Hugging Face's logo

kadriu
/
imja

Automatic Speech Recognition
Transformers
PyTorch
wav2vec2
Generated from Trainer
Inference Endpoints
Model card Files Community
1
kadriu commited on
Commit
cc457f3
·
1 Parent(s): 894de66

Upload 9 files

Browse files
Files changed (9) hide show
  1. README.md +110 -0
  2. added_tokens.json +1 -0
  3. config.json +116 -0
  4. preprocessor_config.json +9 -0
  5. pytorch_model.bin +3 -0
  6. special_tokens_map.json +1 -0
  7. tokenizer_config.json +1 -0
  8. training_args.bin +3 -0
  9. vocab.json +1 -0
README.md ADDED
@@ -0,0 +1,110 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ tags:
4
+ - generated_from_trainer
5
+ model-index:
6
+ - name: wav2vec2-large-xlsr-53-Total2e-4_4
7
+ results: []
8
+ ---
9
+
10
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
11
+ should probably proofread and complete it, then remove this comment. -->
12
+
13
+ # wav2vec2-large-xlsr-53-Total2e-4_4
14
+
15
+ This model is a fine-tuned version of [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on the None dataset.
16
+ It achieves the following results on the evaluation set:
17
+ - Loss: 0.2474
18
+ - Wer: 0.1951
19
+
20
+ ## Model description
21
+
22
+ More information needed
23
+
24
+ ## Intended uses & limitations
25
+
26
+ More information needed
27
+
28
+ ## Training and evaluation data
29
+
30
+ More information needed
31
+
32
+ ## Training procedure
33
+
34
+ ### Training hyperparameters
35
+
36
+ The following hyperparameters were used during training:
37
+ - learning_rate: 0.0002
38
+ - train_batch_size: 8
39
+ - eval_batch_size: 8
40
+ - seed: 42
41
+ - gradient_accumulation_steps: 2
42
+ - total_train_batch_size: 16
43
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
44
+ - lr_scheduler_type: linear
45
+ - lr_scheduler_warmup_steps: 500
46
+ - num_epochs: 5
47
+ - mixed_precision_training: Native AMP
48
+
49
+ ### Training results
50
+
51
+ | Training Loss | Epoch | Step | Validation Loss | Wer |
52
+ |:-------------:|:-----:|:-----:|:---------------:|:------:|
53
+ | 5.5015 | 0.1 | 200 | 2.9261 | 0.9707 |
54
+ | 2.9197 | 0.2 | 400 | 2.7757 | 0.9707 |
55
+ | 1.7594 | 0.3 | 600 | 0.6117 | 0.5746 |
56
+ | 1.0908 | 0.4 | 800 | 0.4673 | 0.4530 |
57
+ | 0.9441 | 0.5 | 1000 | 0.4142 | 0.4010 |
58
+ | 0.8688 | 0.6 | 1200 | 0.3909 | 0.3675 |
59
+ | 0.849 | 0.7 | 1400 | 0.3649 | 0.3360 |
60
+ | 0.8223 | 0.8 | 1600 | 0.3532 | 0.3334 |
61
+ | 0.821 | 0.9 | 1800 | 0.3513 | 0.3185 |
62
+ | 0.7839 | 1.0 | 2000 | 0.3373 | 0.3039 |
63
+ | 0.714 | 1.1 | 2200 | 0.3210 | 0.2922 |
64
+ | 0.7129 | 1.2 | 2400 | 0.3216 | 0.2860 |
65
+ | 0.7076 | 1.3 | 2600 | 0.3279 | 0.2843 |
66
+ | 0.73 | 1.4 | 2800 | 0.3111 | 0.2662 |
67
+ | 0.7256 | 1.5 | 3000 | 0.3032 | 0.2625 |
68
+ | 0.72 | 1.6 | 3200 | 0.3066 | 0.2571 |
69
