m",
3
+ "activation_dropout": 0.0,
4
+ "adapter_attn_dim": null,
5
+ "adapter_kernel_size": 3,
6
+ "adapter_stride": 2,
7
+ "add_adapter": false,
8
+ "apply_spec_augment": true,
9
+ "architectures": [
10
+ "Wav2Vec2ForCTC"
11
+ ],
12
+ "attention_dropout": 0.1,
13
+ "bos_token_id": 1,
14
+ "classifier_proj_size": 256,
15
+ "codevector_dim": 768,
16
+ "contrastive_logits_temperature": 0.1,
17
+ "conv_bias": true,
18
+ "conv_dim": [
19
+ 512,
20
+ 512,
21
+ 512,
22
+ 512,
23
+ 512,
24
+ 512,
25
+ 512
26
+ ],
27
+ "conv_kernel": [
28
+ 10,
29
+ 3,
30
+ 3,
31
+ 3,
32
+ 3,
33
+ 2,
34
+ 2
35
+ ],
36
+ "conv_stride": [
37
+ 5,
38
+ 2,
39
+ 2,
40
+ 2,
41
+ 2,
42
+ 2,
43
+ 2
44
+ ],
45
+ "ctc_loss_reduction": "mean",
46
+ "ctc_zero_infinity": false,
47
+ "diversity_loss_weight": 0.1,
48
+ "do_stable_layer_norm": true,
49
+ "eos_token_id": 2,
50
+ "feat_extract_activation": "gelu",
51
+ "feat_extract_dropout": 0.0,
52
+ "feat_extract_norm": "layer",
53
+ "feat_proj_dropout": 0.1,
54
+ "feat_quantizer_dropout": 0.0,
55
+ "final_dropout": 0.0,
56
+ "gradient_checkpointing": false,
57
+ "hidden_act": "gelu",
58
+ "hidden_dropout": 0.1,
59
+ "hidden_size": 1024,
60
+ "initializer_range": 0.02,
61
+ "intermediate_size": 4096,
62
+ "layer_norm_eps": 1e-05,
63
+ "layerdrop": 0.1,
64
+ "mask_feature_length": 10,
65
+ "mask_feature_min_masks": 0,
66
+ "mask_feature_prob": 0.0,
67
+ "mask_time_length": 10,
68
+ "mask_time_min_masks": 2,
69
+ "mask_time_prob": 0.075,
70
+ "model_type": "wav2vec2",
71
+ "num_adapter_layers": 3,
72
+ "num_attention_heads": 16,
73
+ "num_codevector_groups": 2,
74
+ "num_codevectors_per_group": 320,
75
+ "num_conv_pos_embedding_groups": 16,
76
+ "num_conv_pos_embeddings": 128,
77
+ "num_feat_extract_layers": 7,
78
+ "num_hidden_layers": 24,
79
+ "num_negatives": 100,
80
+ "output_hidden_size": 1024,
81
+ "pad_token_id": 37,
82
+ "proj_codevector_dim": 768,
83
+ "tdnn_dilation": [
84
+ 1,
85
+ 2,
86
+ 3,
87
+ 1,
88
+ 1
89
+ ],
90
+ "tdnn_dim": [
91
+ 512,
92
+ 512,
93
+ 512,
94
+ 512,
95
+ 1500
96
+ ],
97
+ "tdnn_kernel": [
98
+ 5,
99
+ 3,
100
+ 3,
101
+ 1,
102
+ 1
103
+ ],
104
+ "torch_dtype": "float32",
105
+ "transformers_version": "4.35.2",
106
+ "use_weighted_layer_sum": false,
107
+ "vocab_size": 40,
108
+ "xvector_output_dim": 512
109
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c0cbe8c19f62e189a227adebd79ebed35d4cd889b2df97a4e1333718945146d1
3
+ size 1261971480
preprocessor_config.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_normalize": true,
3
+ "feature_extractor_type": "Wav2Vec2FeatureExtractor",
4
+ "feature_size": 1,
5
+ "padding_side": "right",
6
+ "padding_value": 0,
7
+ "processor_class": "Wav2Vec2Processor",
8
+ "return_attention_mask": true,
9
+ "sampling_rate": 16000
10
+ }
trainer_state.json ADDED
@@ -0,0 +1,348 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 0.36759981513023376,
