yesj1234 commited on
Commit
b7dbb58
1 Parent(s): 762a6c2

Upload folder using huggingface_hub

Browse files
README.md ADDED
@@ -0,0 +1,78 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ base_model: facebook/wav2vec2-large-xlsr-53
4
+ tags:
5
+ - automatic-speech-recognition
6
+ - ./sample_speech.py
7
+ - generated_from_trainer
8
+ metrics:
9
+ - wer
10
+ model-index:
11
+ - name: en-xlsr
12
+ results: []
13
+ ---
14
+
15
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
16
+ should probably proofread and complete it, then remove this comment. -->
17
+
18
+ # en-xlsr
19
+
20
+ This model is a fine-tuned version of [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on the ./SAMPLE_SPEECH.PY - NA dataset.
21
+ It achieves the following results on the evaluation set:
22
+ - Loss: 0.5574
23
+ - Cer: 0.0835
24
+ - Wer: 0.1854
25
+
26
+ ## Model description
27
+
28
+ More information needed
29
+
30
+ ## Intended uses & limitations
31
+
32
+ More information needed
33
+
34
+ ## Training and evaluation data
35
+
36
+ More information needed
37
+
38
+ ## Training procedure
39
+
40
+ ### Training hyperparameters
41
+
42
+ The following hyperparameters were used during training:
43
+ - learning_rate: 0.0003
44
+ - train_batch_size: 4
45
+ - eval_batch_size: 4
46
+ - seed: 42
47
+ - distributed_type: multi-GPU
48
+ - num_devices: 4
49
+ - gradient_accumulation_steps: 4
50
+ - total_train_batch_size: 64
51
+ - total_eval_batch_size: 16
52
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
53
+ - lr_scheduler_type: linear
54
+ - lr_scheduler_warmup_ratio: 0.01
55
+ - num_epochs: 30
56
+
57
+ ### Training results
58
+
59
+ | Training Loss | Epoch | Step | Validation Loss | Cer | Wer |
60
+ |:-------------:|:-----:|:----:|:---------------:|:------:|:------:|
61
+ | 0.6992 | 2.79 | 600 | 0.4981 | 0.1370 | 0.3376 |
62
+ | 0.3394 | 5.58 | 1200 | 0.3934 | 0.1057 | 0.2467 |
63
+ | 0.2376 | 8.37 | 1800 | 0.4123 | 0.1015 | 0.2356 |
64
+ | 0.1877 | 11.16 | 2400 | 0.4269 | 0.0928 | 0.2136 |
65
+ | 0.1494 | 13.95 | 3000 | 0.4648 | 0.0922 | 0.2102 |
66
+ | 0.1186 | 16.74 | 3600 | 0.4835 | 0.0919 | 0.2058 |
67
+ | 0.0966 | 19.53 | 4200 | 0.4986 | 0.0875 | 0.1978 |
68
+ | 0.083 | 22.33 | 4800 | 0.5179 | 0.0862 | 0.1927 |
69
+ | 0.071 | 25.12 | 5400 | 0.5539 | 0.0857 | 0.1908 |
70
+ | 0.0648 | 27.91 | 6000 | 0.5583 | 0.0844 | 0.1870 |
71
+
72
+
73
+ ### Framework versions
74
+
75
+ - Transformers 4.34.1
76
+ - Pytorch 2.1.0+cu121
77
+ - Datasets 2.14.5
78
+ - Tokenizers 0.14.1
added_tokens.json ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ {
2
+ "</s>": 32,
3
+ "<s>": 31
4
+ }
all_results.json ADDED
@@ -0,0 +1,22 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 30.0,
3
+ "eval_cer": 0.0835270292039675,
4
+ "eval_loss": 0.5573986172676086,
5
+ "eval_runtime": 22.2886,
6
+ "eval_samples": 1738,
7
+ "eval_samples_per_second": 77.977,
8
+ "eval_steps_per_second": 4.89,
9
+ "eval_wer": 0.1853669294811711,
10
+ "predict_cer": 0.08437524931900295,
11
+ "predict_loss": 0.5953414440155029,
12
+ "predict_runtime": 22.9108,
13
+ "predict_samples": 1744,
14
+ "predict_samples_per_second": 76.121,
15
+ "predict_steps_per_second": 4.758,
16
+ "predict_wer": 0.19078335975509758,
17
+ "train_loss": 0.3642665664539781,
18
+ "train_runtime": 8377.9265,
19
+ "train_samples": 13754,
20
+ "train_samples_per_second": 49.251,
21
+ "train_steps_per_second": 0.77
22
+ }
checkpoint-6000/added_tokens.json ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ {
2
+ "</s>": 32,
3
+ "<s>": 31
4
+ }
checkpoint-6000/config.json ADDED
@@ -0,0 +1,117 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "facebook/wav2vec2-large-xlsr-53",
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": 0.0125,
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": 0.0125,
54
+ "feat_quantizer_dropout": 0.0,
55
+ "final_dropout": 0.0,
56
+ "gradient_checkpointing": false,
57
+ "hidden_act": "gelu",
58
+ "hidden_dropout": 0.0125,
59
+ "hidden_size": 1024,
60
+ "initializer_range": 0.02,
61
+ "intermediate_size": 4096,
62
+ "layer_norm_eps": 1e-05,
63
+ "layerdrop": 0.0125,
64
+ "mask_channel_length": 10,
65
+ "mask_channel_min_space": 1,
66
+ "mask_channel_other": 0.0,
67
+ "mask_channel_prob": 0.0,
68
+ "mask_channel_selection": "static",
69
+ "mask_feature_length": 10,
70
+ "mask_feature_min_masks": 0,
71
+ "mask_feature_prob": 0.0,
72
+ "mask_time_length": 10,
73
+ "mask_time_min_masks": 2,
74
+ "mask_time_min_space": 1,
75
+ "mask_time_other": 0.0,
76
+ "mask_time_prob": 0.0125,
77
+ "mask_time_selection": "static",
78
+ "model_type": "wav2vec2",
79
