updating the repo with the fine-tuned model
Browse files- .gitattributes +1 -0
- README.md +82 -0
- added_tokens.json +4 -0
- all_results.json +14 -0
- alphabet.json +1 -0
- config.json +107 -0
- eval_results.json +9 -0
- language_model/atcosim_uwb_atcc_4g.binary +3 -0
- language_model/attrs.json +1 -0
- language_model/unigrams.txt +0 -0
- log/eval_model_train +15 -0
- log/train_log +0 -0
- preprocessor_config.json +10 -0
- pytorch_model.bin +3 -0
- special_tokens_map.json +6 -0
- tokenizer_config.json +11 -0
- train_results.json +8 -0
- trainer_state.json +265 -0
- training_args.bin +3 -0
- vocab.json +31 -0
.gitattributes
CHANGED
@@ -32,3 +32,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
32 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
33 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
34 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
32 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
33 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
34 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
35 |
+
language_model/atcosim_uwb_atcc_4g.binary filter=lfs diff=lfs merge=lfs -text
|
README.md
ADDED
@@ -0,0 +1,82 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
tags:
|
4 |
+
- automatic-speech-recognition
|
5 |
+
- experiments/data/atcosim_uwb_atcc/train
|
6 |
+
- generated_from_trainer
|
7 |
+
metrics:
|
8 |
+
- wer
|
9 |
+
model-index:
|
10 |
+
- name: 0.0ld_0.0ad_0.0attd_0.05fpd_0.075mtp_12mtl_0.0mfp_12mfl_1acc
|
11 |
+
results: []
|
12 |
+
---
|
13 |
+
|
14 |
+
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
15 |
+
should probably proofread and complete it, then remove this comment. -->
|
16 |
+
|
17 |
+
# 0.0ld_0.0ad_0.0attd_0.05fpd_0.075mtp_12mtl_0.0mfp_12mfl_1acc
|
18 |
+
|
19 |
+
This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the EXPERIMENTS/DATA/ATCOSIM_UWB_ATCC/TRAIN - NA dataset.
|
20 |
+
It achieves the following results on the evaluation set:
|
21 |
+
- Loss: 0.5595
|
22 |
+
- Wer: 0.1687
|
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: 0.0001
|
42 |
+
- train_batch_size: 24
|
43 |
+
- eval_batch_size: 12
|
44 |
+
- seed: 42
|
45 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
46 |
+
- lr_scheduler_type: linear
|
47 |
+
- lr_scheduler_warmup_steps: 1000
|
48 |
+
- training_steps: 10000
|
49 |
+
- mixed_precision_training: Native AMP
|
50 |
+
|
51 |
+
### Training results
|
52 |
+
|
53 |
+
| Training Loss | Epoch | Step | Validation Loss | Wer |
|
54 |
+
|:-------------:|:-----:|:-----:|:---------------:|:------:|
|
55 |
+
| No log | 0.63 | 500 | 3.0458 | 1.0 |
|
56 |
+
| 2.9181 | 1.27 | 1000 | 1.1503 | 0.4723 |
|
57 |
+
| 2.9181 | 1.9 | 1500 | 0.8275 | 0.3500 |
|
58 |
+
| 0.7646 | 2.53 | 2000 | 0.6990 | 0.2845 |
|
59 |
+
| 0.7646 | 3.17 | 2500 | 0.5828 | 0.2509 |
|
60 |
+
