Masaki Hattori
commited on
Commit
•
6a9faba
1
Parent(s):
8e3fa5f
jdrt_byclass_rinnna_hubert_asr_2
Browse files- README.md +111 -0
- config.json +72 -0
- preprocessor_config.json +9 -0
- pytorch_model.bin +3 -0
- training_args.bin +3 -0
README.md
ADDED
@@ -0,0 +1,111 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
base_model: rinna/japanese-hubert-base
|
4 |
+
tags:
|
5 |
+
- generated_from_trainer
|
6 |
+
metrics:
|
7 |
+
- wer
|
8 |
+
model-index:
|
9 |
+
- name: jdrt_byclass_rinnna_hubert_asr_2
|
10 |
+
results: []
|
11 |
+
---
|
12 |
+
|
13 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
14 |
+
should probably proofread and complete it, then remove this comment. -->
|
15 |
+
|
16 |
+
# jdrt_byclass_rinnna_hubert_asr_2
|
17 |
+
|
18 |
+
This model is a fine-tuned version of [rinna/japanese-hubert-base](https://huggingface.co/rinna/japanese-hubert-base) on the None dataset.
|
19 |
+
It achieves the following results on the evaluation set:
|
20 |
+
- Loss: 0.4574
|
21 |
+
- Wer: 0.4461
|
22 |
+
- Cer: 0.3253
|
23 |
+
|
24 |
+
## Model description
|
25 |
+
|
26 |
+
More information needed
|
27 |
+
|
28 |
+
## Intended uses & limitations
|
29 |
+
|
30 |
+
More information needed
|
31 |
+
|
32 |
+
## Training and evaluation data
|
33 |
+
|
34 |
+
More information needed
|
35 |
+
|
36 |
+
## Training procedure
|
37 |
+
|
38 |
+
### Training hyperparameters
|
39 |
+
|
40 |
+
The following hyperparameters were used during training:
|
41 |
+
- learning_rate: 5e-05
|
42 |
+
- train_batch_size: 256
|
43 |
+
- eval_batch_size: 256
|
44 |
+
- seed: 42
|
45 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
46 |
+
- lr_scheduler_type: cosine
|
47 |
+
- lr_scheduler_warmup_steps: 260
|
48 |
+
- num_epochs: 50
|
49 |
+
|
50 |
+
### Training results
|
51 |
+
|
52 |
+
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|
53 |
+
|:-------------:|:-----:|:----:|:---------------:|:------:|:------:|
|
54 |
+
| 11.8822 | 1.0 | 53 | 8.0828 | 0.9838 | 0.9911 |
|
55 |
+
| 6.3754 | 2.0 | 106 | 4.6774 | 0.9838 | 0.9911 |
|
56 |
+
| 4.5264 | 3.0 | 159 | 3.9963 | 0.9838 | 0.9911 |
|
57 |
+
| 3.7867 | 4.0 | 212 | 3.3575 | 0.9838 | 0.9911 |
|
58 |
+
| 3.2022 | 5.0 | 265 | 2.9007 | 0.9838 | 0.9911 |
|
59 |
+
| 2.8134 | 6.0 | 318 | 2.7079 | 0.9838 | 0.9911 |
|
60 |
+
| 2.708 | 7.0 | 371 | 2.6779 | 0.9838 | 0.9911 |
|
61 |
+
| 2.6923 | 8.0 | 424 | 2.6719 | 0.9838 | 0.9911 |
|
62 |
+
| 2.6555 | 9.0 | 477 | 2.4997 | 0.9838 | 0.9911 |
|
63 |
+
