marczenko commited on
Commit
692eb53
1 Parent(s): 658c1c6

SpeechResearch/wtimit-base-normal-all-nofreeze

Browse files
Files changed (5) hide show
  1. README.md +199 -0
  2. config.json +118 -0
  3. model.safetensors +3 -0
  4. preprocessor_config.json +9 -0
  5. training_args.bin +3 -0
README.md ADDED
@@ -0,0 +1,199 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ base_model: facebook/wav2vec2-base
4
+ tags:
5
+ - generated_from_trainer
6
+ datasets:
7
+ - wtimit_asr
8
+ metrics:
9
+ - wer
10
+ model-index:
11
+ - name: wtimit-base-normal-all-nofreeze
12
+ results:
13
+ - task:
14
+ name: Automatic Speech Recognition
15
+ type: automatic-speech-recognition
16
+ dataset:
17
+ name: wtimit_asr
18
+ type: wtimit_asr
19
+ config: clean
20
+ split: None
21
+ args: clean
22
+ metrics:
23
+ - name: Wer
24
+ type: wer
25
+ value: 0.09987953700309014
26
+ ---
27
+
28
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
29
+ should probably proofread and complete it, then remove this comment. -->
30
+
31
+ # wtimit-base-normal-all-nofreeze
32
+
33
+ This model is a fine-tuned version of [facebook/wav2vec2-base](https://huggingface.co/facebook/wav2vec2-base) on the wtimit_asr dataset.
34
+ It achieves the following results on the evaluation set:
35
+ - Loss: 0.3190
36
+ - Wer: 0.0999
37
+
38
+ ## Model description
39
+
40
+ More information needed
41
+
42
+ ## Intended uses & limitations
43
+
44
+ More information needed
45
+
46
+ ## Training and evaluation data
47
+
48
+ More information needed
49
+
50
+ ## Training procedure
51
+
52
+ ### Training hyperparameters
53
+
54
+ The following hyperparameters were used during training:
55
+ - learning_rate: 0.0001
56
+ - train_batch_size: 8
57
+ - eval_batch_size: 8
58
+ - seed: 42
59
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
60
+ - lr_scheduler_type: linear
61
+ - lr_scheduler_warmup_steps: 1000
62
+ - num_epochs: 50
63
+ - mixed_precision_training: Native AMP
64
+
65
+ ### Training results
66
+
67
+ | Training Loss | Epoch | Step | Validation Loss | Wer |
68
+ |:-------------:|:-----:|:------:|:---------------:|:------:|
69
+ | 1.5076 | 0.4 | 1000 | 1.1220 | 0.6793 |
70
+ | 0.4102 | 0.81 | 2000 | 0.7851 | 0.4338 |
71
+ | 0.2278 | 1.21 | 3000 | 0.6897 | 0.3203 |
72
+ | 0.1723 | 1.61 | 4000 | 0.5668 | 0.2890 |
73
+ | 0.1407 | 2.02 | 5000 | 0.4399 | 0.2362 |
74
+ | 0.117 | 2.42 | 6000 | 0.4853 | 0.2508 |
75
+ | 0.098 | 2.83 | 7000 | 0.6732 | 0.2871 |
76
+ | 0.0862 | 3.23 | 8000 | 0.5802 | 0.2680 |
77
+ | 0.0806 | 3.63 | 9000 | 0.4730 | 0.2488 |
78
+ | 0.0706 | 4.04 | 10000 | 0.4001 | 0.1953 |
79
+ | 0.061 | 4.44 | 11000 | 0.4108 | 0.1971 |
80
+ | 0.063 | 4.84 | 12000 | 0.4544 | 0.2056 |
81
+ | 0.0527 | 5.25 | 13000 | 0.4235 | 0.1938 |
82
+ | 0.049 | 5.65 | 14000 | 0.4375 | 0.2054 |