+ "num_adapter_layers": 3,
80
+ "num_attention_heads": 16,
81
+ "num_codevector_groups": 2,
82
+ "num_codevectors_per_group": 320,
83
+ "num_conv_pos_embedding_groups": 16,
84
+ "num_conv_pos_embeddings": 128,
85
+ "num_feat_extract_layers": 7,
86
+ "num_hidden_layers": 24,
87
+ "num_negatives": 100,
88
+ "output_hidden_size": 1024,
89
+ "pad_token_id": 30,
90
+ "proj_codevector_dim": 768,
91
+ "tdnn_dilation": [
92
+ 1,
93
+ 2,
94
+ 3,
95
+ 1,
96
+ 1
97
+ ],
98
+ "tdnn_dim": [
99
+ 512,
100
+ 512,
101
+ 512,
102
+ 512,
103
+ 1500
104
+ ],
105
+ "tdnn_kernel": [
106
+ 5,
107
+ 3,
108
+ 3,
109
+ 1,
110
+ 1
111
+ ],
112
+ "torch_dtype": "float32",
113
+ "transformers_version": "4.34.1",
114
+ "use_weighted_layer_sum": false,
115
+ "vocab_size": 33,
116
+ "xvector_output_dim": 512
117
+ }
checkpoint-6000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe7b3490bc172b849427309b8795bc762f78a725e4abfb5f4cb4902cb532e6e3
3
+ size 2490430390
checkpoint-6000/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
+ }
checkpoint-6000/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f45ac5e952c77b447f72b13c7751afc473a7c2c72e3e04ad84b43e1f62eb874f
3
+ size 1262037674
checkpoint-6000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aa102ff95efd149fa2cd8efec2289a718668c739aed81b75a7702305e150d7df
3
+ size 15024
checkpoint-6000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cd679681cbdac7648a0d27c4af465274796151eda01674dad77ea20d4002d54e
3
+ size 15024
checkpoint-6000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6d64158e248fca2adafdd6d7acb063c8266c6d5ba5442082c3a815363e87afe4
3
+ size 15088
checkpoint-6000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:38e2907b83c90962f6e3fa305a5cf6daadef666689d4e71500c50430a1172ed9
3
+ size 15088
checkpoint-6000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d8d9645dcdd8f59b86ddf7b65ba1bc656747784eb1e271097b4b6aadce19d468
3
+ size 1064
checkpoint-6000/special_tokens_map.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<s>",
4
+ "lstrip": false,
5
+ "normalized": true,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "</s>",
11
+ "lstrip": false,
12
+ "normalized": true,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": {
17
+ "content": "[PAD]",
18
+ "lstrip": true,
19
+ "normalized": false,
20
+ "rstrip": true,
21
+ "single_word": false
22
+ },
23
+ "unk_token": {
24
+ "content": "[UNK]",
25
+ "lstrip": true,
26
+ "normalized": false,
27
+ "rstrip": true,
28
+ "single_word": false
29
+ }
30
+ }
checkpoint-6000/tokenizer_config.json ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "added_tokens_decoder": {
3
+ "29": {
4
+ "content": "[UNK]",
5
+ "lstrip": true,
6
+ "normalized": false,
7
+ "rstrip": true,
8
+ "single_word": false,
9
+ "special": false
10
+ },
11
+ "30": {
12
+ "content": "[PAD]",
13
+ "lstrip": true,
14
+ "normalized": false,
15
+ "rstrip": true,
16
+ "single_word": false,
17
+ "special": false
18
+ },
19
+ "31": {
20
+ "content": "<s>",
21
+ "lstrip": false,
22
+ "normalized": true,
23
+ "rstrip": false,
24
+ "single_word": false,
25
+ "special": true
26
+ },
27
+ "32": {
28
+ "content": "</s>",
29
+ "lstrip": false,
30
+ "normalized": true,
31
+ "rstrip": false,
32
+ "single_word": false,
33
+ "special": true
34
+ }
35
+ },
36
+ "bos_token": "<s>",
37
+ "clean_up_tokenization_spaces": true,
38
+ "config": null,
39
+ "do_lower_case": false,
40
+ "eos_token": "</s>",
41
+ "model_max_length": 1000000000000000019884624838656,
42
+ "pad_token": "[PAD]",
43
+ "processor_class": "Wav2Vec2Processor",
44
+ "replace_word_delimiter_char": " ",
45
+ "target_lang": null,
46
+ "tokenizer_class": "Wav2Vec2CTCTokenizer",
47
+ "tokenizer_type": "wav2vec2",
48
+ "trust_remote_code": false,
49
+ "unk_token": "[UNK]",
50
+ "word_delimiter_token": "|"
51
+ }
checkpoint-6000/trainer_state.json ADDED
@@ -0,0 +1,479 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 0.18702100678171693,
3
+ "best_model_checkpoint": "./en-xlsr/checkpoint-6000",
4
+ "epoch": 27.906976744186046,
5
+ "eval_steps": 600,
6
+ "global_step": 6000,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.47,
13
+ "learning_rate": 0.0002983555207517619,
14
+ "loss": 4.4931,
15
+ "step": 100
16
+ },
17
+ {
18
+ "epoch": 0.93,
19
+ "learning_rate": 0.0002936570086139389,
20
+ "loss": 2.9634,
21
+ "step": 200
22
+ },
23
+ {
24
+ "epoch": 1.4,
25
+ "learning_rate": 0.0002889584964761159,
26
+ "loss": 2.8686,
27
+ "step": 300
28
+ },
29
+ {
30
+ "epoch": 1.86,
31
+ "learning_rate": 0.00028425998433829287,
32
+ "loss": 1.9458,
33
+ "step": 400
34
+ },
35
+ {
36
+ "epoch": 2.33,
37