| 2.3308 | 10.0 | 530 | 2.1709 | 0.9960 | 0.9522 |
|
64 |
+
| 2.1951 | 11.0 | 583 | 2.0579 | 0.9960 | 0.9522 |
|
65 |
+
| 2.0032 | 12.0 | 636 | 1.8255 | 0.9989 | 0.9302 |
|
66 |
+
| 1.8394 | 13.0 | 689 | 1.6870 | 0.9999 | 0.9021 |
|
67 |
+
| 1.6562 | 14.0 | 742 | 1.4793 | 0.9999 | 0.8139 |
|
68 |
+
| 1.5057 | 15.0 | 795 | 1.2942 | 0.9997 | 0.7813 |
|
69 |
+
| 1.4022 | 16.0 | 848 | 1.2146 | 0.9997 | 0.7594 |
|
70 |
+
| 1.2945 | 17.0 | 901 | 1.1512 | 0.9941 | 0.7395 |
|
71 |
+
| 1.2062 | 18.0 | 954 | 1.0732 | 0.9844 | 0.6726 |
|
72 |
+
| 1.1396 | 19.0 | 1007 | 1.0457 | 0.9694 | 0.6559 |
|
73 |
+
| 1.0419 | 20.0 | 1060 | 0.8567 | 0.5929 | 0.4214 |
|
74 |
+
| 0.8799 | 21.0 | 1113 | 0.7172 | 0.5273 | 0.3951 |
|
75 |
+
| 0.8249 | 22.0 | 1166 | 0.6980 | 0.5204 | 0.4048 |
|
76 |
+
| 0.7649 | 23.0 | 1219 | 0.6157 | 0.4923 | 0.3737 |
|
77 |
+
| 0.7401 | 24.0 | 1272 | 0.6171 | 0.4992 | 0.3722 |
|
78 |
+
| 0.7074 | 25.0 | 1325 | 0.5948 | 0.4835 | 0.3669 |
|
79 |
+
| 0.6543 | 26.0 | 1378 | 0.5520 | 0.4759 | 0.3537 |
|
80 |
+
| 0.6164 | 27.0 | 1431 | 0.5512 | 0.4745 | 0.3467 |
|
81 |
+
| 0.5758 | 28.0 | 1484 | 0.5286 | 0.4677 | 0.3416 |
|
82 |
+
| 0.5648 | 29.0 | 1537 | 0.5099 | 0.4644 | 0.3447 |
|
83 |
+
| 0.578 | 30.0 | 1590 | 0.5262 | 0.4701 | 0.3530 |
|
84 |
+
| 0.5343 | 31.0 | 1643 | 0.5025 | 0.4621 | 0.3463 |
|
85 |
+
| 0.5236 | 32.0 | 1696 | 0.4927 | 0.4588 | 0.3411 |
|
86 |
+
| 0.5256 | 33.0 | 1749 | 0.4971 | 0.4614 | 0.3434 |
|
87 |
+
| 0.5309 | 34.0 | 1802 | 0.4882 | 0.4553 | 0.3400 |
|
88 |
+
| 0.4938 | 35.0 | 1855 | 0.4663 | 0.4497 | 0.3309 |
|
89 |
+
| 0.4874 | 36.0 | 1908 | 0.4734 | 0.4508 | 0.3316 |
|
90 |
+
| 0.4519 | 37.0 | 1961 | 0.4642 | 0.4483 | 0.3289 |
|
91 |
+
| 0.5137 | 38.0 | 2014 | 0.4655 | 0.4490 | 0.3303 |
|
92 |
+
| 0.4832 | 39.0 | 2067 | 0.4613 | 0.4474 | 0.3282 |
|
93 |
+
| 0.4581 | 40.0 | 2120 | 0.4565 | 0.4450 | 0.3236 |
|
94 |
+
| 0.4542 | 41.0 | 2173 | 0.4550 | 0.4461 | 0.3262 |
|
95 |
+
| 0.4478 | 42.0 | 2226 | 0.4548 | 0.4442 | 0.3247 |
|
96 |
+
| 0.448 | 43.0 | 2279 | 0.4523 | 0.4450 | 0.3238 |
|
97 |
+
| 0.4468 | 44.0 | 2332 | 0.4560 | 0.4455 | 0.3224 |
|
98 |
+
| 0.4681 | 45.0 | 2385 | 0.4599 | 0.4468 | 0.3253 |
|
99 |
+
| 0.455 | 46.0 | 2438 | 0.4589 | 0.4472 | 0.3246 |
|
100 |
+
| 0.4479 | 47.0 | 2491 | 0.4583 | 0.4465 | 0.3245 |
|
101 |
+
| 0.4533 | 48.0 | 2544 | 0.4582 | 0.4469 | 0.3253 |
|
102 |
+
| 0.4486 | 49.0 | 2597 | 0.4575 | 0.4462 | 0.3253 |
|
103 |
+
| 0.4466 | 50.0 | 2650 | 0.4574 | 0.4461 | 0.3253 |
|
