jonatasgrosman commited on
Commit
1374a29
1 Parent(s): 4c59763

update model

Browse files
Files changed (5) hide show
  1. README.md +17 -10
  2. config.json +19 -4
  3. preprocessor_config.json +1 -0
  4. pytorch_model.bin +2 -2
  5. vocab.json +1 -1
README.md CHANGED
@@ -24,10 +24,10 @@ model-index:
24
  metrics:
25
  - name: Test WER
26
  type: wer
27
- value: 13.60
28
  - name: Test CER
29
  type: cer
30
- value: 4.45
31
  ---
32
 
33
  # Wav2Vec2-Large-XLSR-53-Dutch
@@ -49,7 +49,7 @@ from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
49
 
50
  LANG_ID = "nl"
51
  MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-dutch"
52
- SAMPLES = 5
53
 
54
  test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
55
 
@@ -81,11 +81,16 @@ for i, predicted_sentence in enumerate(predicted_sentences):
81
 
82
  | Reference | Prediction |
83
  | ------------- | ------------- |
84
- | DE ABORIGINALS ZIJN DE OORSPRONKELIJKE BEWONERS VAN AUSTRALIË. | DE ABORIGONALS ZIJN DE OORSPRONKELIJKE BEWONERS VAN AUSTRALIË |
85
- | MIJN TOETSENBORD ZIT VOL STOF | MIJN TOETSEN BORT ZIT VOL STOF. |
86
  | ZE HAD DE BANK BESCHADIGD MET HAAR SKATEBOARD. | ZE HAD DE BANK BESCHADIGD MET HAAR SCHEETBOORD |
87
- | WAAR LAAT JIJ JE ONDERHOUD DOEN? | WAAR LAAT JIJ JE ONDERHOUD DOEN |
88
- | NA HET LEZEN VAN VELE BEOORDELINGEN HAD ZE EINDELIJK HAAR OOG LATEN VALLEN OP EEN LAPTOP MET EEN QWERTY TOETSENBORD. | NA HET LEZEN VAN VELE BEOORDELINGEN HAD ZE EINDELIJK HAAR OOG LATEN VALLEN OP EEN LAPTOP MET EEN KWERTIETOETSENBORD |
 
 
 
 
 
89
 
90
  ## Evaluation
91
 
@@ -102,9 +107,11 @@ LANG_ID = "nl"
102
  MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-dutch"
103
  DEVICE = "cuda"
104
 
105
- CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
106
  "؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
107
- "=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。"]
 
 
108
 
109
  test_dataset = load_dataset("common_voice", LANG_ID, split="test")
110
 
@@ -156,7 +163,7 @@ In the table below I report the Word Error Rate (WER) and the Character Error Ra
156
 
157
  | Model | WER | CER |
158
  | ------------- | ------------- | ------------- |
159
- | jonatasgrosman/wav2vec2-large-xlsr-53-dutch | **13.60%** | **4.45%** |
160
  | wietsedv/wav2vec2-large-xlsr-53-dutch | 16.78% | 5.60% |
161
  | facebook/wav2vec2-large-xlsr-53-dutch | 20.97% | 7.24% |
162
  | nithinholla/wav2vec2-large-xlsr-53-dutch | 21.39% | 7.29% |
 
24
  metrics:
25
  - name: Test WER
26
  type: wer
27
+ value: 15.76
28
  - name: Test CER
29
  type: cer
30
+ value: 5.50
31
  ---
32
 