| 0.5394 | 3.8 | 3000 | 0.5363 | 0.2487 |
|
61 |
+
| 0.5394 | 4.44 | 3500 | 0.5467 | 0.2171 |
|
62 |
+
| 0.4558 | 5.07 | 4000 | 0.5290 | 0.2090 |
|
63 |
+
| 0.4558 | 5.7 | 4500 | 0.4992 | 0.2046 |
|
64 |
+
| 0.3773 | 6.34 | 5000 | 0.4934 | 0.2052 |
|
65 |
+
| 0.3773 | 6.97 | 5500 | 0.4700 | 0.1983 |
|
66 |
+
| 0.3301 | 7.6 | 6000 | 0.4938 | 0.1874 |
|
67 |
+
| 0.3301 | 8.24 | 6500 | 0.5364 | 0.1893 |
|
68 |
+
| 0.2938 | 8.87 | 7000 | 0.5170 | 0.1830 |
|
69 |
+
| 0.2938 | 9.51 | 7500 | 0.5408 | 0.1815 |
|
70 |
+
| 0.2674 | 10.14 | 8000 | 0.5581 | 0.1733 |
|
71 |
+
| 0.2674 | 10.77 | 8500 | 0.5389 | 0.1719 |
|
72 |
+
| 0.24 | 11.41 | 9000 | 0.5344 | 0.1714 |
|
73 |
+
| 0.24 | 12.04 | 9500 | 0.5503 | 0.1686 |
|
74 |
+
| 0.211 | 12.67 | 10000 | 0.5595 | 0.1687 |
|
75 |
+
|
76 |
+
|
77 |
+
### Framework versions
|
78 |
+
|
79 |
+
- Transformers 4.24.0
|
80 |
+
- Pytorch 1.13.0+cu117
|
81 |
+
- Datasets 2.6.1
|
82 |
+
- Tokenizers 0.13.2
|
added_tokens.json
ADDED
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"</s>": 30,
|
3 |
+
"<s>": 29
|
4 |
+
}
|
all_results.json
ADDED
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 12.67,
|
3 |
+
"eval_loss": 0.5595183372497559,
|
4 |
+
"eval_runtime": 161.956,
|
5 |
+
"eval_samples": 4723,
|
6 |
+
"eval_samples_per_second": 29.162,
|
7 |
+
"eval_steps_per_second": 2.433,
|
8 |
+
"eval_wer": 0.16865713938706936,
|
9 |
+
"train_loss": 0.6397389862060547,
|
10 |
+
"train_runtime": 13054.0758,
|
11 |
+
"train_samples": 18925,
|
12 |
+
"train_samples_per_second": 18.385,
|
13 |
+
"train_steps_per_second": 0.766
|
14 |
+
}
|
alphabet.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"labels": [" ", "a", "b", "c", "d", "e", "f", "g", "h", "i", "j", "k", "l", "m", "n", "o", "p", "q", "r", "s", "t", "u", "v", "w", "x", "y", "z", "\u2047", "", "<s>", "</s>"], "is_bpe": false}
|
config.json
ADDED
@@ -0,0 +1,107 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"_name_or_path": "facebook/wav2vec2-xls-r-300m",
|
3 |
+
"activation_dropout": 0.0,
|
4 |
+
"adapter_kernel_size": 3,
|
5 |
+
"adapter_stride": 2,
|
6 |
+
"add_adapter": false,
|
7 |
+
"apply_spec_augment": true,
|
8 |
+
"architectures": [
|
9 |
+
"Wav2Vec2ForCTC"
|
10 |
+
],
|
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": true,
|
46 |
+
"diversity_loss_weight": 0.1,
|
47 |
+
"do_stable_layer_norm": true,
|
48 |
+
"eos_token_id": 2,
|
49 |
+
"feat_extract_activation": "gelu",
|
50 |
+
"feat_extract_dropout": 0.0,
|
51 |
+
"feat_extract_norm": "layer",
|
52 |
+
"feat_proj_dropout": 0.05,
|
53 |
+
"feat_quantizer_dropout": 0.0,
|
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": 12,
|
63 |
+
"mask_feature_min_masks": 0,
|
64 |
+
"mask_feature_prob": 0.0,
|
65 |
+
"mask_time_length": 12,
|
66 |
+
"mask_time_min_masks": 2,
|
67 |
+
"mask_time_prob": 0.075,
|
