jonatasgrosman commited on
Commit
fbfdf44
1 Parent(s): b3c86a5

update model

Browse files
Files changed (5) hide show
  1. README.md +13 -6
  2. config.json +10 -2
  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: 11.90
28
  - name: Test CER
29
  type: cer
30
- value: 2.94
31
  ---
32
 
33
  # Wav2Vec2-Large-XLSR-53-Italian
@@ -49,7 +49,7 @@ from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
49
 
50
  LANG_ID = "it"
51
  MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-italian"
52
- SAMPLES = 5
53
 
54
  test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
55
 
@@ -86,6 +86,11 @@ for i, predicted_sentence in enumerate(predicted_sentences):
86
  | "FIN DALL'INIZIO LA SEDE EPISCOPALE È STATA IMMEDIATAMENTE SOGGETTA ALLA SANTA SEDE." | FIN DALL'INIZIO LA SEDE EPISCOPALE È STATA IMMEDIATAMENTE SOGGETTA ALLA SANTA SEDE |
87
  | IL VUOTO ASSOLUTO? | IL VUOTO ASSOLUTO |
88
  | DOPO ALCUNI ANNI, EGLI DECISE DI TORNARE IN INDIA PER RACCOGLIERE ALTRI INSEGNAMENTI. | DOPO ALCUNI ANNI EGLI DECISE DI TORNARE IN INDIA PER RACCOGLIERE ALTRI INSEGNAMENTI |
 
 
 
 
 
89
 
90
  ## Evaluation
91
 
@@ -102,9 +107,11 @@ LANG_ID = "it"
102
  MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-italian"
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-italian | **11.90%** | **2.94%** |
160
  | joorock12/wav2vec2-large-xlsr-italian | 12.60% | 3.18% |
161
  | gchhablani/wav2vec2-large-xlsr-it | 12.99% | 3.11% |
162
  | facebook/wav2vec2-large-xlsr-53-italian | 22.08% | 6.36% |
 
24
  metrics:
25
  - name: Test WER
26
  type: wer
27
+ value: 9.36
28
  - name: Test CER
29
  type: cer
30
+ value: 2.33
31
  ---
32
 
33
  # Wav2Vec2-Large-XLSR-53-Italian
 
49
 
50
  LANG_ID = "it"
51
  MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-italian"
52
+ SAMPLES = 10
53
 
54
  test_dataset = load_dataset("common_voice", LANG_ID, split=f"test[:{SAMPLES}]")
55
 
 
86
  | "FIN DALL'INIZIO LA SEDE EPISCOPALE È STATA IMMEDIATAMENTE SOGGETTA ALLA SANTA SEDE." | FIN DALL'INIZIO LA SEDE EPISCOPALE È STATA IMMEDIATAMENTE SOGGETTA ALLA SANTA SEDE |
87
  | IL VUOTO ASSOLUTO? | IL VUOTO ASSOLUTO |
88
  | DOPO ALCUNI ANNI, EGLI DECISE DI TORNARE IN INDIA PER RACCOGLIERE ALTRI INSEGNAMENTI. | DOPO ALCUNI ANNI EGLI DECISE DI TORNARE IN INDIA PER RACCOGLIERE ALTRI INSEGNAMENTI |
89
+ | SALVATION SUE | SALVATION SOO |
90
+ | IN QUESTO MODO, DECIO OTTENNE IL POTERE IMPERIALE. | IN QUESTO MODO DECHO OTTENNE IL POTERE IMPERIALE |
91
+ | SPARTA NOVARA ACQUISISCE IL TITOLO SPORTIVO PER GIOCARE IN PRIMA CATEGORIA. | PARCANOVARACFILISCE IL TITOLO SPORTIVO PER GIOCARE IN PRIMA CATEGORIA |
92
+ | IN SEGUITO, KYGO E SHEAR HANNO PROPOSTO DI CONTINUARE A LAVORARE SULLA CANZONE. | IN SEGUITO KIGO E SHIAR HANNO PROPOSTO DI CONTINUARE A LAVORARE SULLA CANZONE |
93
+ | ALAN CLARKE | ALAN CLARK |
94
 
95
  ## Evaluation
96
 
 
107
  MODEL_ID = "jonatasgrosman/wav2vec2-large-xlsr-53-italian"
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-italian | **9.36%** | **2.33%** |
167
  | joorock12/wav2vec2-large-xlsr-italian | 12.60% | 3.18% |
168
  | gchhablani/wav2vec2-large-xlsr-it | 12.99% | 3.11% |
169
  | facebook/wav2vec2-large-xlsr-53-italian | 22.08% | 6.36% |
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,
@@ -37,12 +39,14 @@
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": 0.0,
47
  "gradient_checkpointing": true,
48
  "hidden_act": "gelu",
@@ -66,11 +70,15 @@
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": 0,
74
- "transformers_version": "4.5.0.dev0",
75
- "vocab_size": 75
 
76
  }
 
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",
 
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": 44
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 sha256:2d9909cce880814662ccf4b45b2b3dcaf90801c7343acf2884e100ca50555b83
3
- size 1262241303
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:14040a54d1e0caecd8b2271fdd2d96653be45702c7bc8765354140d864fb445a
3
+ size 1262114199
vocab.json CHANGED
@@ -1 +1 @@
1
- {"<pad>": 0, "<s>": 1, "</s>": 2, "<unk>": 3, "|": 4, "Ð": 5, "Ó": 6, "Ł": 7, "Ô": 8, "O": 9, "Å": 10, "S": 11, "L": 12, "Ě": 13, "Ř": 14, "Ū": 15, "Ş": 16, "Ê": 17, "J": 18, "Q": 19, "Ō": 20, "U": 21, "Ž": 22, "A": 23, "Ć": 24, "B": 25, "Ï": 26, "Č": 27, "Ì": 28, "D": 29, "Y": 30, "Ğ": 31, "$": 32, "C": 33, "Á": 34, "P": 35, "Î": 36, "Ń": 37, "N": 38, "X": 39, "À": 40, "Ë": 41, "'": 42, "Ñ": 43, "Ī": 44, "È": 45, "Š": 46, "F": 47, "Ø": 48, "E": 49, "Ö": 50, "Ź": 51, "K": 52, "G": 53, "Ä": 54, "Ã": 55, "Μ": 56, "T": 57, "Ę": 58, "V": 59, "I": 60, "É": 61, "Z": 62, "R": 63, "-": 64, "W": 65, "Æ": 66, "M": 67, "Í": 68, "Œ": 69, "Ù": 70, "H": 71, "Ḥ": 72, "Ò": 73, "Ú": 74}
 
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, "À": 33, "Á": 34, "È": 35, "É": 36, "Ì": 37, "Í": 38, "Ò": 39, "Ó": 40, "Ù": 41, "Ú": 42, "Š": 43}