33
  # Wav2Vec2-Large-XLSR-53-Dutch
 
49
 
50
  LANG_ID = "nl"
51
  MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-dutch"
52
+ SAMPLES = 10
53
 
54
  test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
55
 
 
81
 
82
  | Reference | Prediction |
83
  | ------------- | ------------- |
84
+ | DE ABORIGINALS ZIJN DE OORSPRONKELIJKE BEWONERS VAN AUSTRALIË. | DE ABBORIGENALS ZIJN DE OORSPRONKELIJKE BEWONERS VAN AUSTRALIË |
85
+ | MIJN TOETSENBORD ZIT VOL STOF. | MIJN TOETSENBORD ZIT VOL STOF |
86
  | ZE HAD DE BANK BESCHADIGD MET HAAR SKATEBOARD. | ZE HAD DE BANK BESCHADIGD MET HAAR SCHEETBOORD |
87
+ | WAAR LAAT JIJ JE ONDERHOUD DOEN? | WAAR LAAT JIJ HET ONDERHOUD DOEN |
88
+ | NA HET LEZEN VAN VELE BEOORDELINGEN HAD ZE EINDELIJK HAAR OOG LATEN VALLEN OP EEN LAPTOP MET EEN QWERTY TOETSENBORD. | NA HET LEZEN VAN VELE BEOORDELINGEN HAD ZE EINDELIJK HAAR OOG LATEN VALLEN OP EEN LAPTOP MET EEN QUERTITOETSEMBORD |
89
+ | DE TAMPONS ZIJN OP. | DE TAPONT ZIJN OP |
90
+ | MARIJKE KENT OLIVIER NU AL MEER DAN TWEE JAAR. | MAARRIJKEN KENT OLIEVIER NU AL MEER DAN TWEE JAAR |
91
+ | HET VOEREN VAN BROOD AAN EENDEN IS EIGENLIJK ONGEZOND VOOR DE BEESTEN. | HET VOEREN VAN BEUROT AAN EINDEN IS EIGENLIJK ONGEZOND VOOR DE BEESTEN |
92
+ | PARKET MOET JE STOFZUIGEN, TEGELS MOET JE DWEILEN. | PARKET MOET JE STOF ZUIGEN MAAR TEGELS MOET JE DWEILEN |
93
+ | IN ONZE BUURT KENT IEDEREEN ELKAAR. | IN ONZE BUURT KENT IEDEREEN ELKAAR |
94
 
95
  ## Evaluation
96
 
 
107
  MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-dutch"
108
  DEVICE = "cuda"
109
 
110
+ CHARS_TO_IGNORE = [",", "?", "¿", ".", "!", "¡", ";", ";", ":", '""', "%", '"', "�", "ʿ", "·", "჻", "~", "՞",
111
  "؟", "،", "।", "॥", "«", "»", "„", "“", "”", "「", "」", "‘", "’", "《", "》", "(", ")", "[", "]",
112
+ "{", "}", "=", "`", "_", "+", "<", ">", "…", "–", "°", "´", "ʾ", "‹", "›", "©", "®", "—", "→", "。",
113
+ "、", "﹂", "﹁", "‧", "~", "﹏", ",", "{", "}", "(", ")", "[", "]", "【", "】", "‥", "〽",
114
+ "『", "』", "〝", "〟", "⟨", "⟩", "〜", ":", "!", "?", "♪", "؛", "/", "\\", "º", "−", "^", "ʻ", "ˆ"]
115
 
116
  test_dataset = load_dataset("common_voice", LANG_ID, split="test")
117
 
 
163
 
164
  | Model | WER | CER |
165
  | ------------- | ------------- | ------------- |
166
+ | jonatasgrosman/wav2vec2-large-xlsr-53-dutch | **15.76%** | **5.50%** |
167
  | wietsedv/wav2vec2-large-xlsr-53-dutch | 16.78% | 5.60% |
168
  | facebook/wav2vec2-large-xlsr-53-dutch | 20.97% | 7.24% |
169
  | nithinholla/wav2vec2-large-xlsr-53-dutch | 21.39% | 7.29% |
config.json CHANGED
@@ -7,6 +7,8 @@
7
  ],
8
  "attention_dropout": 0.1,
9
  "bos_token_id": 1,
 
 
10
  "conv_bias": true,
11
  "conv_dim": [
12
  512,
@@ -37,33 +39,46 @@
37
  ],
38
  "ctc_loss_reduction": "mean",
39
  "ctc_zero_infinity": true,
 
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": 0.05,
46
- "final_dropout": 0.1,
 
47
  "gradient_checkpointing": true,
48
  "hidden_act": "gelu",
49
  "hidden_dropout": 0.05,
50
- "hidden_dropout_prob": 0.1,
51
  "hidden_size": 1024,
52
  "initializer_range": 0.02,
53
  "intermediate_size": 4096,
54
  "layer_norm_eps": 1e-05,
55
  "layerdrop": 0.05,
 
 
 
 
 
56
  "mask_feature_length": 10,
57
  "mask_feature_prob": 0.0,
58
  "mask_time_length": 10,
 
 
59
  "mask_time_prob": 0.05,
 
60
  "model_type": "wav2vec2",
61
  "num_attention_heads": 16,
 
 
62
  "num_conv_pos_embedding_groups": 16,
63
  "num_conv_pos_embeddings": 128,
64
  "num_feat_extract_layers": 7,
65
  "num_hidden_layers": 24,
 
