Vadim commited on
Commit
e1d3334
·
1 Parent(s): 8ce8347

update README.md

Browse files
Files changed (2) hide show
  1. .gitignore +86 -0
  2. README.md +95 -3
.gitignore ADDED
@@ -0,0 +1,86 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ .idea
2
+ __pycache__/
3
+ *.py[cod]
4
+ *$py.class
5
+
6
+ # C extensions
7
+ *.so
8
+
9
+ # Distribution / packaging
10
+ .Python
11
+ build/
12
+ develop-eggs/
13
+ dist/
14
+ downloads/
15
+ eggs/
16
+ .eggs/
17
+ lib/
18
+ lib64/
19
+ parts/
20
+ sdist/
21
+ var/
22
+ wheels/
23
+ share/python-wheels/
24
+ *.egg-info/
25
+ .installed.cfg
26
+ *.egg
27
+ MANIFEST
28
+
29
+ # PyInstaller
30
+ # Usually these files are written by a python script from a template
31
+ # before PyInstaller builds the exe, so as to inject date/other infos into it.
32
+ *.manifest
33
+ *.spec
34
+
35
+ # Installer logs
36
+ pip-log.txt
37
+ pip-delete-this-directory.txt
38
+
39
+ # Flask stuff:
40
+ instance/
41
+ .webassets-cache
42
+
43
+ # Scrapy stuff:
44
+ .scrapy
45
+
46
+ # Sphinx documentation
47
+ docs/_build/
48
+
49
+ # PyBuilder
50
+ .pybuilder/
51
+ target/
52
+
53
+ # Jupyter Notebook
54
+ .ipynb_checkpoints
55
+
56
+ # IPython
57
+ profile_default/
58
+ ipython_config.py
59
+
60
+ # PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
61
+ __pypackages__/
62
+
63
+ # Celery stuff
64
+ celerybeat-schedule
65
+ celerybeat.pid
66
+
67
+ # SageMath parsed files
68
+ *.sage.py
69
+
70
+ # Environments
71
+ .env
72
+ .venv
73
+ env/
74
+ venv/
75
+ ENV/
76
+ env.bak/
77
+ venv.bak/
78
+
79
+ *.py
80
+
81
+ best_dev-*.data-*-of-*
82
+ best_dev-*.index
83
+ best_dev-*.meta
84
+ train-*.data-*-of-*
85
+ train-*.index
86
+ train-*.meta
README.md CHANGED
@@ -4,10 +4,102 @@ language:
4
  - ru
5
  ---
6
 
 
7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8
  ```shell
9
- apt-get install libopusfile0 libopus-dev libopusfile-dev
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
10
 
11
- pip install -U pip
12
- pip install coqui_stt_training
13
  ```
 
4
  - ru
5
  ---
6
 
7
+ ### [coqui-ai](https://github.com/coqui-ai/STT)
8
 
9
+ ### Models
10
+ - v51
11
+ - language: **uk**
12
+ - Epoch: **9**
13
+ - Loss: **5.813825**
14
+ - Test: **WER: 0.163878, CER: 0.033506, loss: 4.564916**
15
+ - v50
16
+ - language: **uk-RU**
17
+ - Epoch: **9**
18
+ - Loss: **8.292612**
19
+ - v49
20
+ - language: **uk-RU**
21
+ - Epoch: **8**
22
+ - Loss: **30.810781**
23
+ - Test: **WER: 0.400063, CER: 0.189420, loss: 53.694748**
24
+ - v48
25
+ - language: **uk**
26
+ - Epoch: **25**
27
+ - Loss: **26.365822**
28
+ - Test: **WER: 0.239203, CER: 0.123846, loss: 32.882496**
29
+
30
+
31
+ ### Використання:
32
  ```shell
33
+ docker run --gpus all --ulimit memlock=-1 --ulimit stack=67108864 --shm-size=18gb -it --rm -v ${PWD}/ds:/ds -v ${PWD}/../../datasets:/datasets ghcr.io/coqui-ai/stt-train:latest
34
+ ```
35
+
36
+ ### Приклад тренування:
37
+ ```python
38
+ #!/usr/bin/env python
39
+ import os
40
+ from coqui_stt_training.util.config import initialize_globals_from_args
41
+ from coqui_stt_training.train import train
42
+ from coqui_stt_training.evaluate import test
43
+
44
+ # only one GPU for only one training sample
45
+ # os.environ["CUDA_VISIBLE_DEVICES"] = "0"
46
+ # os.environ["TF_ENABLE_DEPRECATION_WARNINGS"] = "1"
47
+
48
+ # TF_ENABLE_DEPRECATION_WARNINGS=1 python /ds/train_uk.py
49
+ # CUDA_VISIBLE_DEVICES=0 python /ds/train_uk.py
50
+
51
+ initialize_globals_from_args(
52
+ # drop_source_layers=1,
53
+ # load_train='last',
54
+ use_allow_growth=True,
55
+ bytes_output_mode=True,
56
+ force_initialize_learning_rate=False,
57
+
58
+ # load_train="auto",
59
+ # load_evaluate="auto",
60
+ #
61
+ load_cudnn=False,
62
+ train_cudnn=True,
63
+ n_hidden=1024,
64
+ epochs=5,
65
+ learning_rate=0.01,
66
+ dropout_rate=0.5,
67
+
68
+ # shuffle_batches=True,
69
+
70
+ show_progressbar=True,
71
+ early_stop=False,
72
+
73
+ # alphabet_config_path="/ds/alphabet.txt",
74
+ train_files=['/ds/train_uk.sdb'],
75
+ train_batch_size=32,
76
+ dev_files=['/ds/dev_uk.sdb'],
77
+ dev_batch_size=32,
78
+ test_files=['/ds/test_uk.sdb'],
79
+ test_batch_size=12,
80
+
81
+ feature_cache='/ds/tmp/feature_cache_sdb',
82
+
83
+ checkpoint_dir='/ds/ckpt_sdb',
84
+ # load_checkpoint_dir='/ds/coqui-stt-1.4.0-checkpoint',
85
+ # save_checkpoint_dir='/ds/ckpt_sdb',
86
+
87
+ scorer_path='',
88
+
89
+ export_dir='/ds/export_train_sdb',
90
+ export_tflite=False,
91
+ export_language='Ukraine (uk-UA)',
92
+ export_author_id="Vadim",
93
+ export_model_name="ds_uk",
94
+ export_model_version="0.0.1",
95
+ # audio_sample_rate=16000,
96
+ # max_to_keep=5,
97
+ # plateau_epochs=3,
98
+ # checkpoint_secs=1200,
99
+ # use_allow_growth=True,
100
+ )
101
+
102
+ train()
103
 
104
+ test()
 
105
  ```