Baybars commited on
Commit
0f68f26
1 Parent(s): 56c12f3

Model save

Browse files
config.json CHANGED
@@ -1,5 +1,5 @@
1
  {
2
- "_name_or_path": "hf-test/xls-r-dummy",
3
  "activation_dropout": 0.0,
4
  "adapter_kernel_size": 3,
5
  "adapter_stride": 2,
@@ -11,23 +11,35 @@
11
  "attention_dropout": 0.0,
12
  "bos_token_id": 1,
13
  "classifier_proj_size": 256,
14
- "codevector_dim": 256,
15
  "contrastive_logits_temperature": 0.1,
16
- "conv_bias": false,
17
  "conv_dim": [
18
- 32,
19
- 32,
20
- 32
 
 
 
 
21
  ],
22
  "conv_kernel": [
23
- 8,
24
- 8,
25
- 8
 
 
 
 
26
  ],
27
  "conv_stride": [
28
- 4,
29
- 4,
30
- 4
 
 
 
 
31
  ],
32
  "ctc_loss_reduction": "mean",
33
  "ctc_zero_infinity": false,
@@ -42,31 +54,30 @@
42
  "final_dropout": 0.0,
43
  "hidden_act": "gelu",
44
  "hidden_dropout": 0.0,
45
- "hidden_dropout_prob": 0.1,
46
- "hidden_size": 16,
47
  "initializer_range": 0.02,
48
- "intermediate_size": 20,
49
  "layer_norm_eps": 1e-05,
50
  "layerdrop": 0.0,
51
- "mask_feature_length": 10,
52
  "mask_feature_min_masks": 0,
53
- "mask_feature_prob": 0.0,
54
  "mask_time_length": 10,
55
  "mask_time_min_masks": 2,
56
- "mask_time_prob": 0.05,
57
  "model_type": "wav2vec2",
58
  "num_adapter_layers": 3,
59
- "num_attention_heads": 2,
60
  "num_codevector_groups": 2,
61
  "num_codevectors_per_group": 320,
62
- "num_conv_pos_embedding_groups": 2,
63
- "num_conv_pos_embeddings": 16,
64
- "num_feat_extract_layers": 3,
65
- "num_hidden_layers": 4,
66
- "num_negatives": 10,
67
- "output_hidden_size": 16,
68
  "pad_token_id": 38,
69
- "proj_codevector_dim": 256,
70
  "tdnn_dilation": [
71
  1,
72
  2,
 
1
  {
2
+ "_name_or_path": "facebook/wav2vec2-xls-r-300m",
3
  "activation_dropout": 0.0,
4
  "adapter_kernel_size": 3,
5
  "adapter_stride": 2,
 
11
  "attention_dropout": 0.0,
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": false,
 
54
  "final_dropout": 0.0,
55
  "hidden_act": "gelu",
56
  "hidden_dropout": 0.0,
57
+ "hidden_size": 1024,
 
58
  "initializer_range": 0.02,
59
+ "intermediate_size": 4096,
60
  "layer_norm_eps": 1e-05,
61
  "layerdrop": 0.0,
62
+ "mask_feature_length": 64,
63
  "mask_feature_min_masks": 0,
64
+ "mask_feature_prob": 0.1,
65
  "mask_time_length": 10,
66
  "mask_time_min_masks": 2,
67
+ "mask_time_prob": 0.3,
68
  "model_type": "wav2vec2",
69
  "num_adapter_layers": 3,
70
+ "num_attention_heads": 16,
71
  "num_codevector_groups": 2,
72
  "num_codevectors_per_group": 320,
73
+ "num_conv_pos_embedding_groups": 16,
74
+ "num_conv_pos_embeddings": 128,
75
+ "num_feat_extract_layers": 7,
76
+ "num_hidden_layers": 24,
77
+ "num_negatives": 100,
78
+ "output_hidden_size": 1024,
79
  "pad_token_id": 38,
80
+ "proj_codevector_dim": 768,
81
  "tdnn_dilation": [
82
  1,
83
  2,
preprocessor_config.json CHANGED
@@ -3,7 +3,7 @@
3
  "feature_extractor_type": "Wav2Vec2FeatureExtractor",
4
  "feature_size": 1,
5
  "padding_side": "right",
6
- "padding_value": 0.0,
7
- "return_attention_mask": false,
8
  "sampling_rate": 16000
9
  }
 