3
+ "best_model_checkpoint": "nrshoudi/wav2vec-arabic-V2-10/checkpoint-2952",
4
+ "epoch": 20.0,
5
+ "eval_steps": 500,
6
+ "global_step": 3280,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 1.0,
13
+ "learning_rate": 5e-05,
14
+ "loss": 22.0067,
15
+ "step": 164
16
+ },
17
+ {
18
+ "epoch": 1.0,
19
+ "eval_loss": 5.513646125793457,
20
+ "eval_per": 1.0,
21
+ "eval_runtime": 124.4365,
22
+ "eval_samples_per_second": 6.533,
23
+ "eval_steps_per_second": 3.271,
24
+ "eval_wer": 1.0,
25
+ "step": 164
26
+ },
27
+ {
28
+ "epoch": 2.0,
29
+ "learning_rate": 0.0001,
30
+ "loss": 3.7377,
31
+ "step": 328
32
+ },
33
+ {
34
+ "epoch": 2.0,
35
+ "eval_loss": 3.274935483932495,
36
+ "eval_per": 1.0,
37
+ "eval_runtime": 123.268,
38
+ "eval_samples_per_second": 6.595,
39
+ "eval_steps_per_second": 3.302,
40
+ "eval_wer": 1.0,
41
+ "step": 328
42
+ },
43
+ {
44
+ "epoch": 3.0,
45
+ "learning_rate": 9.444444444444444e-05,
46
+ "loss": 3.2647,
47
+ "step": 492
48
+ },
49
+ {
50
+ "epoch": 3.0,
51
+ "eval_loss": 3.316337823867798,
52
+ "eval_per": 1.0,
53
+ "eval_runtime": 123.6463,
54
+ "eval_samples_per_second": 6.575,
55
+ "eval_steps_per_second": 3.292,
56
+ "eval_wer": 1.0,
57
+ "step": 492
58
+ },
59
+ {
60
+ "epoch": 4.0,
61
+ "learning_rate": 8.888888888888889e-05,
62
+ "loss": 3.2513,
63
+ "step": 656
64
+ },
65
+ {
66
+ "epoch": 4.0,
67
+ "eval_loss": 3.2318897247314453,
68
+ "eval_per": 1.0,
69
+ "eval_runtime": 123.6244,
70
+ "eval_samples_per_second": 6.576,
71
+ "eval_steps_per_second": 3.292,
72
+ "eval_wer": 1.0,
73
+ "step": 656
74
+ },
75
+ {
76
+ "epoch": 5.0,
77
+ "learning_rate": 8.333333333333334e-05,
78
+ "loss": 3.2314,
79
+ "step": 820
80
+ },
81
+ {
82
+ "epoch": 5.0,
83
+ "eval_loss": 3.290160894393921,
84
+ "eval_per": 1.0,
85
+ "eval_runtime": 123.9537,
86
+ "eval_samples_per_second": 6.559,
87
+ "eval_steps_per_second": 3.283,
88
+ "eval_wer": 1.0,
89
+ "step": 820
90
+ },
91
+ {
92
+ "epoch": 6.0,
93
+ "learning_rate": 7.777777777777778e-05,
94
+ "loss": 3.1754,
95
+ "step": 984
96
+ },
97
+ {
98
+ "epoch": 6.0,
99
+ "eval_loss": 3.0333683490753174,
100
+ "eval_per": 1.0,
101
+ "eval_runtime": 125.221,
102
+ "eval_samples_per_second": 6.493,
103
+ "eval_steps_per_second": 3.25,
104
+ "eval_wer": 1.0,
105
+ "step": 984
106
+ },
107
+ {
108
+ "epoch": 7.0,
109
+ "learning_rate": 7.222222222222222e-05,
110
+ "loss": 2.8137,
111
+ "step": 1148
112
+ },
113
+ {
114
+ "epoch": 7.0,
115
+ "eval_loss": 2.5538418292999268,
116
+ "eval_per": 0.9959248091962687,
117
+ "eval_runtime": 125.8623,
118
+ "eval_samples_per_second": 6.459,
119
+ "eval_steps_per_second": 3.234,
120
+ "eval_wer": 0.9930474546218095,
121
+ "step": 1148
122
+ },
123
+ {
124
+ "epoch": 8.0,
125
+ "learning_rate": 6.666666666666667e-05,
126
+ "loss": 2.1146,
127
+ "step": 1312
128
+ },
129
+ {
130
+ "epoch": 8.0,