+ "learning_rate": 0.0002795614722004698,
38
+ "loss": 0.8991,
39
+ "step": 500
40
+ },
41
+ {
42
+ "epoch": 2.79,
43
+ "learning_rate": 0.0002748629600626468,
44
+ "loss": 0.6992,
45
+ "step": 600
46
+ },
47
+ {
48
+ "epoch": 2.79,
49
+ "eval_cer": 0.13698708949899893,
50
+ "eval_loss": 0.4980902671813965,
51
+ "eval_runtime": 22.423,
52
+ "eval_samples_per_second": 77.51,
53
+ "eval_steps_per_second": 4.861,
54
+ "eval_wer": 0.33759717704140707,
55
+ "step": 600
56
+ },
57
+ {
58
+ "epoch": 3.26,
59
+ "learning_rate": 0.0002701644479248238,
60
+ "loss": 0.5755,
61
+ "step": 700
62
+ },
63
+ {
64
+ "epoch": 3.72,
65
+ "learning_rate": 0.0002654659357870008,
66
+ "loss": 0.502,
67
+ "step": 800
68
+ },
69
+ {
70
+ "epoch": 4.19,
71
+ "learning_rate": 0.00026076742364917777,
72
+ "loss": 0.4539,
73
+ "step": 900
74
+ },
75
+ {
76
+ "epoch": 4.65,
77
+ "learning_rate": 0.0002560689115113547,
78
+ "loss": 0.4015,
79
+ "step": 1000
80
+ },
81
+ {
82
+ "epoch": 5.12,
83
+ "learning_rate": 0.0002513703993735317,
84
+ "loss": 0.3877,
85
+ "step": 1100
86
+ },
87
+ {
88
+ "epoch": 5.58,
89
+ "learning_rate": 0.0002466718872357087,
90
+ "loss": 0.3394,
91
+ "step": 1200
92
+ },
93
+ {
94
+ "epoch": 5.58,
95
+ "eval_cer": 0.10568890525395255,
96
+ "eval_loss": 0.393413245677948,
97
+ "eval_runtime": 22.3286,
98
+ "eval_samples_per_second": 77.838,
99
+ "eval_steps_per_second": 4.882,
100
+ "eval_wer": 0.24673319733142196,
101
+ "step": 1200
102
+ },
103
+ {
104
+ "epoch": 6.05,
105
+ "learning_rate": 0.00024197337509788565,
106
+ "loss": 0.336,
107
+ "step": 1300
108
+ },
109
+ {
110
+ "epoch": 6.51,
111
+ "learning_rate": 0.00023727486296006264,
112
+ "loss": 0.289,
113
+ "step": 1400
114
+ },
115
+ {
116
+ "epoch": 6.98,
117
+ "learning_rate": 0.0002325763508222396,
118
+ "loss": 0.2961,
119
+ "step": 1500
120
+ },
121
+ {
122
+ "epoch": 7.44,
123
+ "learning_rate": 0.0002278778386844166,
124
+ "loss": 0.2615,
125
+ "step": 1600
126
+ },
127
+ {
128
+ "epoch": 7.91,
129
+ "learning_rate": 0.00022317932654659358,
130
+ "loss": 0.2656,
131
+ "step": 1700
132
+ },
133
+ {
134
+ "epoch": 8.37,
135
+ "learning_rate": 0.00021848081440877052,
136
+ "loss": 0.2376,
137
+ "step": 1800
138
+ },
139
+ {
140
+ "epoch": 8.37,
141
+ "eval_cer": 0.10152348514486917,
142
+ "eval_loss": 0.41225335001945496,
143
+ "eval_runtime": 22.5992,
144
+ "eval_samples_per_second": 76.905,
145
+ "eval_steps_per_second": 4.823,
146
+ "eval_wer": 0.2355957435077466,
147
+ "step": 1800
148
+ },
149
+ {
150
+ "epoch": 8.84,
151
+ "learning_rate": 0.0002137823022709475,
152
+ "loss": 0.227,
153
+ "step": 1900
154
+ },
155
+ {
156
+ "epoch": 9.3,
157
+ "learning_rate": 0.00020908379013312447,
158
+ "loss": 0.2208,
159
+ "step": 2000
160
+ },
161
+ {
162
+ "epoch": 9.77,
163
+ "learning_rate": 0.00020438527799530146,
164
+ "loss": 0.2109,
165
+ "step": 2100
166
+ },
167
+ {
168
+ "epoch": 10.23,
169
+ "learning_rate": 0.00019968676585747845,
170
+ "loss": 0.2003,
171
+ "step": 2200
172
+ },
173
+ {
174
+ "epoch": 10.7,
175
+ "learning_rate": 0.00019498825371965542,
176
+ "loss": 0.1948,
177
+ "step": 2300
178
+ },
179
+ {
180
+ "epoch": 11.16,
181
+ "learning_rate": 0.0001902897415818324,
182
+ "loss": 0.1877,
183
+ "step": 2400
184
+ },
185
+ {
186
+ "epoch": 11.16,
187
+ "eval_cer": 0.09280141762363933,
188
+ "eval_loss": 0.4269295632839203,
189
+ "eval_runtime": 22.6943,
190
+ "eval_samples_per_second": 76.583,
191
+ "eval_steps_per_second": 4.803,
192
+ "eval_wer": 0.21359651541048685,
193
+ "step": 2400
194
+ },
195
+ {
196
+ "epoch": 11.63,
197
+ "learning_rate": 0.0001855912294440094,
198
+ "loss": 0.1777,
199
+ "step": 2500
200
+ },
201
+ {
202
+ "epoch": 12.09,
203
+ "learning_rate": 0.00018089271730618636,
204
+ "loss": 0.1749,
205
+ "step": 2600
206
+ },
207
+ {
208
+ "epoch": 12.56,
209
+ "learning_rate": 0.00017619420516836332,
210
+ "loss": 0.1607,
211
+ "step": 2700
212
+ },
213
+ {
214
+ "epoch": 13.02,
215
+ "learning_rate": 0.00017149569303054031,
216
+ "loss": 0.1595,
217
+ "step": 2800
218
+ },
219
+ {
220
+ "epoch": 13.49,
221
+ "learning_rate": 0.00016679718089271728,
222
+ "loss": 0.1513,
223
+ "step": 2900
224
+ },
225
+ {
226
+ "epoch": 13.95,
227
+ "learning_rate": 0.00016209866875489427,
228
+ "loss": 0.1494,
229
+ "step": 3000
230
+ },
231
+ {
232
+ "epoch": 13.95,
233
+ "eval_cer": 0.09215704324212368,
234
+ "eval_loss": 0.4648122489452362,
235
+ "eval_runtime": 22.977,
236
+ "eval_samples_per_second": 75.641,
237
+ "eval_steps_per_second": 4.744,
238
+ "eval_wer": 0.21023322489937696,
239
+ "step": 3000
240
+ },
241
+ {
242
+ "epoch": 14.42,
243