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 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:1f1900fda595fc11d224093d778614c0532b93a9ba52d38f0313c2cb347dfff4
3
- size 143078
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bc548f439611bd98eab1f2cbb578ef3fd4344acf3ed24a78ee45778dbdd9db60
3
+ size 1262091761
run-test.shs ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ python run_speech_recognition_ctc.py \
2
+ --dataset_name="common_voice" \
3
+ --model_name_or_path="hf-test/xls-r-dummy" \
4
+ --dataset_config_name="tr" \
5
+ --output_dir="./" \
6
+ --overwrite_output_dir \
7
+ --num_train_epochs="5" \
8
+ --per_device_train_batch_size="16" \
9
+ --gradient_accumulation_steps="2" \
10
+ --learning_rate="3e-4" \
11
+ --warmup_steps="500" \
12
+ --evaluation_strategy="steps" \
13
+ --text_column_name="sentence" \
14
+ --length_column_name="input_length" \
15
+ --save_steps="400" \
16
+ --eval_steps="100" \
17
+ --layerdrop="0.0" \
18
+ --save_total_limit="3" \
19
+ --freeze_feature_encoder \
20
+ --gradient_checkpointing \
21
+ --chars_to_ignore , ? . ! - \; \: \" “ % ‘ ” � \
22
+ --fp16 \
23
+ --group_by_length \
24
+ --push_to_hub \
25
+ --do_train --do_eval
run.sh CHANGED
@@ -1,25 +1,29 @@
1
  python run_speech_recognition_ctc.py \
2
  --dataset_name="common_voice" \
3
- --model_name_or_path="hf-test/xls-r-dummy" \
4
  --dataset_config_name="tr" \
5
  --output_dir="./" \
6
  --overwrite_output_dir \
7
- --num_train_epochs="5" \
8
- --per_device_train_batch_size="16" \
9
- --gradient_accumulation_steps="2" \
10
- --learning_rate="3e-4" \
11
  --warmup_steps="500" \
12
  --evaluation_strategy="steps" \
13
  --text_column_name="sentence" \
14
- --length_column_name="input_length" \
15
- --save_steps="400" \
16
- --eval_steps="100" \
17
  --layerdrop="0.0" \
18
- --save_total_limit="3" \
19
- --freeze_feature_encoder \
20
- --gradient_checkpointing \
 
 
 
21
  --chars_to_ignore , ? . ! - \; \: \" “ % ‘ ” � \
22
  --fp16 \
23
  --group_by_length \
24
  --push_to_hub \
25
- --do_train --do_eval
 
 
1
  python run_speech_recognition_ctc.py \
2
  --dataset_name="common_voice" \
3
+ --model_name_or_path="facebook/wav2vec2-xls-r-300m" \
4
  --dataset_config_name="tr" \
5
  --output_dir="./" \
6
  --overwrite_output_dir \
7
+ --num_train_epochs="2" \
8
+ --per_device_train_batch_size="32" \
9
+ --gradient_accumulation_steps="1" \
10
+ --learning_rate="5e-4" \
11
  --warmup_steps="500" \
12
  --evaluation_strategy="steps" \
13
  --text_column_name="sentence" \
14
+ --save_steps="500" \
15
+ --eval_steps="500" \
16
+ --logging_steps="1" \
17
  --layerdrop="0.0" \
18
+ --eval_metrics wer cer \
19
+ --save_total_limit="1" \
20
+ --mask_time_prob="0.3" \
21
+ --mask_time_length="10" \
22
+ --mask_feature_prob="0.1" \
23
+ --mask_feature_length="64" \
24
  --chars_to_ignore , ? . ! - \; \: \" “ % ‘ ” � \
25
  --fp16 \
26
  --group_by_length \
27
  --push_to_hub \
28
+ --do_train --do_eval \
29
+ --gradient_checkpointing
special_tokens_map.json CHANGED
@@ -1 +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}]}
 
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}]}
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:fb45223fb5a36e43839ad3901144a0cc22ece7adfab348b54d71e1678afd9fca
3
  size 2991
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c741e2d2d858980a8cfdf104cef10aacd51be9a34f3f0fb299372de4f16f9e54
3
  size 2991