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": 28,
|
80 |
+
"proj_codevector_dim": 768,
|
81 |
+
"tdnn_dilation": [
|
82 |
+
1,
|
83 |
+
2,
|
84 |
+
3,
|
85 |
+
1,
|
86 |
+
1
|
87 |
+
],
|
88 |
+
"tdnn_dim": [
|
89 |
+
512,
|
90 |
+
512,
|
91 |
+
512,
|
92 |
+
512,
|
93 |
+
1500
|
94 |
+
],
|
95 |
+
"tdnn_kernel": [
|
96 |
+
5,
|
97 |
+
3,
|
98 |
+
3,
|
99 |
+
1,
|
100 |
+
1
|
101 |
+
],
|
102 |
+
"torch_dtype": "float32",
|
103 |
+
"transformers_version": "4.24.0",
|
104 |
+
"use_weighted_layer_sum": false,
|
105 |
+
"vocab_size": 31,
|
106 |
+
"xvector_output_dim": 512
|
107 |
+
}
|
eval_results.json
ADDED
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 12.67,
|
3 |
+
"eval_loss": 0.5595183372497559,
|
4 |
+
"eval_runtime": 161.956,
|
5 |
+
"eval_samples": 4723,
|
6 |
+
"eval_samples_per_second": 29.162,
|
7 |
+
"eval_steps_per_second": 2.433,
|
8 |
+
"eval_wer": 0.16865713938706936
|
9 |
+
}
|
language_model/atcosim_uwb_atcc_4g.binary
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:974c8f8db889c30ac4060a741144847c7b71cad0a8bd6714d81a265794b235d0
|
3 |
+
size 1360621
|
language_model/attrs.json
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
{"alpha": 0.5, "beta": 1.5, "unk_score_offset": -10.0, "score_boundary": true}
|
language_model/unigrams.txt
ADDED
File without changes
|
log/eval_model_train
ADDED
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Running on vgnf004
|
2 |
+
# Started at Wed 30 Nov 07:37:30 CET 2022
|
3 |
+
# python3 src/eval_model.py --language-model experiments/data/atcosim_uwb_atcc/train/lm/atcosim_uwb_atcc_4g.binary --pretrained-model experiments/results/bs_exp/linear/wav2vec2-xls-r-300m/atcosim_uwb_atcc/0.0ld_0.0ad_0.0attd_0.05fpd_0.075mtp_12mtl_0.0mfp_12mfl_1acc/ --print-output true --test-set experiments/data/atcosim_corpus/train
|
4 |
+
Integrating a LM by shallow fusion, results should be better
|
5 |
+
*** Loading the Wav2Vec 2.0 model, loading... ***
|
6 |
+
Traceback (most recent call last):
|
7 |
+
File "/remote/idiap.svm/temp.speech01/jzuluaga/experiments/journal/asr/github/w2v2-air-traffic/src/eval_model.py", line 250, in <module>
|
8 |
+
main()
|
9 |
+
File "/remote/idiap.svm/temp.speech01/jzuluaga/experiments/journal/asr/github/w2v2-air-traffic/src/eval_model.py", line 152, in main
|
10 |
+
processor, processor_ctc_kenlm, model = get_kenlm_processor(path_model, path_lm)
|
11 |
+
File "/remote/idiap.svm/temp.speech01/jzuluaga/experiments/journal/asr/github/w2v2-air-traffic/src/eval_model.py", line 65, in get_kenlm_processor
|
12 |
+
path_tokenizer + "/vocab.json",
|
13 |
+
TypeError: unsupported operand type(s) for +: 'PosixPath' and 'str'
|
14 |
+
# Accounting: time=18 threads=1
|
15 |
+
# Finished at Wed 30 Nov 07:37:48 CET 2022 with status 1
|
log/train_log
ADDED
The diff for this file is too large to render.