+ "learning_rate": 0.00015740015661707126,
244
+ "loss": 0.1414,
245
+ "step": 3100
246
+ },
247
+ {
248
+ "epoch": 14.88,
249
+ "learning_rate": 0.00015270164447924822,
250
+ "loss": 0.1363,
251
+ "step": 3200
252
+ },
253
+ {
254
+ "epoch": 15.35,
255
+ "learning_rate": 0.00014800313234142518,
256
+ "loss": 0.1319,
257
+ "step": 3300
258
+ },
259
+ {
260
+ "epoch": 15.81,
261
+ "learning_rate": 0.00014330462020360218,
262
+ "loss": 0.1244,
263
+ "step": 3400
264
+ },
265
+ {
266
+ "epoch": 16.28,
267
+ "learning_rate": 0.00013860610806577917,
268
+ "loss": 0.1194,
269
+ "step": 3500
270
+ },
271
+ {
272
+ "epoch": 16.74,
273
+ "learning_rate": 0.00013390759592795613,
274
+ "loss": 0.1186,
275
+ "step": 3600
276
+ },
277
+ {
278
+ "epoch": 16.74,
279
+ "eval_cer": 0.09190389616367109,
280
+ "eval_loss": 0.48351120948791504,
281
+ "eval_runtime": 22.4373,
282
+ "eval_samples_per_second": 77.46,
283
+ "eval_steps_per_second": 4.858,
284
+ "eval_wer": 0.20576721618790317,
285
+ "step": 3600
286
+ },
287
+ {
288
+ "epoch": 17.21,
289
+ "learning_rate": 0.00012920908379013312,
290
+ "loss": 0.118,
291
+ "step": 3700
292
+ },
293
+ {
294
+ "epoch": 17.67,
295
+ "learning_rate": 0.00012451057165231008,
296
+ "loss": 0.1098,
297
+ "step": 3800
298
+ },
299
+ {
300
+ "epoch": 18.14,
301
+ "learning_rate": 0.00011981205951448707,
302
+ "loss": 0.1048,
303
+ "step": 3900
304
+ },
305
+ {
306
+ "epoch": 18.6,
307
+ "learning_rate": 0.00011511354737666405,
308
+ "loss": 0.1051,
309
+ "step": 4000
310
+ },
311
+ {
312
+ "epoch": 19.07,
313
+ "learning_rate": 0.00011041503523884103,
314
+ "loss": 0.1014,
315
+ "step": 4100
316
+ },
317
+ {
318
+ "epoch": 19.53,
319
+ "learning_rate": 0.000105716523101018,
320
+ "loss": 0.0966,
321
+ "step": 4200
322
+ },
323
+ {
324
+ "epoch": 19.53,
325
+ "eval_cer": 0.08753135571767197,
326
+ "eval_loss": 0.49859338998794556,
327
+ "eval_runtime": 22.5733,
328
+ "eval_samples_per_second": 76.994,
329
+ "eval_steps_per_second": 4.829,
330
+ "eval_wer": 0.19782764514528312,
331
+ "step": 4200
332
+ },
333
+ {
334
+ "epoch": 20.0,
335
+ "learning_rate": 0.00010101801096319498,
336
+ "loss": 0.097,
337
+ "step": 4300
338
+ },
339
+ {
340
+ "epoch": 20.47,
341
+ "learning_rate": 9.631949882537196e-05,
342
+ "loss": 0.09,
343
+ "step": 4400
344
+ },
345
+ {
346
+ "epoch": 20.93,
347
+ "learning_rate": 9.162098668754895e-05,
348
+ "loss": 0.0904,
349
+ "step": 4500
350
+ },
351
+ {
352
+ "epoch": 21.4,
353
+ "learning_rate": 8.692247454972591e-05,
354
+ "loss": 0.0822,
355
+ "step": 4600
356
+ },
357
+ {
358
+ "epoch": 21.86,
359
+ "learning_rate": 8.222396241190289e-05,
360
+ "loss": 0.0869,
361
+ "step": 4700
362
+ },
363
+ {
364
+ "epoch": 22.33,
365
+ "learning_rate": 7.752545027407988e-05,
366
+ "loss": 0.083,
367
+ "step": 4800
368
+ },
369
+ {
370
+ "epoch": 22.33,
371
+ "eval_cer": 0.08620808689848802,
372
+ "eval_loss": 0.5179353952407837,
373
+ "eval_runtime": 22.6454,
374
+ "eval_samples_per_second": 76.748,
375
+ "eval_steps_per_second": 4.813,
376
+ "eval_wer": 0.192700005513591,
377
+ "step": 4800
378
+ },
379
+ {
380
+ "epoch": 22.79,
381
+ "learning_rate": 7.282693813625684e-05,
382
+ "loss": 0.0817,
383
+ "step": 4900
384
+ },
385
+ {
386
+ "epoch": 23.26,
387
+ "learning_rate": 6.812842599843383e-05,
388
+ "loss": 0.0774,
389
+ "step": 5000
390
+ },
391
+ {
392
+ "epoch": 23.72,
393
+ "learning_rate": 6.34299138606108e-05,
394
+ "loss": 0.0743,
395
+ "step": 5100
396
+ },
397
+ {
398
+ "epoch": 24.19,
399
+ "learning_rate": 5.873140172278778e-05,
400
+ "loss": 0.073,
401
+ "step": 5200
402
+ },
403
+ {
404
+ "epoch": 24.65,
405
+ "learning_rate": 5.4032889584964754e-05,
406
+ "loss": 0.0699,
407
+ "step": 5300
408
+ },
409
+ {
410
+ "epoch": 25.12,
411
+ "learning_rate": 4.933437744714174e-05,
412
+ "loss": 0.071,
413
+ "step": 5400
414
+ },
415
+ {
416
+ "epoch": 25.12,
417
+ "eval_cer": 0.08569028605619865,
418
+ "eval_loss": 0.5539018511772156,
419
+ "eval_runtime": 22.6635,
420
+ "eval_samples_per_second": 76.687,
421
+ "eval_steps_per_second": 4.81,
422
+ "eval_wer": 0.19082538457297238,
423
+ "step": 5400
424
+ },
425
+ {
426
+ "epoch": 25.58,
427
+ "learning_rate": 4.463586530931871e-05,
428
+ "loss": 0.0706,
429
+ "step": 5500
430
+ },
431
+ {
432
+ "epoch": 26.05,
433
+ "learning_rate": 3.993735317149569e-05,
434
+ "loss": 0.0696,
435
+ "step": 5600
436
+ },
437
+ {
438
+ "epoch": 26.51,
439
+ "learning_rate": 3.523884103367267e-05,
440
+ "loss": 0.0702,
441
+ "step": 5700
442
+ },
443
+ {
444
+ "epoch": 26.98,
445
+ "learning_rate": 3.0540328895849644e-05,
446
+ "loss": 0.0656,
