ilyes rebai commited on
Commit
f154900
1 Parent(s): 472d9af

add new models with high performance

Browse files

README.md CHANGED
@@ -23,6 +23,9 @@ model-index:
23
  value: 20.89%
24
  ---
25
  ## Evaluation on Common Voice FR Test
 
 
 
26
  ```python
27
  import torch
28
  import torchaudio
@@ -73,6 +76,20 @@ print(wer.compute(predictions=result["predicted"], references=result["target"]))
73
 
74
  ## Results
75
 
 
 
76
  WER=18.29%
77
 
78
  SER=71.44%
 
 
 
 
 
 
 
 
 
 
 
 
23
  value: 20.89%
24
  ---
25
  ## Evaluation on Common Voice FR Test
26
+ The script used for training and evaluation can be found here: https://github.com/irebai/wav2vec2
27
+
28
+
29
  ```python
30
  import torch
31
  import torchaudio
76
 
77
  ## Results
78
 
79
+ # v0.1
80
+
81
  WER=18.29%
82
 
83
  SER=71.44%
84
+
85
+ # v1.0
86
+
87
+ WER=15.97%
88
+
89
+ CER=5.51%
90
+
91
+ # v2.0
92
+
93
+ WER=14.71%
94
+
95
+ CER=5.06%
config.json → v0.1/config.json RENAMED
File without changes
preprocessor_config.json → v0.1/preprocessor_config.json RENAMED
File without changes
pytorch_model.bin → v0.1/pytorch_model.bin RENAMED
File without changes
special_tokens_map.json → v0.1/special_tokens_map.json RENAMED
File without changes
tokenizer_config.json → v0.1/tokenizer_config.json RENAMED
File without changes
vocab.json → v0.1/vocab.json RENAMED
File without changes
v1.0/config.json ADDED
@@ -0,0 +1,78 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "/workspace/output_models/wav2vec2-large-xlsr-53/checkpoint-71900",
3
+ "activation_dropout": 0.055,
4
+ "apply_spec_augment": true,
5
+ "architectures": [
6
+ "Wav2Vec2ForCTC"
7
+ ],
8
+ "attention_dropout": 0.094,
9
+ "bos_token_id": 1,
10
+ "conv_bias": true,
11
+ "conv_dim": [
12
+ 512,
13
+ 512,
14
+ 512,
15
+ 512,
16
+ 512,
17
+ 512,
18
+ 512
19
+ ],
20
+ "conv_kernel": [
21
+ 10,
22
+ 3,
23
+ 3,
24
+ 3,
25
+ 3,
26
+ 2,
27
+ 2
28
+ ],
29
+ "conv_stride": [
30
+ 5,
31
+ 2,
32
+ 2,
33
+ 2,
34
+ 2,
35
+ 2,
36
+ 2
37
+ ],
38
+ "ctc_loss_reduction": "mean",
39
+ "ctc_zero_infinity": false,
40
+ "do_stable_layer_norm": true,
41
+ "eos_token_id": 2,
42
+ "feat_extract_activation": "gelu",
43
+ "feat_extract_dropout": 0.0,
44
+ "feat_extract_norm": "layer",
45
+ "feat_proj_dropout": 0.04,
46
+ "final_dropout": 0.0,
47
+ "gradient_checkpointing": true,
48
+ "hidden_act": "gelu",
49
+ "hidden_dropout": 0.047,
50
+ "hidden_size": 1024,
51
+ "initializer_range": 0.02,
52
+ "intermediate_size": 4096,
53
+ "layer_norm_eps": 1e-05,
54
+ "layerdrop": 0.041,
55
+ "mask_channel_length": 10,
56
+ "mask_channel_min_space": 1,
57
+ "mask_channel_other": 0.0,
58
+ "mask_channel_prob": 0.0,
59
+ "mask_channel_selection": "static",
60
+ "mask_feature_length": 10,
61
+ "mask_feature_prob": 0.0,
62
+ "mask_time_length": 10,
63
+ "mask_time_min_space": 1,
64
+ "mask_time_other": 0.0,
65
+ "mask_time_prob": 0.4,
66
+ "mask_time_selection": "static",
67
+ "model_type": "wav2vec2",
68
+ "num_attention_heads": 16,
69
+ "num_conv_pos_embedding_groups": 16,
70
+ "num_conv_pos_embeddings": 128,
71
+ "num_feat_extract_layers": 7,
72
+ "num_hidden_layers": 24,
73
+ "pad_token_id": 46,
74
+ "pooling_type": "max",
75
+ "time_pooling_size": 1,
76
+ "transformers_version": "4.5.0.dev0",
77
+ "vocab_size": 47
78
+ }
v1.0/preprocessor_config.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_normalize": true,
3
+ "feature_size": 1,
4
+ "padding_side": "right",
5
+ "padding_value": 0.0,
6
+ "return_attention_mask": true,
7
+ "sampling_rate": 16000
8
+ }
v1.0/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0922012f44439677c60c929ef48938ec2c3ae6ec717523cae13988a2d22bfd99
3
+ size 1262144708
v1.0/results.txt ADDED
@@ -0,0 +1,2 @@
 