83
+ | 0.0489 | 6.06 | 15000 | 0.5451 | 0.2522 |
84
+ | 0.0473 | 6.46 | 16000 | 0.3939 | 0.1868 |
85
+ | 0.0442 | 6.86 | 17000 | 0.5662 | 0.2548 |
86
+ | 0.0428 | 7.27 | 18000 | 0.6695 | 0.2755 |
87
+ | 0.0379 | 7.67 | 19000 | 0.3929 | 0.1947 |
88
+ | 0.0398 | 8.07 | 20000 | 0.4446 | 0.2066 |
89
+ | 0.0336 | 8.48 | 21000 | 0.5409 | 0.2260 |
90
+ | 0.0316 | 8.88 | 22000 | 0.3819 | 0.1715 |
91
+ | 0.0322 | 9.29 | 23000 | 0.3861 | 0.1711 |
92
+ | 0.0352 | 9.69 | 24000 | 0.4063 | 0.1728 |
93
+ | 0.0315 | 10.09 | 25000 | 0.4992 | 0.2146 |
94
+ | 0.0254 | 10.5 | 26000 | 0.5838 | 0.2158 |
95
+ | 0.0243 | 10.9 | 27000 | 0.3458 | 0.1523 |
96
+ | 0.0245 | 11.3 | 28000 | 0.5121 | 0.1953 |
97
+ | 0.0231 | 11.71 | 29000 | 0.3773 | 0.1616 |
98
+ | 0.0202 | 12.11 | 30000 | 0.4110 | 0.1715 |
99
+ | 0.0261 | 12.52 | 31000 | 0.5376 | 0.2116 |
100
+ | 0.0243 | 12.92 | 32000 | 0.4066 | 0.1569 |
101
+ | 0.0201 | 13.32 | 33000 | 0.5944 | 0.2276 |
102
+ | 0.0211 | 13.73 | 34000 | 0.4670 | 0.1997 |
103
+ | 0.0249 | 14.13 | 35000 | 0.5521 | 0.2254 |
104
+ | 0.021 | 14.53 | 36000 | 0.4602 | 0.2061 |
105
+ | 0.0169 | 14.94 | 37000 | 0.4870 | 0.1690 |
106
+ | 0.0184 | 15.34 | 38000 | 0.6038 | 0.2208 |
107
+ | 0.0207 | 15.74 | 39000 | 0.5266 | 0.2068 |
108
+ | 0.0209 | 16.15 | 40000 | 0.5197 | 0.2083 |
109
+ | 0.0175 | 16.55 | 41000 | 0.5074 | 0.1927 |
110
+ | 0.0164 | 16.96 | 42000 | 0.4594 | 0.1615 |
111
+ | 0.0164 | 17.36 | 43000 | 0.2956 | 0.1151 |
112
+ | 0.0142 | 17.76 | 44000 | 0.3834 | 0.1580 |
113
+ | 0.0139 | 18.17 | 45000 | 0.5316 | 0.2175 |
114
+ | 0.0181 | 18.57 | 46000 | 0.5226 | 0.1890 |
115
+ | 0.0159 | 18.97 | 47000 | 0.4914 | 0.1689 |
116
+ | 0.0127 | 19.38 | 48000 | 0.5454 | 0.1957 |
117
+ | 0.0136 | 19.78 | 49000 | 0.5530 | 0.2172 |
118
+ | 0.0129 | 20.19 | 50000 | 0.6980 | 0.2636 |
119
+ | 0.0131 | 20.59 | 51000 | 0.3984 | 0.1379 |
120
+ | 0.0123 | 20.99 | 52000 | 0.4925 | 0.1843 |
121
+ | 0.0095 | 21.4 | 53000 | 0.5367 | 0.1931 |
122
+ | 0.0124 | 21.8 | 54000 | 0.4299 | 0.1763 |
123
+ | 0.0115 | 22.2 | 55000 | 0.4797 | 0.1803 |
124
+ | 0.0136 | 22.61 | 56000 | 0.6638 | 0.2300 |
125
+ | 0.0121 | 23.01 | 57000 | 0.4292 | 0.1530 |
126
+ | 0.0097 | 23.42 | 58000 | 0.4064 | 0.1520 |
127
+ | 0.0143 | 23.82 | 59000 | 0.4691 | 0.1771 |
128
+ | 0.0092 | 24.22 | 60000 | 0.5134 | 0.2009 |
129
+ | 0.0097 | 24.63 | 61000 | 0.6165 | 0.2281 |
130
+ | 0.0078 | 25.03 | 62000 | 0.4828 | 0.1863 |
131
+ | 0.0114 | 25.43 | 63000 | 0.4817 | 0.1868 |
132
+ | 0.0089 | 25.84 | 64000 | 0.5137 | 0.2003 |
133
+ | 0.0083 | 26.24 | 65000 | 0.4194 | 0.1524 |
134
+ | 0.01 | 26.65 | 66000 | 0.3416 | 0.1332 |
135
+ | 0.0102 | 27.05 | 67000 | 0.3834 | 0.1475 |
136
+ | 0.0076 | 27.45 | 68000 | 0.3390 | 0.1277 |
137
+ | 0.0085 | 27.86 | 69000 | 0.4708 | 0.1843 |
138