See raw diff
|
|
preprocessor_config.json
ADDED
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"do_normalize": true,
|
3 |
+
"feature_extractor_type": "Wav2Vec2FeatureExtractor",
|
4 |
+
"feature_size": 1,
|
5 |
+
"padding_side": "right",
|
6 |
+
"padding_value": 0,
|
7 |
+
"processor_class": "Wav2Vec2ProcessorWithLM",
|
8 |
+
"return_attention_mask": true,
|
9 |
+
"sampling_rate": 16000
|
10 |
+
}
|
pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:955e67ecbe9217132afe903b7188c9207bbfd552090feecd9fa0f1076b3dc3d8
|
3 |
+
size 1262028973
|
special_tokens_map.json
ADDED
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token": "<s>",
|
3 |
+
"eos_token": "</s>",
|
4 |
+
"pad_token": "[PAD]",
|
5 |
+
"unk_token": "[UNK]"
|
6 |
+
}
|
tokenizer_config.json
ADDED
@@ -0,0 +1,11 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"bos_token": "<s>",
|
3 |
+
"do_lower_case": false,
|
4 |
+
"eos_token": "</s>",
|
5 |
+
"pad_token": "[PAD]",
|
6 |
+
"processor_class": "Wav2Vec2ProcessorWithLM",
|
7 |
+
"replace_word_delimiter_char": " ",
|
8 |
+
"tokenizer_class": "Wav2Vec2CTCTokenizer",
|
9 |
+
"unk_token": "[UNK]",
|
10 |
+
"word_delimiter_token": "|"
|
11 |
+
}
|
train_results.json
ADDED
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"epoch": 12.67,
|
3 |
+
"train_loss": 0.6397389862060547,
|
4 |
+
"train_runtime": 13054.0758,
|
5 |
+
"train_samples": 18925,
|
6 |
+
"train_samples_per_second": 18.385,
|
7 |
+
"train_steps_per_second": 0.766
|
8 |
+
}
|
trainer_state.json
ADDED
@@ -0,0 +1,265 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": null,
|
3 |
+
"best_model_checkpoint": null,
|
4 |
+
"epoch": 12.67427122940431,
|
5 |
+
"global_step": 10000,
|
6 |
+
"is_hyper_param_search": false,
|
7 |
+
"is_local_process_zero": true,
|
8 |
+
"is_world_process_zero": true,
|
9 |
+
"log_history": [
|
10 |
+
{
|
11 |
+
"epoch": 0.63,
|
12 |
+
"eval_loss": 3.0458309650421143,
|
13 |
+
"eval_runtime": 143.841,
|
14 |
+
"eval_samples_per_second": 32.835,
|
15 |
+
"eval_steps_per_second": 2.739,
|
16 |
+
"eval_wer": 1.0,
|
17 |
+
"step": 500
|
18 |
+
},
|
19 |
+
{
|
20 |
+
"epoch": 1.27,
|
21 |
+
"learning_rate": 9.98e-05,
|
22 |
+
"loss": 2.9181,
|
23 |
+
"step": 1000
|
24 |
+
},
|
25 |
+
{
|
26 |
+
"epoch": 1.27,
|
27 |
+
"eval_loss": 1.1502786874771118,
|
28 |
+
"eval_runtime": 143.5317,
|
29 |
+
"eval_samples_per_second": 32.906,
|
30 |
+
"eval_steps_per_second": 2.745,
|
31 |
+
"eval_wer": 0.4722885713129023,
|
32 |
+
"step": 1000
|
33 |
+
},
|
34 |
+
{
|
35 |