447
+ "step": 5800
448
+ },
449
+ {
450
+ "epoch": 27.44,
451
+ "learning_rate": 2.584181675802662e-05,
452
+ "loss": 0.0662,
453
+ "step": 5900
454
+ },
455
+ {
456
+ "epoch": 27.91,
457
+ "learning_rate": 2.1143304620203598e-05,
458
+ "loss": 0.0648,
459
+ "step": 6000
460
+ },
461
+ {
462
+ "epoch": 27.91,
463
+ "eval_cer": 0.08437852392239892,
464
+ "eval_loss": 0.5582728981971741,
465
+ "eval_runtime": 22.4539,
466
+ "eval_samples_per_second": 77.403,
467
+ "eval_steps_per_second": 4.854,
468
+ "eval_wer": 0.18702100678171693,
469
+ "step": 6000
470
+ }
471
+ ],
472
+ "logging_steps": 100,
473
+ "max_steps": 6450,
474
+ "num_train_epochs": 30,
475
+ "save_steps": 600,
476
+ "total_flos": 4.161559358313857e+19,
477
+ "trial_name": null,
478
+ "trial_params": null
479
+ }
checkpoint-6000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8dc5ca704c2a09f0fa35a3898e0f40a7516ae8d96ac92fa615bf691eda5db270
3
+ size 4536
checkpoint-6000/vocab.json ADDED
@@ -0,0 +1,33 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "\"": 1,
3
+ "-": 2,
4
+ "[PAD]": 30,
5
+ "[UNK]": 29,
6
+ "a": 3,
7
+ "b": 4,
8
+ "c": 5,
9
+ "d": 6,
10
+ "e": 7,
11
+ "f": 8,
12
+ "g": 9,
13
+ "h": 10,
14
+ "i": 11,
15
+ "j": 12,
16
+ "k": 13,
17
+ "l": 14,
18
+ "m": 15,
19
+ "n": 16,
20
+ "o": 17,
21
+ "p": 18,
22
+ "q": 19,
23
+ "r": 20,
24
+ "s": 21,
25
+ "t": 22,
26
+ "u": 23,
27
+ "v": 24,
28
+ "w": 25,
29
+ "x": 26,
30
+ "y": 27,
31
+ "z": 28,
32
+ "|": 0
33
+ }
config.json ADDED
@@ -0,0 +1,117 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "facebook/wav2vec2-large-xlsr-53",
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": 0.0125,
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": 0.0125,
54
+ "feat_quantizer_dropout": 0.0,
55
+ "final_dropout": 0.0,
56
+ "gradient_checkpointing": false,
57
+ "hidden_act": "gelu",
58
+ "hidden_dropout": 0.0125,
59
+ "hidden_size": 1024,
60
+ "initializer_range": 0.02,
61
+ "intermediate_size": 4096,
62
+ "layer_norm_eps": 1e-05,
63
+ "layerdrop": 0.0125,
64
+ "mask_channel_length": 10,
65
+ "mask_channel_min_space": 1,
66
+ "mask_channel_other": 0.0,
67
+ "mask_channel_prob": 0.0,
68
+ "mask_channel_selection": "static",
69
+ "mask_feature_length": 10,
70
+ "mask_feature_min_masks": 0,
71
+ "mask_feature_prob": 0.0,
72
+ "mask_time_length": 10,
73
+ "mask_time_min_masks": 2,
74
+ "mask_time_min_space": 1,
75
+ "mask_time_other": 0.0,
76
+ "mask_time_prob": 0.0125,
77
+ "mask_time_selection": "static",
78
+ "model_type": "wav2vec2",
79
+ "num_adapter_layers": 3,
80
+ "num_attention_heads": 16,
81
+ "num_codevector_groups": 2,
82
+ "num_codevectors_per_group": 320,
83
+ "num_conv_pos_embedding_groups": 16,
84
+ "num_conv_pos_embeddings": 128,
85
+ "num_feat_extract_layers": 7,
86
+ "num_hidden_layers": 24,
87
+ "num_negatives": 100,
88
+ "output_hidden_size": 1024,
89
+ "pad_token_id": 30,
90
+ "proj_codevector_dim": 768,
91
+ "tdnn_dilation": [
92
+ 1,
93
+ 2,
94
+ 3,
95
+ 1,
96
+ 1
97
+ ],
98
+ "tdnn_dim": [
99
+ 512,
100
+ 512,
101
+ 512,
102
+ 512,
103
+ 1500
104
+ ],
105
+ "tdnn_kernel": [
106
+ 5,
107
+ 3,
108
+ 3,
109
+ 1,
110
+ 1
111
+ ],
112
+ "torch_dtype": "float32",
113
+ "transformers_version": "4.34.1",
114
+ "use_weighted_layer_sum": false,
115
+ "vocab_size": 33,
116
+ "xvector_output_dim": 512
117
+ }
eval_results.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 30.0,
3
+ "eval_cer": 0.0835270292039675,
4
+ "eval_loss": 0.5573986172676086,
5
+ "eval_runtime": 22.2886,
6
+ "eval_samples": 1738,
7
+ "eval_samples_per_second": 77.977,
8
+ "eval_steps_per_second": 4.89,
9
+ "eval_wer": 0.1853669294811711
10
+ }
predict_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "predict_cer": 0.08437524931900295,
3
+ "predict_loss": 0.5953414440155029,
4
+ "predict_runtime": 22.9108,
5
+ "predict_samples": 1744,
6
+ "predict_samples_per_second": 76.121,
7
+ "predict_steps_per_second": 4.758,
8
+ "predict_wer": 0.19078335975509758
9
+ }
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
+ }
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f45ac5e952c77b447f72b13c7751afc473a7c2c72e3e04ad84b43e1f62eb874f
3
+ size 1262037674
special_tokens_map.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<s>",
4
+ "lstrip": false,
5
+ "normalized": true,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "</s>",
11
+ "lstrip": false,
12
+ "normalized": true,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": {
17
+ "content": "[PAD]",
18
+ "lstrip": true,
19
+ "normalized": false,
20
+ "rstrip": true,
21
+ "single_word": false
22
+ },
23
+ "unk_token": {
24
+ "content": "[UNK]",
25