+ | 0.0074 | 28.26 | 70000 | 0.4434 | 0.1530 |
139
+ | 0.0078 | 28.66 | 71000 | 0.2942 | 0.1104 |
140
+ | 0.0075 | 29.07 | 72000 | 0.3623 | 0.1442 |
141
+ | 0.0066 | 29.47 | 73000 | 0.4709 | 0.1547 |
142
+ | 0.0073 | 29.87 | 74000 | 0.5198 | 0.1750 |
143
+ | 0.0056 | 30.28 | 75000 | 0.3083 | 0.1211 |
144
+ | 0.0066 | 30.68 | 76000 | 0.3204 | 0.1243 |
145
+ | 0.0048 | 31.09 | 77000 | 0.3713 | 0.1326 |
146
+ | 0.0047 | 31.49 | 78000 | 0.3121 | 0.1018 |
147
+ | 0.0066 | 31.89 | 79000 | 0.4510 | 0.1473 |
148
+ | 0.0053 | 32.3 | 80000 | 0.3599 | 0.1130 |
149
+ | 0.0058 | 32.7 | 81000 | 0.4256 | 0.1463 |
150
+ | 0.0056 | 33.1 | 82000 | 0.4393 | 0.1605 |
151
+ | 0.0046 | 33.51 | 83000 | 0.6327 | 0.2056 |
152
+ | 0.0049 | 33.91 | 84000 | 0.4069 | 0.1360 |
153
+ | 0.0031 | 34.32 | 85000 | 0.4359 | 0.1458 |
154
+ | 0.0052 | 34.72 | 86000 | 0.2825 | 0.1032 |
155
+ | 0.0039 | 35.12 | 87000 | 0.3545 | 0.1256 |
156
+ | 0.003 | 35.53 | 88000 | 0.3674 | 0.1252 |
157
+ | 0.004 | 35.93 | 89000 | 0.3849 | 0.1288 |
158
+ | 0.0029 | 36.33 | 90000 | 0.3465 | 0.1130 |
159
+ | 0.003 | 36.74 | 91000 | 0.4034 | 0.1294 |
160
+ | 0.0036 | 37.14 | 92000 | 0.3456 | 0.1209 |
161
+ | 0.0033 | 37.55 | 93000 | 0.3882 | 0.1407 |
162
+ | 0.0037 | 37.95 | 94000 | 0.3372 | 0.1094 |
163
+ | 0.0025 | 38.35 | 95000 | 0.3601 | 0.1137 |
164
+ | 0.0037 | 38.76 | 96000 | 0.2804 | 0.1027 |
165
+ | 0.0022 | 39.16 | 97000 | 0.4160 | 0.1354 |
166
+ | 0.0027 | 39.56 | 98000 | 0.3379 | 0.1202 |
167
+ | 0.002 | 39.97 | 99000 | 0.3462 | 0.1171 |
168
+ | 0.0021 | 40.37 | 100000 | 0.3694 | 0.1272 |
169
+ | 0.0014 | 40.78 | 101000 | 0.3315 | 0.1048 |
170
+ | 0.0025 | 41.18 | 102000 | 0.3316 | 0.1088 |
171
+ | 0.002 | 41.58 | 103000 | 0.3776 | 0.1319 |
172
+ | 0.0028 | 41.99 | 104000 | 0.3024 | 0.1028 |
173
+ | 0.0015 | 42.39 | 105000 | 0.3087 | 0.1102 |
174
+ | 0.0018 | 42.79 | 106000 | 0.3254 | 0.1067 |
175
+ | 0.0028 | 43.2 | 107000 | 0.3305 | 0.1081 |
176
+ | 0.002 | 43.6 | 108000 | 0.3445 | 0.1120 |
177
+ | 0.0019 | 44.0 | 109000 | 0.3264 | 0.1082 |
178
+ | 0.0019 | 44.41 | 110000 | 0.3650 | 0.1202 |
179
+ | 0.001 | 44.81 | 111000 | 0.3415 | 0.1133 |
180
+ | 0.0015 | 45.22 | 112000 | 0.3194 | 0.1044 |
181
+ | 0.0011 | 45.62 | 113000 | 0.3302 | 0.1085 |
182
+ | 0.0013 | 46.02 | 114000 | 0.3083 | 0.1053 |
183
+ | 0.0008 | 46.43 | 115000 | 0.2976 | 0.0982 |
184
+ | 0.0019 | 46.83 | 116000 | 0.3212 | 0.1057 |
185
+ | 0.0006 | 47.23 | 117000 | 0.3415 | 0.1089 |
186
+ | 0.0025 | 47.64 | 118000 | 0.3188 | 0.1043 |
187
+ | 0.0009 | 48.04 | 119000 | 0.3136 | 0.1025 |
188
+ | 0.0015 | 48.45 | 120000 | 0.3180 | 0.1050 |
189
+ | 0.0013 | 48.85 | 121000 | 0.3439 | 0.1110 |
190
+ | 0.0007 | 49.25 | 122000 | 0.3286 | 0.1048 |
191
+ | 0.0014 | 49.66 | 123000 | 0.3190 | 0.0999 |
192
+
193
+
194