66
  "pad_token_id": 0,
67
- "transformers_version": "4.5.0.dev0",
68
- "vocab_size": 50
 
69
  }
 
7
  ],
8
  "attention_dropout": 0.1,
9
  "bos_token_id": 1,
10
+ "codevector_dim": 768,
11
+ "contrastive_logits_temperature": 0.1,
12
  "conv_bias": true,
13
  "conv_dim": [
14
  512,
 
39
  ],
40
  "ctc_loss_reduction": "mean",
41
  "ctc_zero_infinity": true,
42
+ "diversity_loss_weight": 0.1,
43
  "do_stable_layer_norm": true,
44
  "eos_token_id": 2,
45
  "feat_extract_activation": "gelu",
46
  "feat_extract_dropout": 0.0,
47
  "feat_extract_norm": "layer",
48
  "feat_proj_dropout": 0.05,
49
+ "feat_quantizer_dropout": 0.0,
50
+ "final_dropout": 0.0,
51
  "gradient_checkpointing": true,
52
  "hidden_act": "gelu",
53
  "hidden_dropout": 0.05,
 
54
  "hidden_size": 1024,
55
  "initializer_range": 0.02,
56
  "intermediate_size": 4096,
57
  "layer_norm_eps": 1e-05,
58
  "layerdrop": 0.05,
59
+ "mask_channel_length": 10,
60
+ "mask_channel_min_space": 1,
61
+ "mask_channel_other": 0.0,
62
+ "mask_channel_prob": 0.0,
63
+ "mask_channel_selection": "static",
64
  "mask_feature_length": 10,
65
  "mask_feature_prob": 0.0,
66
  "mask_time_length": 10,
67
+ "mask_time_min_space": 1,
68
+ "mask_time_other": 0.0,
69
  "mask_time_prob": 0.05,
70
+ "mask_time_selection": "static",
71
  "model_type": "wav2vec2",
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
  "pad_token_id": 0,
81
+ "proj_codevector_dim": 768,
82
+ "transformers_version": "4.7.0.dev0",
83
+ "vocab_size": 39
84
  }
preprocessor_config.json CHANGED
@@ -1,5 +1,6 @@
1
  {
2
  "do_normalize": true,
 
3
  "feature_size": 1,
4
  "padding_side": "right",
5
  "padding_value": 0.0,
 
1
  {
2
  "do_normalize": true,
3
+ "feature_extractor_type": "Wav2Vec2FeatureExtractor",
4
  "feature_size": 1,
5
  "padding_side": "right",
6
  "padding_value": 0.0,
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:391fd46405658415cd0f6a93b77443052f5931cf2dca07756f7c109135feeafe
3
- size 1262138839
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f307fd2a7a323dbc4dc6fda728e97d51c0be50fed198cb4d1a14d275acb255cd
3
+ size 1262093720
vocab.json CHANGED
@@ -1 +1 @@
1
- {"<pad>": 0, "<s>": 1, "</s>": 2, "<unk>": 3, "E": 4, "N": 5, "A": 6, "I": 7, "T": 8, "O": 9, "D": 10, "R": 11, "|": 12, "L": 13, "S": 14, "H": 15, "G": 16, "M": 17, "K": 18, "V": 19, "J": 20, "W": 21, "Z": 22, "U": 23, "B": 24, "C": 25, "P": 26, "F": 27, "Y": 28, "É": 29, "'": 30, "X": 31, "Ë": 32, "Q": 33, "-": 34, "Ê": 35, "À": 36, "Ä": 37, "È": 38, "Ï": 39, "Â": 40, "Û": 41, "Ö": 42, "Ô": 43, "Ü": 44, "Î": 45, "Ç": 46, "Æ": 47, "Ù": 48, "Œ": 49}
 
1
+ {"<pad>": 0, "<s>": 1, "</s>": 2, "<unk>": 3, "|": 4, "'": 5, "-": 6, "A": 7, "B": 8, "C": 9, "D": 10, "E": 11, "F": 12, "G": 13, "H": 14, "I": 15, "J": 16, "K": 17, "L": 18, "M": 19, "N": 20, "O": 21, "P": 22, "Q": 23, "R": 24, "S": 25, "T": 26, "U": 27, "V": 28, "W": 29, "X": 30, "Y": 31, "Z": 32, "É": 33, "Ë": 34, "Ï": 35, "Ó": 36, "Ö": 37, "Ü": 38}