104 |
+
|
105 |
+
|
106 |
+
### Framework versions
|
107 |
+
|
108 |
+
- Transformers 4.34.0.dev0
|
109 |
+
- Pytorch 2.0.1+cu118
|
110 |
+
- Datasets 2.14.5
|
111 |
+
- Tokenizers 0.13.3
|
config.json
ADDED
@@ -0,0 +1,72 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "rinna/japanese-hubert-base",
|
3 |
+
"activation_dropout": 0.1,
|
4 |
+
"apply_spec_augment": true,
|
5 |
+
"architectures": [
|
6 |
+
"HubertForCTC"
|
7 |
+
],
|
8 |
+
"attention_dropout": 0.1,
|
9 |
+
"bos_token_id": 1,
|
10 |
+
"classifier_proj_size": 256,
|
11 |
+
"conv_bias": false,
|
12 |
+
"conv_dim": [
|
13 |
+
512,
|
14 |
+
512,
|
15 |
+
512,
|
16 |
+
512,
|
17 |
+
512,
|
18 |
+
512,
|
19 |
+
512
|
20 |
+
],
|
21 |
+
"conv_kernel": [
|
22 |
+
10,
|
23 |
+
3,
|
24 |
+
3,
|
25 |
+
3,
|
26 |
+
3,
|
27 |
+
2,
|
28 |
+
2
|
29 |
+
],
|
30 |
+
"conv_stride": [
|
31 |
+
5,
|
32 |
+
2,
|
33 |
+
2,
|
34 |
+
2,
|
35 |
+
2,
|
36 |
+
2,
|
37 |
+
2
|
38 |
+
],
|
39 |
+
"ctc_loss_reduction": "mean",
|
40 |
+
"ctc_zero_infinity": false,
|
41 |
+
"do_stable_layer_norm": false,
|
42 |
+
"eos_token_id": 2,
|
43 |
+
"feat_extract_activation": "gelu",
|
44 |
+
"feat_extract_norm": "group",
|
45 |
+
"feat_proj_dropout": 0.1,
|
46 |
+
"feat_proj_layer_norm": true,
|
47 |
+
"final_dropout": 0.1,
|
48 |
+
"hidden_act": "gelu",
|
49 |
+
"hidden_dropout": 0,
|
50 |
+
"hidden_size": 768,
|
51 |
+
"initializer_range": 0.02,
|
52 |
+
"intermediate_size": 3072,
|
53 |
+
"layer_norm_eps": 1e-05,
|
54 |
+
"layerdrop": 0.1,
|
55 |
+
"mask_feature_length": 10,
|
56 |
+
"mask_feature_min_masks": 0,
|
57 |
+
"mask_feature_prob": 0.0,
|
58 |
+
"mask_time_length": 10,
|
59 |
+
"mask_time_min_masks": 2,
|
60 |
+
"mask_time_prob": 0.05,
|
61 |
+
"model_type": "hubert",
|
62 |
+
"num_attention_heads": 12,
|
63 |
+
"num_conv_pos_embedding_groups": 16,
|
64 |
+
"num_conv_pos_embeddings": 128,
|
65 |
+
"num_feat_extract_layers": 7,
|
66 |
+
"num_hidden_layers": 12,
|
67 |
+
"pad_token_id": 28,
|
68 |
+
"torch_dtype": "float32",
|
69 |
+
"transformers_version": "4.34.0.dev0",
|
70 |
+
"use_weighted_layer_sum": false,
|
71 |
+
"vocab_size": 29
|
72 |
+
}
|
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": 0.0,
|
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:ce138f33909c9f4784589f06de30fa10b0ee265d231bdbb1bbe65ad0a619a5bd
|
3 |
+
size 377649057
|
training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c4712b520abc1536421c411c214980d15f7ee55b49d90fa0a792055a41ae2904
|
3 |
+
size 4027
|