131
+ "eval_loss": 1.657683253288269,
132
+ "eval_per": 0.6457881843022708,
133
+ "eval_runtime": 126.4952,
134
+ "eval_samples_per_second": 6.427,
135
+ "eval_steps_per_second": 3.218,
136
+ "eval_wer": 0.6541925248471839,
137
+ "step": 1312
138
+ },
139
+ {
140
+ "epoch": 9.0,
141
+ "learning_rate": 6.111111111111112e-05,
142
+ "loss": 1.2679,
143
+ "step": 1476
144
+ },
145
+ {
146
+ "epoch": 9.0,
147
+ "eval_loss": 0.8286688923835754,
148
+ "eval_per": 0.18025063601243757,
149
+ "eval_runtime": 126.0524,
150
+ "eval_samples_per_second": 6.45,
151
+ "eval_steps_per_second": 3.229,
152
+ "eval_wer": 0.20890299099435397,
153
+ "step": 1476
154
+ },
155
+ {
156
+ "epoch": 10.0,
157
+ "learning_rate": 5.555555555555556e-05,
158
+ "loss": 0.6981,
159
+ "step": 1640
160
+ },
161
+ {
162
+ "epoch": 10.0,
163
+ "eval_loss": 0.580149233341217,
164
+ "eval_per": 0.12289173654951474,
165
+ "eval_runtime": 125.8066,
166
+ "eval_samples_per_second": 6.462,
167
+ "eval_steps_per_second": 3.235,
168
+ "eval_wer": 0.14227054267183054,
169
+ "step": 1640
170
+ },
171
+ {
172
+ "epoch": 11.0,
173
+ "learning_rate": 5e-05,
174
+ "loss": 0.4737,
175
+ "step": 1804
176
+ },
177
+ {
178
+ "epoch": 11.0,
179
+ "eval_loss": 0.48674020171165466,
180
+ "eval_per": 0.10169132196362951,
181
+ "eval_runtime": 126.8326,
182
+ "eval_samples_per_second": 6.41,
183
+ "eval_steps_per_second": 3.209,
184
+ "eval_wer": 0.1219261816994074,
185
+ "step": 1804
186
+ },
187
+ {
188
+ "epoch": 12.0,
189
+ "learning_rate": 4.4444444444444447e-05,
190
+ "loss": 0.3733,
191
+ "step": 1968
192
+ },
193
+ {
194
+ "epoch": 12.0,
195
+ "eval_loss": 0.4481865465641022,
196
+ "eval_per": 0.09288137190238387,
197
+ "eval_runtime": 125.9688,
198
+ "eval_samples_per_second": 6.454,
199
+ "eval_steps_per_second": 3.231,
200
+ "eval_wer": 0.10979422332135691,
201
+ "step": 1968
202
+ },
203
+ {
204
+ "epoch": 13.0,
205
+ "learning_rate": 3.888888888888889e-05,
206
+ "loss": 0.2913,
207
+ "step": 2132
208
+ },
209
+ {
210
+ "epoch": 13.0,
211
+ "eval_loss": 0.4176279902458191,
212
+ "eval_per": 0.09330538019410158,
213
+ "eval_runtime": 125.8355,
214
+ "eval_samples_per_second": 6.461,
215
+ "eval_steps_per_second": 3.234,
216
+ "eval_wer": 0.10918762540245439,
217
+ "step": 2132
218
+ },
219
+ {
220
+ "epoch": 14.0,
221
+ "learning_rate": 3.3333333333333335e-05,
222
+ "loss": 0.2358,
223
+ "step": 2296
224
+ },
225
+ {
226
+ "epoch": 14.0,
227
+ "eval_loss": 0.41998350620269775,
228
+ "eval_per": 0.0915151229624046,
229
+ "eval_runtime": 126.5555,
230
+ "eval_samples_per_second": 6.424,
231
+ "eval_steps_per_second": 3.216,
232
+ "eval_wer": 0.10844104334842052,
233
+ "step": 2296
234
+ },
235
+ {
236
+ "epoch": 15.0,
237
+ "learning_rate": 2.777777777777778e-05,
238
+ "loss": 0.2196,
239
+ "step": 2460
240
+ },
241
+ {
242
+ "epoch": 15.0,
243
+ "eval_loss": 0.399908185005188,