+ ### Framework versions
195
+
196
+ - Transformers 4.39.3
197
+ - Pytorch 2.0.1+cu117
198
+ - Datasets 2.18.0
199
+ - Tokenizers 0.15.2
config.json ADDED
@@ -0,0 +1,118 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "facebook/wav2vec2-base",
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": 256,
16
+ "contrastive_logits_temperature": 0.1,
17
+ "conv_bias": false,
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": false,
49
+ "eos_token_id": 2,
50
+ "feat_extract_activation": "gelu",
51
+ "feat_extract_norm": "group",
52
+ "feat_proj_dropout": 0.1,
53
+ "feat_quantizer_dropout": 0.0,
54
+ "final_dropout": 0.0,
55
+ "freeze_feat_extract_train": true,
56
+ "hidden_act": "gelu",
57
+ "hidden_dropout": 0.1,
58
+ "hidden_size": 768,
59
+ "initializer_range": 0.02,
60
+ "intermediate_size": 3072,
61
+ "layer_norm_eps": 1e-05,
62
+ "layerdrop": 0.0,
63
+ "mask_channel_length": 10,
64
+ "mask_channel_min_space": 1,
65
+ "mask_channel_other": 0.0,
66
+ "mask_channel_prob": 0.0,
67
+ "mask_channel_selection": "static",
68
+ "mask_feature_length": 10,
69
+ "mask_feature_min_masks": 0,
70
+ "mask_feature_prob": 0.0,
71
+ "mask_time_length": 10,
72
+ "mask_time_min_masks": 2,
73
+ "mask_time_min_space": 1,
74
+ "mask_time_other": 0.0,
75
+ "mask_time_prob": 0.05,
76
+ "mask_time_selection": "static",
77
+ "model_type": "wav2vec2",
78
+ "no_mask_channel_overlap": false,
79
+ "no_mask_time_overlap": false,
80
+ "num_adapter_layers": 3,
81
+ "num_attention_heads": 12,
82
+ "num_codevector_groups": 2,
83
+ "num_codevectors_per_group": 320,
84
+ "num_conv_pos_embedding_groups": 16,
85
+ "num_conv_pos_embeddings": 128,
86
+ "num_feat_extract_layers": 7,
87
+ "num_hidden_layers": 12,
88
+ "num_negatives": 100,
89
+ "output_hidden_size": 768,
90
+ "pad_token_id": 29,
91
+ "proj_codevector_dim": 256,
92
+ "tdnn_dilation": [
93
+ 1,
94
+ 2,
95
+ 3,
96
+ 1,
97
+ 1
98
+ ],
99
+ "tdnn_dim": [
100
+ 512,
101
+ 512,
102
+ 512,
103
+ 512,
104
+ 1500
105
+ ],
106
+ "tdnn_kernel": [
107
+ 5,
108
+ 3,
109
+ 3,
110
+ 1,
111
+ 1
112
+ ],
113
+ "torch_dtype": "float32",
114
+ "transformers_version": "4.39.3",
115
+ "use_weighted_layer_sum": false,
116
+ "vocab_size": 32,
117
+ "xvector_output_dim": 512
118
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f575b545fffd1953aa07fe018005f19e497e8a4fcb5d96b8e42bb7c6a94e4b33
3
+ size 377611072
preprocessor_config.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_normalize": true,
3
+ "feature_extractor_type": "Wav2Vec2FeatureExtractor",
4
+ "feature_size": 1,
5
+ "padding_side": "right",
6
+ "padding_value": 0.0,
7
+ "return_attention_mask": false,
8
+ "sampling_rate": 16000
9
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:045d0b96326a5a4cff916fed2db4d066d80cccd2322ab47447c9947983f250eb
3
+ size 4475