+
"epoch": 1.9,
|
36 |
+
"eval_loss": 0.8274601101875305,
|
37 |
+
"eval_runtime": 142.4484,
|
38 |
+
"eval_samples_per_second": 33.156,
|
39 |
+
"eval_steps_per_second": 2.766,
|
40 |
+
"eval_wer": 0.3499858305331768,
|
41 |
+
"step": 1500
|
42 |
+
},
|
43 |
+
{
|
44 |
+
"epoch": 2.53,
|
45 |
+
"learning_rate": 8.89111111111111e-05,
|
46 |
+
"loss": 0.7646,
|
47 |
+
"step": 2000
|
48 |
+
},
|
49 |
+
{
|
50 |
+
"epoch": 2.53,
|
51 |
+
"eval_loss": 0.6990081667900085,
|
52 |
+
"eval_runtime": 145.8955,
|
53 |
+
"eval_samples_per_second": 32.372,
|
54 |
+
"eval_steps_per_second": 2.701,
|
55 |
+
"eval_wer": 0.2845228938099672,
|
56 |
+
"step": 2000
|
57 |
+
},
|
58 |
+
{
|
59 |
+
"epoch": 3.17,
|
60 |
+
"eval_loss": 0.5828294157981873,
|
61 |
+
"eval_runtime": 147.7617,
|
62 |
+
"eval_samples_per_second": 31.964,
|
63 |
+
"eval_steps_per_second": 2.666,
|
64 |
+
"eval_wer": 0.2508602890571232,
|
65 |
+
"step": 2500
|
66 |
+
},
|
67 |
+
{
|
68 |
+
"epoch": 3.8,
|
69 |
+
"learning_rate": 7.780000000000001e-05,
|
70 |
+
"loss": 0.5394,
|
71 |
+
"step": 3000
|
72 |
+
},
|
73 |
+
{
|
74 |
+
"epoch": 3.8,
|
75 |
+
"eval_loss": 0.5363064408302307,
|
76 |
+
"eval_runtime": 147.8939,
|
77 |
+
"eval_samples_per_second": 31.935,
|
78 |
+
"eval_steps_per_second": 2.664,
|
79 |
+
"eval_wer": 0.24873486903364236,
|
80 |
+
"step": 3000
|
81 |
+
},
|
82 |
+
{
|
83 |
+
"epoch": 4.44,
|
84 |
+
"eval_loss": 0.5467007756233215,
|
85 |
+
"eval_runtime": 151.8112,
|
86 |
+
"eval_samples_per_second": 31.111,
|
87 |
+
"eval_steps_per_second": 2.595,
|
88 |
+
"eval_wer": 0.21711671592243229,
|
89 |
+
"step": 3500
|
90 |
+
},
|
91 |
+
{
|
92 |
+
"epoch": 5.07,
|
93 |
+
"learning_rate": 6.66888888888889e-05,
|
94 |
+
"loss": 0.4558,
|
95 |
+
"step": 4000
|
96 |
+
},
|
97 |
+
{
|
98 |
+
"epoch": 5.07,
|
99 |
+
"eval_loss": 0.5289521813392639,
|
100 |
+
"eval_runtime": 152.3052,
|
101 |
+
"eval_samples_per_second": 31.01,
|
102 |
+
"eval_steps_per_second": 2.587,
|
103 |
+
"eval_wer": 0.20901987773774341,
|
104 |
+
"step": 4000
|
105 |
+
},
|
106 |
+
{
|
107 |
+
"epoch": 5.7,
|
108 |
+
"eval_loss": 0.4992470443248749,
|
109 |
+
"eval_runtime": 148.0991,
|
110 |
+
"eval_samples_per_second": 31.891,
|
111 |
+
"eval_steps_per_second": 2.66,
|
112 |
+
"eval_wer": 0.20464758511801143,
|
113 |
+
"step": 4500
|
114 |
+
},
|
115 |
+
{
|
116 |
+
"epoch": 6.34,
|
117 |
+
"learning_rate": 5.558888888888889e-05,
|
118 |
+
"loss": 0.3773,
|
119 |
+
"step": 5000
|
120 |
+
},
|
121 |
+
{
|
122 |
+
"epoch": 6.34,
|