+ | 0.6754 | 1.7 | 3400 | 0.2999 | 0.2581 |
70
+ | 0.6859 | 1.8 | 3600 | 0.2935 | 0.2562 |
71
+ | 0.6966 | 1.9 | 3800 | 0.2858 | 0.2469 |
72
+ | 0.6791 | 2.0 | 4000 | 0.2857 | 0.2393 |
73
+ | 0.6412 | 2.1 | 4200 | 0.2815 | 0.2392 |
74
+ | 0.6356 | 2.2 | 4400 | 0.2836 | 0.2343 |
75
+ | 0.6048 | 2.3 | 4600 | 0.2824 | 0.2422 |
76
+ | 0.6473 | 2.4 | 4800 | 0.2805 | 0.2316 |
77
+ | 0.659 | 2.5 | 5000 | 0.2775 | 0.2262 |
78
+ | 0.6412 | 2.6 | 5200 | 0.2729 | 0.2249 |
79
+ | 0.6167 | 2.7 | 5400 | 0.2719 | 0.2227 |
80
+ | 0.6226 | 2.8 | 5600 | 0.2661 | 0.2193 |
81
+ | 0.6168 | 2.9 | 5800 | 0.2615 | 0.2172 |
82
+ | 0.6145 | 3.0 | 6000 | 0.2608 | 0.2148 |
83
+ | 0.593 | 3.1 | 6200 | 0.2643 | 0.2123 |
84
+ | 0.5919 | 3.2 | 6400 | 0.2617 | 0.2131 |
85
+ | 0.6115 | 3.3 | 6600 | 0.2589 | 0.2114 |
86
+ | 0.5859 | 3.4 | 6800 | 0.2591 | 0.2100 |
87
+ | 0.5919 | 3.5 | 7000 | 0.2564 | 0.2103 |
88
+ | 0.5873 | 3.6 | 7200 | 0.2572 | 0.2074 |
89
+ | 0.561 | 3.7 | 7400 | 0.2561 | 0.2056 |
90
+ | 0.5808 | 3.8 | 7600 | 0.2538 | 0.2062 |
91
+ | 0.5701 | 3.9 | 7800 | 0.2517 | 0.2029 |
92
+ | 0.5722 | 4.0 | 8000 | 0.2523 | 0.2007 |
93
+ | 0.5508 | 4.1 | 8200 | 0.2570 | 0.2023 |
94
+ | 0.5591 | 4.2 | 8400 | 0.2502 | 0.2029 |
95
+ | 0.5697 | 4.3 | 8600 | 0.2478 | 0.1991 |
96
+ | 0.5689 | 4.4 | 8800 | 0.2492 | 0.2021 |
97
+ | 0.5345 | 4.5 | 9000 | 0.2498 | 0.2005 |
98
+ | 0.5726 | 4.6 | 9200 | 0.2492 | 0.1983 |
99
+ | 0.5382 | 4.7 | 9400 | 0.2487 | 0.1974 |
100
+ | 0.5614 | 4.8 | 9600 | 0.2481 | 0.1957 |
101
+ | 0.5568 | 4.9 | 9800 | 0.2477 | 0.1955 |
102
+ | 0.5631 | 5.0 | 10000 | 0.2474 | 0.1951 |
103
+
104
+
105
+ ### Framework versions
106
+
107
+ - Transformers 4.15.0
108
+ - Pytorch 1.10.0+cu111
109
+ - Datasets 1.18.3
110
+ - Tokenizers 0.10.3
added_tokens.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"<s>": 30, "</s>": 31}
config.json ADDED
@@ -0,0 +1,116 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "facebook/wav2vec2-large-xlsr-53",
3
+ "activation_dropout": 0.0,
4
+ "adapter_kernel_size": 3,
5
+ "adapter_stride": 2,
6
+ "add_adapter": false,
7
+ "apply_spec_augment": true,
8
+ "architectures": [
9
+ "Wav2Vec2ForCTC"
10
+ ],
11
+ "attention_dropout": 0.1,
12
+ "bos_token_id": 1,
13
+ "classifier_proj_size": 256,
14
+ "codevector_dim": 768,
15
+ "contrastive_logits_temperature": 0.1,
16
+ "conv_bias": true,
17
+ "conv_dim": [
18
+ 512,
19
+ 512,
20
+ 512,
21
+ 512,
22
+ 512,
23
+ 512,
24
+ 512
25
+ ],
26
+ "conv_kernel": [
27
+ 10,
28
+ 3,
29
+ 3,
30
+ 3,
31
+ 3,
32
+ 2,
33
+ 2
34
+ ],
35
+ "conv_stride": [
36
+ 5,
37
+ 2,
38
+ 2,
39
+ 2,
40
+ 2,
41
+ 2,
42
+ 2
43
+ ],
44
+ "ctc_loss_reduction": "mean",
45
+ "ctc_zero_infinity": true,
46
+ "diversity_loss_weight": 0.1,
47
+ "do_stable_layer_norm": true,
48
+ "eos_token_id": 2,
49
+ "feat_extract_activation": "gelu",
50