 
1
+ WER=15.97816949508748
2
+ CER=5.518223907734944
v1.0/special_tokens_map.json ADDED
@@ -0,0 +1 @@
 
1
+ {"unk_token": "<unk>", "pad_token": "<pad>"}
v1.0/text.txt ADDED
The diff for this file is too large to render. See raw diff
v1.0/tokenizer_config.json ADDED
@@ -0,0 +1 @@
 
1
+ {"unk_token": "<unk>", "bos_token": null, "eos_token": null, "pad_token": "<pad>", "do_lower_case": false, "word_delimiter_token": "|"}
v1.0/trans.txt ADDED
The diff for this file is too large to render. See raw diff
v1.0/vocab.json ADDED
@@ -0,0 +1 @@
 
1
+ {"a": 0, "e": 1, "i": 2, "o": 3, "u": 4, "y": 5, "b": 6, "c": 7, "d": 8, "f": 9, "g": 10, "h": 11, "j": 12, "k": 13, "l": 14, "m": 15, "n": 16, "p": 17, "q": 18, "r": 19, "s": 20, "t": 21, "v": 22, "w": 23, "x": 24, "z": 25, "à": 26, "â": 27, "æ": 28, "ç": 29, "è": 30, "é": 31, "ê": 32, "ë": 33, "î": 34, "ï": 35, "ô": 36, "œ": 37, "ù": 38, "û": 39, "ü": 40, "ÿ": 41, "'": 42, "-": 43, "|": 44, "<unk>": 45, "<pad>": 46}
v2.0/config.json ADDED
@@ -0,0 +1,78 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "facebook/wav2vec2-large-xlsr-53",
3
+ "activation_dropout": 0.055,
4
+ "apply_spec_augment": true,
5
+ "architectures": [
6
+ "Wav2Vec2ForCTC"
7
+ ],
8
+ "attention_dropout": 0.094,
9
+ "bos_token_id": 1,
10
+ "conv_bias": true,
11
+ "conv_dim": [
12
+ 512,
13
+ 512,
14
+ 512,
15
+ 512,
16
+ 512,
17
+ 512,
18
+ 512
19
+ ],
20
+ "conv_kernel": [
21
+ 10,
22
+ 3,
23
+ 3,
24
+ 3,
25
+ 3,
26
+ 2,
27
+ 2
28
+ ],
29
+ "conv_stride": [
30
+ 5,
31
+ 2,
32
+ 2,
33
+ 2,
34
+ 2,
35
+ 2,
36
+ 2
37
+ ],
38
+ "ctc_loss_reduction": "mean",
39
+ "ctc_zero_infinity": false,
40
+ "do_stable_layer_norm": true,
41
+ "eos_token_id": 2,
42
+ "feat_extract_activation": "gelu",
43
+ "feat_extract_dropout": 0.0,
44
+ "feat_extract_norm": "layer",
45
+ "feat_proj_dropout": 0.04,
46
+ "final_dropout": 0.0,
47
+ "gradient_checkpointing": true,
48
+ "hidden_act": "gelu",
49
+ "hidden_dropout": 0.047,
50
+ "hidden_size": 1024,
51
+ "initializer_range": 0.02,
52
+ "intermediate_size": 4096,
53
+ "layer_norm_eps": 1e-05,
54
+ "layerdrop": 0.041,
55
+ "mask_channel_length": 10,
56
+ "mask_channel_min_space": 1,