123 |
+
"eval_loss": 0.49340707063674927,
|
124 |
+
"eval_runtime": 155.8909,
|
125 |
+
"eval_samples_per_second": 30.297,
|
126 |
+
"eval_steps_per_second": 2.527,
|
127 |
+
"eval_wer": 0.2051738796000162,
|
128 |
+
"step": 5000
|
129 |
+
},
|
130 |
+
{
|
131 |
+
"epoch": 6.97,
|
132 |
+
"eval_loss": 0.47003018856048584,
|
133 |
+
"eval_runtime": 146.5312,
|
134 |
+
"eval_samples_per_second": 32.232,
|
135 |
+
"eval_steps_per_second": 2.689,
|
136 |
+
"eval_wer": 0.19827132504756892,
|
137 |
+
"step": 5500
|
138 |
+
},
|
139 |
+
{
|
140 |
+
"epoch": 7.6,
|
141 |
+
"learning_rate": 4.447777777777778e-05,
|
142 |
+
"loss": 0.3301,
|
143 |
+
"step": 6000
|
144 |
+
},
|
145 |
+
{
|
146 |
+
"epoch": 7.6,
|
147 |
+
"eval_loss": 0.4938233494758606,
|
148 |
+
"eval_runtime": 154.4839,
|
149 |
+
"eval_samples_per_second": 30.573,
|
150 |
+
"eval_steps_per_second": 2.55,
|
151 |
+
"eval_wer": 0.18738107768916237,
|
152 |
+
"step": 6000
|
153 |
+
},
|
154 |
+
{
|
155 |
+
"epoch": 8.24,
|
156 |
+
"eval_loss": 0.5363861322402954,
|
157 |
+
"eval_runtime": 163.224,
|
158 |
+
"eval_samples_per_second": 28.936,
|
159 |
+
"eval_steps_per_second": 2.414,
|
160 |
+
"eval_wer": 0.18934456094894944,
|
161 |
+
"step": 6500
|
162 |
+
},
|
163 |
+
{
|
164 |
+
"epoch": 8.87,
|
165 |
+
"learning_rate": 3.336666666666667e-05,
|
166 |
+
"loss": 0.2938,
|
167 |
+
"step": 7000
|
168 |
+
},
|
169 |
+
{
|
170 |
+
"epoch": 8.87,
|
171 |
+
"eval_loss": 0.5169841051101685,
|
172 |
+
"eval_runtime": 164.5878,
|
173 |
+
"eval_samples_per_second": 28.696,
|
174 |
+
"eval_steps_per_second": 2.394,
|
175 |
+
"eval_wer": 0.18300878506943039,
|
176 |
+
"step": 7000
|
177 |
+
},
|
178 |
+
{
|
179 |
+
"epoch": 9.51,
|
180 |
+
"eval_loss": 0.5407580733299255,
|
181 |
+
"eval_runtime": 163.5687,
|
182 |
+
"eval_samples_per_second": 28.875,
|
183 |
+
"eval_steps_per_second": 2.409,
|
184 |
+
"eval_wer": 0.18149062790980122,
|
185 |
+
"step": 7500
|
186 |
+
},
|
187 |
+
{
|
188 |
+
"epoch": 10.14,
|
189 |
+
"learning_rate": 2.225555555555556e-05,
|
190 |
+
"loss": 0.2674,
|
191 |
+
"step": 8000
|
192 |
+
},
|
193 |
+
{
|
194 |
+
"epoch": 10.14,
|
195 |
+
"eval_loss": 0.5580623149871826,
|
196 |
+
"eval_runtime": 162.7994,
|
197 |
+
"eval_samples_per_second": 29.011,
|
198 |
+
"eval_steps_per_second": 2.42,
|
199 |
+
"eval_wer": 0.173272337152342,
|
200 |
+
"step": 8000
|
201 |
+
},
|
202 |
+
{
|
203 |
+
"epoch": 10.77,
|
204 |
+
"eval_loss": 0.538887083530426,
|
205 |
+
"eval_runtime": 163.5851,
|
206 |
+