+ "lstrip": true,
26
+ "normalized": false,
27
+ "rstrip": true,
28
+ "single_word": false
29
+ }
30
+ }
tokenizer_config.json ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "added_tokens_decoder": {
3
+ "29": {
4
+ "content": "[UNK]",
5
+ "lstrip": true,
6
+ "normalized": false,
7
+ "rstrip": true,
8
+ "single_word": false,
9
+ "special": false
10
+ },
11
+ "30": {
12
+ "content": "[PAD]",
13
+ "lstrip": true,
14
+ "normalized": false,
15
+ "rstrip": true,
16
+ "single_word": false,
17
+ "special": false
18
+ },
19
+ "31": {
20
+ "content": "<s>",
21
+ "lstrip": false,
22
+ "normalized": true,
23
+ "rstrip": false,
24
+ "single_word": false,
25
+ "special": true
26
+ },
27
+ "32": {
28
+ "content": "</s>",
29
+ "lstrip": false,
30
+ "normalized": true,
31
+ "rstrip": false,
32
+ "single_word": false,
33
+ "special": true
34
+ }
35
+ },
36
+ "bos_token": "<s>",
37
+ "clean_up_tokenization_spaces": true,
38
+ "config": null,
39
+ "do_lower_case": false,
40
+ "eos_token": "</s>",
41
+ "model_max_length": 1000000000000000019884624838656,
42
+ "pad_token": "[PAD]",
43
+ "processor_class": "Wav2Vec2Processor",
44
+ "replace_word_delimiter_char": " ",
45
+ "target_lang": null,
46
+ "tokenizer_class": "Wav2Vec2CTCTokenizer",
47
+ "tokenizer_type": "wav2vec2",
48
+ "trust_remote_code": false,
49
+ "unk_token": "[UNK]",
50
+ "word_delimiter_token": "|"
51
+ }
train_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 30.0,
3
+ "train_loss": 0.3642665664539781,
4
+ "train_runtime": 8377.9265,
5
+ "train_samples": 13754,
6
+ "train_samples_per_second": 49.251,
7
+ "train_steps_per_second": 0.77
8
+ }
trainer_state.json ADDED
@@ -0,0 +1,512 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 0.18702100678171693,
3
+ "best_model_checkpoint": "./en-xlsr/checkpoint-6000",
4
+ "epoch": 30.0,
5
+ "eval_steps": 600,
6
+ "global_step": 6450,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.47,
13
+ "learning_rate": 0.0002983555207517619,
14
+ "loss": 4.4931,
15
+ "step": 100
16
+ },
17
+ {
18
+ "epoch": 0.93,
19
+ "learning_rate": 0.0002936570086139389,
20
+ "loss": 2.9634,
21
+ "step": 200
22
+ },
23
+ {
24
+ "epoch": 1.4,
25
+ "learning_rate": 0.0002889584964761159,
26
+ "loss": 2.8686,
27
+ "step": 300
28
+ },
29
+ {
30
+ "epoch": 1.86,
31
+ "learning_rate": 0.00028425998433829287,
32
+ "loss": 1.9458,
33
+ "step": 400
34
+ },
35
+ {
36
+ "epoch": 2.33,
37
+ "learning_rate": 0.0002795614722004698,
38
+ "loss": 0.8991,
39
+ "step": 500
40
+ },
41
+ {
42
+ "epoch": 2.79,
43
+ "learning_rate": 0.0002748629600626468,
44
+ "loss": 0.6992,
45
+ "step": 600
46
+ },
47
+ {
48
+ "epoch": 2.79,
49
+ "eval_cer": 0.13698708949899893,
50
+ "eval_loss": 0.4980902671813965,
51
+ "eval_runtime": 22.423,
52
+ "eval_samples_per_second": 77.51,
53
+ "eval_steps_per_second": 4.861,
54
+ "eval_wer": 0.33759717704140707,
55
+ "step": 600
56
+ },
57
+ {
58
+ "epoch": 3.26,
59
+ "learning_rate": 0.0002701644479248238,
60
+ "loss": 0.5755,
61
+ "step": 700
62
+ },
63
+ {
64
+ "epoch": 3.72,
65
+ "learning_rate": 0.0002654659357870008,
66
+ "loss": 0.502,
67
+ "step": 800
68
+ },
69
+ {
70
+ "epoch": 4.19,
71
+ "learning_rate": 0.00026076742364917777,
72
+ "loss": 0.4539,
73
+ "step": 900
74
+ },
75
+ {
76
+ "epoch": 4.65,
77
+ "learning_rate": 0.0002560689115113547,
78
+ "loss": 0.4015,
79
+ "step": 1000
80
+ },
81
+ {
82
+ "epoch": 5.12,
83
+ "learning_rate": 0.0002513703993735317,
84
+ "loss": 0.3877,
85
+ "step": 1100
86
+ },
87
+ {
88
+ "epoch": 5.58,
89
+ "learning_rate": 0.0002466718872357087,
90
+ "loss": 0.3394,
91
+ "step": 1200
92
+ },
93
+ {
94
+ "epoch": 5.58,
95
+ "eval_cer": 0.10568890525395255,
96
+ "eval_loss": 0.393413245677948,
97
+ "eval_runtime": 22.3286,
98
+ "eval_samples_per_second": 77.838,
99
+ "eval_steps_per_second": 4.882,
100
+ "eval_wer": 0.24673319733142196,
101
+ "step": 1200
102
+ },
103
+ {
104
+ "epoch": 6.05,
105
+ "learning_rate": 0.00024197337509788565,
106
+ "loss": 0.336,
107
+ "step": 1300
108
+ },
109
+ {
110
+ "epoch": 6.51,
111
+ "learning_rate": 0.00023727486296006264,
112
+ "loss": 0.289,
113
+ "step": 1400
114
+ },
115
+ {
116
+ "epoch": 6.98,
117
+ "learning_rate": 0.0002325763508222396,
118
+ "loss": 0.2961,
119
+ "step": 1500
120
+ },
121
+ {
122
+ "epoch": 7.44,
123
+ "learning_rate": 0.0002278778386844166,
124
+ "loss": 0.2615,
125
+ "step": 1600
126
+ },
127
+ {
128
+ "epoch": 7.91,
129
+ "learning_rate": 0.00022317932654659358,
130
+ "loss": 0.2656,
131
+ "step": 1700
132
+ },
133
+ {
134
+ "epoch": 8.37,
135
+ "learning_rate": 0.00021848081440877052,
136
+ "loss": 0.2376,
137
+ "step": 1800