244
+ "eval_per": 0.08235183265806087,
245
+ "eval_runtime": 126.1327,
246
+ "eval_samples_per_second": 6.446,
247
+ "eval_steps_per_second": 3.227,
248
+ "eval_wer": 0.09868881526760301,
249
+ "step": 2460
250
+ },
251
+ {
252
+ "epoch": 16.0,
253
+ "learning_rate": 2.2222222222222223e-05,
254
+ "loss": 0.1845,
255
+ "step": 2624
256
+ },
257
+ {
258
+ "epoch": 16.0,
259
+ "eval_loss": 0.39929312467575073,
260
+ "eval_per": 0.08256383680391972,
261
+ "eval_runtime": 126.7957,
262
+ "eval_samples_per_second": 6.412,
263
+ "eval_steps_per_second": 3.21,
264
+ "eval_wer": 0.10060193178106482,
265
+ "step": 2624
266
+ },
267
+ {
268
+ "epoch": 17.0,
269
+ "learning_rate": 1.6666666666666667e-05,
270
+ "loss": 0.1831,
271
+ "step": 2788
272
+ },
273
+ {
274
+ "epoch": 17.0,
275
+ "eval_loss": 0.3811805844306946,
276
+ "eval_per": 0.0653443889569396,
277
+ "eval_runtime": 126.0916,
278
+ "eval_samples_per_second": 6.448,
279
+ "eval_steps_per_second": 3.228,
280
+ "eval_wer": 0.08100415286267557,
281
+ "step": 2788
282
+ },
283
+ {
284
+ "epoch": 18.0,
285
+ "learning_rate": 1.1111111111111112e-05,
286
+ "loss": 0.1629,
287
+ "step": 2952
288
+ },
289
+ {
290
+ "epoch": 18.0,
291
+ "eval_loss": 0.36759981513023376,
292
+ "eval_per": 0.06357768774144916,
293
+ "eval_runtime": 126.4537,
294
+ "eval_samples_per_second": 6.429,
295
+ "eval_steps_per_second": 3.219,
296
+ "eval_wer": 0.07848443843031123,
297
+ "step": 2952
298
+ },
299
+ {
300
+ "epoch": 19.0,
301
+ "learning_rate": 5.555555555555556e-06,
302
+ "loss": 0.1578,
303
+ "step": 3116
304
+ },
305
+ {
306
+ "epoch": 19.0,
307
+ "eval_loss": 0.3679310083389282,
308
+ "eval_per": 0.06576839724865731,
309
+ "eval_runtime": 126.1007,
310
+ "eval_samples_per_second": 6.447,
311
+ "eval_steps_per_second": 3.228,
312
+ "eval_wer": 0.08063086183565862,
313
+ "step": 3116
314
+ },
315
+ {
316
+ "epoch": 20.0,
317
+ "learning_rate": 0.0,
318
+ "loss": 0.1317,
319
+ "step": 3280
320
+ },
321
+ {
322
+ "epoch": 20.0,
323
+ "eval_loss": 0.3686366677284241,
324
+ "eval_per": 0.06133986620182795,
325
+ "eval_runtime": 126.1017,
326
+ "eval_samples_per_second": 6.447,
327
+ "eval_steps_per_second": 3.228,
328
+ "eval_wer": 0.07629135364658672,
329
+ "step": 3280
330
+ },
331
+ {
332
+ "epoch": 20.0,
333
+ "step": 3280,
334
+ "total_flos": 1.036683517600052e+18,
335
+ "train_loss": 2.3987663961038357,
336
+ "train_runtime": 4491.8663,
337
+ "train_samples_per_second": 1.456,
338
+ "train_steps_per_second": 0.73
339
+ }
340
+ ],
341
+ "logging_steps": 500,
342
+ "max_steps": 3280,
343
+ "num_train_epochs": 20,
344
+ "save_steps": 500,
345
+ "total_flos": 1.036683517600052e+18,
346
+ "trial_name": null,
347
+ "trial_params": null
348
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:096e5362ab5d7fe4ea0e524b7f759a59b92d17bc133cb7c0fad0fdfbb9e98920
3
+ size 4600