"eval_samples_per_second": 28.872,
|
207 |
+
"eval_steps_per_second": 2.409,
|
208 |
+
"eval_wer": 0.1719161167564066,
|
209 |
+
"step": 8500
|
210 |
+
},
|
211 |
+
{
|
212 |
+
"epoch": 11.41,
|
213 |
+
"learning_rate": 1.1144444444444445e-05,
|
214 |
+
"loss": 0.24,
|
215 |
+
"step": 9000
|
216 |
+
},
|
217 |
+
{
|
218 |
+
"epoch": 11.41,
|
219 |
+
"eval_loss": 0.5344488024711609,
|
220 |
+
"eval_runtime": 163.7329,
|
221 |
+
"eval_samples_per_second": 28.846,
|
222 |
+
"eval_steps_per_second": 2.406,
|
223 |
+
"eval_wer": 0.17141006436986356,
|
224 |
+
"step": 9000
|
225 |
+
},
|
226 |
+
{
|
227 |
+
"epoch": 12.04,
|
228 |
+
"eval_loss": 0.5502966046333313,
|
229 |
+
"eval_runtime": 154.8226,
|
230 |
+
"eval_samples_per_second": 30.506,
|
231 |
+
"eval_steps_per_second": 2.545,
|
232 |
+
"eval_wer": 0.16863689729160763,
|
233 |
+
"step": 9500
|
234 |
+
},
|
235 |
+
{
|
236 |
+
"epoch": 12.67,
|
237 |
+
"learning_rate": 3.3333333333333334e-08,
|
238 |
+
"loss": 0.211,
|
239 |
+
"step": 10000
|
240 |
+
},
|
241 |
+
{
|
242 |
+
"epoch": 12.67,
|
243 |
+
"eval_loss": 0.5595183372497559,
|
244 |
+
"eval_runtime": 157.1471,
|
245 |
+
"eval_samples_per_second": 30.055,
|
246 |
+
"eval_steps_per_second": 2.507,
|
247 |
+
"eval_wer": 0.16865713938706936,
|
248 |
+
"step": 10000
|
249 |
+
},
|
250 |
+
{
|
251 |
+
"epoch": 12.67,
|
252 |
+
"step": 10000,
|
253 |
+
"total_flos": 2.7136568655380324e+19,
|
254 |
+
"train_loss": 0.6397389862060547,
|
255 |
+
"train_runtime": 13054.0758,
|
256 |
+
"train_samples_per_second": 18.385,
|
257 |
+
"train_steps_per_second": 0.766
|
258 |
+
}
|
259 |
+
],
|
260 |
+
"max_steps": 10000,
|
261 |
+
"num_train_epochs": 13,
|
262 |
+
"total_flos": 2.7136568655380324e+19,
|
263 |
+
"trial_name": null,
|
264 |
+
"trial_params": null
|
265 |
+
}
|
training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f2cb36695fb66bf763f5c29a45137dfc2f7f8480ce24a65a0341e1f64851ddc3
|
3 |
+
size 3771
|
vocab.json
ADDED
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"[PAD]": 28,
|
3 |
+
"[UNK]": 27,
|
4 |
+
"a": 1,
|
5 |
+
"b": 2,
|
6 |
+
"c": 3,
|
7 |
+
"d": 4,
|
8 |
+
"e": 5,
|
9 |
+
"f": 6,
|
10 |
+
"g": 7,
|
11 |
+
"h": 8,
|
12 |
+
"i": 9,
|
13 |
+
"j": 10,
|
14 |
+
"k": 11,
|
15 |
+
"l": 12,
|
16 |
+
"m": 13,
|
17 |
+
"n": 14,
|
18 |
+
"o": 15,
|
19 |
+
"p": 16,
|
20 |
+
"q": 17,
|
21 |
+
"r": 18,
|
22 |
+
"s": 19,
|
23 |
+
"t": 20,
|
24 |
+
"u": 21,
|
25 |
+
"v": 22,
|
26 |
+
"w": 23,
|
27 |
+
"x": 24,
|
28 |
+
"y": 25,
|
29 |
+
"z": 26,
|
30 |
+
"|": 0
|
31 |
+
}
|