138
+ },
139
+ {
140
+ "epoch": 8.37,
141
+ "eval_cer": 0.10152348514486917,
142
+ "eval_loss": 0.41225335001945496,
143
+ "eval_runtime": 22.5992,
144
+ "eval_samples_per_second": 76.905,
145
+ "eval_steps_per_second": 4.823,
146
+ "eval_wer": 0.2355957435077466,
147
+ "step": 1800
148
+ },
149
+ {
150
+ "epoch": 8.84,
151
+ "learning_rate": 0.0002137823022709475,
152
+ "loss": 0.227,
153
+ "step": 1900
154
+ },
155
+ {
156
+ "epoch": 9.3,
157
+ "learning_rate": 0.00020908379013312447,
158
+ "loss": 0.2208,
159
+ "step": 2000
160
+ },
161
+ {
162
+ "epoch": 9.77,
163
+ "learning_rate": 0.00020438527799530146,
164
+ "loss": 0.2109,
165
+ "step": 2100
166
+ },
167
+ {
168
+ "epoch": 10.23,
169
+ "learning_rate": 0.00019968676585747845,
170
+ "loss": 0.2003,
171
+ "step": 2200
172
+ },
173
+ {
174
+ "epoch": 10.7,
175
+ "learning_rate": 0.00019498825371965542,
176
+ "loss": 0.1948,
177
+ "step": 2300
178
+ },
179
+ {
180
+ "epoch": 11.16,
181
+ "learning_rate": 0.0001902897415818324,
182
+ "loss": 0.1877,
183
+ "step": 2400
184
+ },
185
+ {
186
+ "epoch": 11.16,
187
+ "eval_cer": 0.09280141762363933,
188
+ "eval_loss": 0.4269295632839203,
189
+ "eval_runtime": 22.6943,
190
+ "eval_samples_per_second": 76.583,
191
+ "eval_steps_per_second": 4.803,
192
+ "eval_wer": 0.21359651541048685,
193
+ "step": 2400
194
+ },
195
+ {
196
+ "epoch": 11.63,
197
+ "learning_rate": 0.0001855912294440094,
198
+ "loss": 0.1777,
199
+ "step": 2500
200
+ },
201
+ {
202
+ "epoch": 12.09,
203
+ "learning_rate": 0.00018089271730618636,
204
+ "loss": 0.1749,
205
+ "step": 2600
206
+ },
207
+ {
208
+ "epoch": 12.56,
209
+ "learning_rate": 0.00017619420516836332,
210
+ "loss": 0.1607,
211
+ "step": 2700
212
+ },
213
+ {
214
+ "epoch": 13.02,
215
+ "learning_rate": 0.00017149569303054031,
216
+ "loss": 0.1595,
217
+ "step": 2800
218
+ },
219
+ {
220
+ "epoch": 13.49,
221
+ "learning_rate": 0.00016679718089271728,
222
+ "loss": 0.1513,
223
+ "step": 2900
224
+ },
225
+ {
226
+ "epoch": 13.95,
227
+ "learning_rate": 0.00016209866875489427,
228
+ "loss": 0.1494,
229
+ "step": 3000
230
+ },
231
+ {
232
+ "epoch": 13.95,
233
+ "eval_cer": 0.09215704324212368,
234
+ "eval_loss": 0.4648122489452362,
235
+ "eval_runtime": 22.977,
236
+ "eval_samples_per_second": 75.641,
237
+ "eval_steps_per_second": 4.744,
238
+ "eval_wer": 0.21023322489937696,
239
+ "step": 3000
240
+ },
241
+ {
242
+ "epoch": 14.42,
243
+ "learning_rate": 0.00015740015661707126,
244
+ "loss": 0.1414,
245
+ "step": 3100
246
+ },
247
+ {
248
+ "epoch": 14.88,
249
+ "learning_rate": 0.00015270164447924822,
250
+ "loss": 0.1363,
251
+ "step": 3200
252
+ },
253
+ {
254
+ "epoch": 15.35,
255
+ "learning_rate": 0.00014800313234142518,
256
+ "loss": 0.1319,
257
+ "step": 3300
258
+ },
259
+ {
260
+ "epoch": 15.81,
261
+ "learning_rate": 0.00014330462020360218,
262
+ "loss": 0.1244,
263
+ "step": 3400
264
+ },
265
+ {
266
+ "epoch": 16.28,
267
+ "learning_rate": 0.00013860610806577917,
268
+ "loss": 0.1194,
269
+ "step": 3500
270
+ },
271
+ {
272
+ "epoch": 16.74,
273
+ "learning_rate": 0.00013390759592795613,
274
+ "loss": 0.1186,
275
+ "step": 3600
276
+ },
277
+ {
278
+ "epoch": 16.74,
279
+ "eval_cer": 0.09190389616367109,
280
+ "eval_loss": 0.48351120948791504,
281
+ "eval_runtime": 22.4373,
282
+ "eval_samples_per_second": 77.46,
283
+ "eval_steps_per_second": 4.858,
284
+ "eval_wer": 0.20576721618790317,
285
+ "step": 3600
286
+ },
287
+ {
288
+ "epoch": 17.21,
289
+ "learning_rate": 0.00012920908379013312,
290
+ "loss": 0.118,
291
+ "step": 3700
292
+ },
293
+ {
294
+ "epoch": 17.67,
295
+ "learning_rate": 0.00012451057165231008,
296
+ "loss": 0.1098,
297
+ "step": 3800
298
+ },
299
+ {
300
+ "epoch": 18.14,
301
+ "learning_rate": 0.00011981205951448707,
302
+ "loss": 0.1048,
303
+ "step": 3900
304
+ },
305
+ {
306
+ "epoch": 18.6,
307
+ "learning_rate": 0.00011511354737666405,
308
+ "loss": 0.1051,
309
+ "step": 4000
310
+ },
311
+ {
312
+ "epoch": 19.07,
313
+ "learning_rate": 0.00011041503523884103,
314
+ "loss": 0.1014,
315
+ "step": 4100
316
+ },
317
+ {
318
+ "epoch": 19.53,
319
+ "learning_rate": 0.000105716523101018,
320
+ "loss": 0.0966,
321
+ "step": 4200
322
+ },
323
+ {
324
+ "epoch": 19.53,
325
+ "eval_cer": 0.08753135571767197,
326
+ "eval_loss": 0.49859338998794556,
327
+ "eval_runtime": 22.5733,
328
+ "eval_samples_per_second": 76.994,
329
+ "eval_steps_per_second": 4.829,
330
+ "eval_wer": 0.19782764514528312,
331
+ "step": 4200
332
+ },
333
+ {
334
+ "epoch": 20.0,
335
+ "learning_rate": 0.00010101801096319498,
336
+ "loss": 0.097,
337
+ "step": 4300
338
+ },
339
+ {