57
+ "mask_channel_other": 0.0,
58
+ "mask_channel_prob": 0.0,
59
+ "mask_channel_selection": "static",
60
+ "mask_feature_length": 10,
61
+ "mask_feature_prob": 0.0,
62
+ "mask_time_length": 10,
63
+ "mask_time_min_space": 1,
64
+ "mask_time_other": 0.0,
65
+ "mask_time_prob": 0.2,
66
+ "mask_time_selection": "static",
67
+ "model_type": "wav2vec2",
68
+ "num_attention_heads": 16,
69
+ "num_conv_pos_embedding_groups": 16,
70
+ "num_conv_pos_embeddings": 128,
71
+ "num_feat_extract_layers": 7,
72
+ "num_hidden_layers": 24,
73
+ "pad_token_id": 1,
74
+ "pooling_type": "max",
75
+ "time_pooling_size": 1,
76
+ "transformers_version": "4.5.0.dev0",
77
+ "vocab_size": 47
78
+ }
v2.0/preprocessor_config.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_normalize": true,
3
+ "feature_size": 1,
4
+ "padding_side": "right",
5
+ "padding_value": 0.0,
6
+ "return_attention_mask": true,
7
+ "sampling_rate": 16000
8
+ }
v2.0/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:536491b9fe83421402af342b175daba8ea082099c742492c4e5000403833b7bf
3
+ size 1262144708
v2.0/results.txt ADDED
@@ -0,0 +1,2 @@
 
 
1
+ WER=14.711290915777859
2
+ CER=5.068944791354929
v2.0/special_tokens_map.json ADDED
@@ -0,0 +1 @@
 
1
+ {"bos_token": "<bos>", "eos_token": "<eos>", "unk_token": "<unk>", "pad_token": "<pad>"}
v2.0/text.txt ADDED
The diff for this file is too large to render. See raw diff
v2.0/tokenizer_config.json ADDED
@@ -0,0 +1 @@
 
1
+ {"unk_token": "<unk>", "bos_token": "<bos>", "eos_token": "<eos>", "pad_token": "<pad>", "do_lower_case": false, "word_delimiter_token": "|"}
v2.0/trans.txt ADDED
The diff for this file is too large to render. See raw diff
v2.0/vocab.json ADDED
@@ -0,0 +1 @@
 
1
+ {"<unk>": 0, "<pad>": 1, "a": 2, "e": 3, "i": 4, "o": 5, "u": 6, "y": 7, "b": 8, "c": 9, "d": 10, "f": 11, "g": 12, "h": 13, "j": 14, "k": 15, "l": 16, "m": 17, "n": 18, "p": 19, "q": 20, "r": 21, "s": 22, "t": 23, "v": 24, "w": 25, "x": 26, "z": 27, "à": 28, "â": 29, "æ": 30, "ç": 31, "è": 32, "é": 33, "ê": 34, "ë": 35, "î": 36, "ï": 37, "ô": 38, "œ": 39, "ù": 40, "û": 41, "ü": 42, "ÿ": 43, "'": 44, "-": 45, "|": 46}