+ "feat_extract_dropout": 0.0,
51
+ "feat_extract_norm": "layer",
52
+ "feat_proj_dropout": 0.1,
53
+ "feat_quantizer_dropout": 0.0,
54
+ "final_dropout": 0.0,
55
+ "gradient_checkpointing": false,
56
+ "hidden_act": "gelu",
57
+ "hidden_dropout": 0.1,
58
+ "hidden_size": 1024,
59
+ "initializer_range": 0.02,
60
+ "intermediate_size": 4096,
61
+ "layer_norm_eps": 1e-05,
62
+ "layerdrop": 0.05,
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": 0.1,
76
+ "mask_time_selection": "static",
77
+ "model_type": "wav2vec2",
78
+ "num_adapter_layers": 3,
79
+ "num_attention_heads": 16,
80
+ "num_codevector_groups": 2,
81
+ "num_codevectors_per_group": 320,
82
+ "num_conv_pos_embedding_groups": 16,
83
+ "num_conv_pos_embeddings": 128,
84
+ "num_feat_extract_layers": 7,
85
+ "num_hidden_layers": 24,
86
+ "num_negatives": 100,
87
+ "output_hidden_size": 1024,
88
+ "pad_token_id": 29,
89
+ "proj_codevector_dim": 768,
90
+ "tdnn_dilation": [
91
+ 1,
92
+ 2,
93
+ 3,
94
+ 1,
95
+ 1
96
+ ],
97
+ "tdnn_dim": [
98
+ 512,
99
+ 512,
100
+ 512,
101
+ 512,
102
+ 1500
103
+ ],
104
+ "tdnn_kernel": [
105
+ 5,
106
+ 3,
107
+ 3,
108
+ 1,
109
+ 1
110
+ ],
111
+ "torch_dtype": "float32",
112
+ "transformers_version": "4.15.0",
113
+ "use_weighted_layer_sum": false,
114
+ "vocab_size": 32,
115
+ "xvector_output_dim": 512
116
+ }
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,
7
+ "return_attention_mask": true,
8
+ "sampling_rate": 16000
9
+ }
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:63af4a0ac4bf1792debdfc2722a881cefde059168ae80439210e8f6bc64acfca
3
+ size 1262054897
special_tokens_map.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"bos_token": "<s>", "eos_token": "</s>", "unk_token": "[UNK]", "pad_token": "[PAD]", "additional_special_tokens": [{"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "<s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}, {"content": "</s>", "single_word": false, "lstrip": false, "rstrip": false, "normalized": true}]}
tokenizer_config.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"unk_token": "[UNK]", "bos_token": "<s>", "eos_token": "</s>", "pad_token": "[PAD]", "do_lower_case": false, "word_delimiter_token": "|", "special_tokens_map_file": null, "tokenizer_file": null, "name_or_path": "akadriu/wav2vec2-large-xlsr-53-Total_2e-4_2", "tokenizer_class": "Wav2Vec2CTCTokenizer"}
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:53bbc3262635cdef4102e5d5ddc069810ce5705c242ce47aaa67efbda3b16414
3
+ size 2991
vocab.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"e": 0, "ç": 1, "n": 3, "c": 4, "k": 5, "y": 6, "a": 7, "ë": 8, "z": 9, "j": 10, "r": 11, "g": 12, "b": 13, "v": 14, "h": 15, "o": 16, "u": 17, "s": 18, "f": 19, "d": 20, "l": 21, "q": 22, "x": 23, "m": 24, "t": 25, "p": 26, "i": 27, "|": 2, "[UNK]": 28, "[PAD]": 29}