340
+ "epoch": 20.47,
341
+ "learning_rate": 9.631949882537196e-05,
342
+ "loss": 0.09,
343
+ "step": 4400
344
+ },
345
+ {
346
+ "epoch": 20.93,
347
+ "learning_rate": 9.162098668754895e-05,
348
+ "loss": 0.0904,
349
+ "step": 4500
350
+ },
351
+ {
352
+ "epoch": 21.4,
353
+ "learning_rate": 8.692247454972591e-05,
354
+ "loss": 0.0822,
355
+ "step": 4600
356
+ },
357
+ {
358
+ "epoch": 21.86,
359
+ "learning_rate": 8.222396241190289e-05,
360
+ "loss": 0.0869,
361
+ "step": 4700
362
+ },
363
+ {
364
+ "epoch": 22.33,
365
+ "learning_rate": 7.752545027407988e-05,
366
+ "loss": 0.083,
367
+ "step": 4800
368
+ },
369
+ {
370
+ "epoch": 22.33,
371
+ "eval_cer": 0.08620808689848802,
372
+ "eval_loss": 0.5179353952407837,
373
+ "eval_runtime": 22.6454,
374
+ "eval_samples_per_second": 76.748,
375
+ "eval_steps_per_second": 4.813,
376
+ "eval_wer": 0.192700005513591,
377
+ "step": 4800
378
+ },
379
+ {
380
+ "epoch": 22.79,
381
+ "learning_rate": 7.282693813625684e-05,
382
+ "loss": 0.0817,
383
+ "step": 4900
384
+ },
385
+ {
386
+ "epoch": 23.26,
387
+ "learning_rate": 6.812842599843383e-05,
388
+ "loss": 0.0774,
389
+ "step": 5000
390
+ },
391
+ {
392
+ "epoch": 23.72,
393
+ "learning_rate": 6.34299138606108e-05,
394
+ "loss": 0.0743,
395
+ "step": 5100
396
+ },
397
+ {
398
+ "epoch": 24.19,
399
+ "learning_rate": 5.873140172278778e-05,
400
+ "loss": 0.073,
401
+ "step": 5200
402
+ },
403
+ {
404
+ "epoch": 24.65,
405
+ "learning_rate": 5.4032889584964754e-05,
406
+ "loss": 0.0699,
407
+ "step": 5300
408
+ },
409
+ {
410
+ "epoch": 25.12,
411
+ "learning_rate": 4.933437744714174e-05,
412
+ "loss": 0.071,
413
+ "step": 5400
414
+ },
415
+ {
416
+ "epoch": 25.12,
417
+ "eval_cer": 0.08569028605619865,
418
+ "eval_loss": 0.5539018511772156,
419
+ "eval_runtime": 22.6635,
420
+ "eval_samples_per_second": 76.687,
421
+ "eval_steps_per_second": 4.81,
422
+ "eval_wer": 0.19082538457297238,
423
+ "step": 5400
424
+ },
425
+ {
426
+ "epoch": 25.58,
427
+ "learning_rate": 4.463586530931871e-05,
428
+ "loss": 0.0706,
429
+ "step": 5500
430
+ },
431
+ {
432
+ "epoch": 26.05,
433
+ "learning_rate": 3.993735317149569e-05,
434
+ "loss": 0.0696,
435
+ "step": 5600
436
+ },
437
+ {
438
+ "epoch": 26.51,
439
+ "learning_rate": 3.523884103367267e-05,
440
+ "loss": 0.0702,
441
+ "step": 5700
442
+ },
443
+ {
444
+ "epoch": 26.98,
445
+ "learning_rate": 3.0540328895849644e-05,
446
+ "loss": 0.0656,
447
+ "step": 5800
448
+ },
449
+ {
450
+ "epoch": 27.44,
451
+ "learning_rate": 2.584181675802662e-05,
452
+ "loss": 0.0662,
453
+ "step": 5900
454
+ },
455
+ {
456
+ "epoch": 27.91,
457
+ "learning_rate": 2.1143304620203598e-05,
458
+ "loss": 0.0648,
459
+ "step": 6000
460
+ },
461
+ {
462
+ "epoch": 27.91,
463
+ "eval_cer": 0.08437852392239892,
464
+ "eval_loss": 0.5582728981971741,
465
+ "eval_runtime": 22.4539,
466
+ "eval_samples_per_second": 77.403,
467
+ "eval_steps_per_second": 4.854,
468
+ "eval_wer": 0.18702100678171693,
469
+ "step": 6000
470
+ },
471
+ {
472
+ "epoch": 28.37,
473
+ "learning_rate": 1.6444792482380578e-05,
474
+ "loss": 0.0622,
475
+ "step": 6100
476
+ },
477
+ {
478
+ "epoch": 28.84,
479
+ "learning_rate": 1.1746280344557556e-05,
480
+ "loss": 0.0602,
481
+ "step": 6200
482
+ },
483
+ {
484
+ "epoch": 29.3,
485
+ "learning_rate": 7.047768206734534e-06,
486
+ "loss": 0.06,
487
+ "step": 6300
488
+ },
489
+ {
490
+ "epoch": 29.77,
491
+ "learning_rate": 2.349256068911511e-06,
492
+ "loss": 0.0615,
493
+ "step": 6400
494
+ },
495
+ {
496
+ "epoch": 30.0,
497
+ "step": 6450,
498
+ "total_flos": 4.472815672614715e+19,
499
+ "train_loss": 0.3642665664539781,
500
+ "train_runtime": 8377.9265,
501
+ "train_samples_per_second": 49.251,
502
+ "train_steps_per_second": 0.77
503
+ }
504
+ ],
505
+ "logging_steps": 100,
506
+ "max_steps": 6450,
507
+ "num_train_epochs": 30,
508
+ "save_steps": 600,
509
+ "total_flos": 4.472815672614715e+19,
510
+ "trial_name": null,
511
+ "trial_params": null
512
+ }
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8dc5ca704c2a09f0fa35a3898e0f40a7516ae8d96ac92fa615bf691eda5db270
3
+ size 4536
vocab.json ADDED
@@ -0,0 +1,33 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "\"": 1,
3
+ "-": 2,
4
+ "[PAD]": 30,
5
+ "[UNK]": 29,
6
+ "a": 3,
7
+ "b": 4,
8
+ "c": 5,
9
+ "d": 6,
10
+ "e": 7,
11
+ "f": 8,
12
+ "g": 9,
13
+ "h": 10,
14
+ "i": 11,
15
+ "j": 12,
16
+ "k": 13,
17
+ "l": 14,
18
+ "m": 15,
19
+ "n": 16,
20
+ "o": 17,
21
+ "p": 18,
22
+ "q": 19,
23
+ "r": 20,
24
+ "s": 21,
25
+ "t": 22,
26
+ "u": 23,
27
+ "v": 24,
28
+ "w": 25,
29
+ "x": 26,
30
+ "y": 27,
31
+ "z": 28,
32
+ "|": 0
33
+ }