import from zenodo
Browse files- README.md +50 -0
- exp/asr_stats_raw_char_sp/train/feats_stats.npz +0 -0
- exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/RESULTS.md +231 -0
- exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/config.yaml +224 -0
- exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/acc.png +0 -0
- exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/backward_time.png +0 -0
- exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/cer.png +0 -0
- exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/cer_ctc.png +0 -0
- exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/forward_time.png +0 -0
- exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/iter_time.png +0 -0
- exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/loss.png +0 -0
- exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/loss_att.png +0 -0
- exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/loss_ctc.png +0 -0
- exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/lr_0.png +0 -0
- exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/optim_step_time.png +0 -0
- exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/train_time.png +0 -0
- exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/wer.png +0 -0
- exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/valid.acc.ave_10best.pth +3 -0
- exp/lm_train_lm_transformer_char/config.yaml +174 -0
- exp/lm_train_lm_transformer_char/images/backward_time.png +0 -0
- exp/lm_train_lm_transformer_char/images/forward_time.png +0 -0
- exp/lm_train_lm_transformer_char/images/iter_time.png +0 -0
- exp/lm_train_lm_transformer_char/images/loss.png +0 -0
- exp/lm_train_lm_transformer_char/images/lr_0.png +0 -0
- exp/lm_train_lm_transformer_char/images/optim_step_time.png +0 -0
- exp/lm_train_lm_transformer_char/images/train_time.png +0 -0
- exp/lm_train_lm_transformer_char/perplexity_test/ppl +1 -0
- exp/lm_train_lm_transformer_char/valid.loss.ave_10best.pth +3 -0
- meta.yaml +10 -0
README.md
ADDED
@@ -0,0 +1,50 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
tags:
|
3 |
+
- espnet
|
4 |
+
- audio
|
5 |
+
- automatic-speech-recognition
|
6 |
+
language: en
|
7 |
+
datasets:
|
8 |
+
- reverb
|
9 |
+
license: cc-by-4.0
|
10 |
+
---
|
11 |
+
## Example ESPnet2 ASR model
|
12 |
+
### `kamo-naoyuki/reverb_asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp_valid.acc.ave`
|
13 |
+
♻️ Imported from https://zenodo.org/record/4278363/
|
14 |
+
|
15 |
+
This model was trained by kamo-naoyuki using reverb/asr1 recipe in [espnet](https://github.com/espnet/espnet/).
|
16 |
+
### Demo: How to use in ESPnet2
|
17 |
+
```python
|
18 |
+
# coming soon
|
19 |
+
```
|
20 |
+
### Citing ESPnet
|
21 |
+
```BibTex
|
22 |
+
@inproceedings{watanabe2018espnet,
|
23 |
+
author={Shinji Watanabe and Takaaki Hori and Shigeki Karita and Tomoki Hayashi and Jiro Nishitoba and Yuya Unno and Nelson {Enrique Yalta Soplin} and Jahn Heymann and Matthew Wiesner and Nanxin Chen and Adithya Renduchintala and Tsubasa Ochiai},
|
24 |
+
title={{ESPnet}: End-to-End Speech Processing Toolkit},
|
25 |
+
year={2018},
|
26 |
+
booktitle={Proceedings of Interspeech},
|
27 |
+
pages={2207--2211},
|
28 |
+
doi={10.21437/Interspeech.2018-1456},
|
29 |
+
url={http://dx.doi.org/10.21437/Interspeech.2018-1456}
|
30 |
+
}
|
31 |
+
@inproceedings{hayashi2020espnet,
|
32 |
+
title={{Espnet-TTS}: Unified, reproducible, and integratable open source end-to-end text-to-speech toolkit},
|
33 |
+
author={Hayashi, Tomoki and Yamamoto, Ryuichi and Inoue, Katsuki and Yoshimura, Takenori and Watanabe, Shinji and Toda, Tomoki and Takeda, Kazuya and Zhang, Yu and Tan, Xu},
|
34 |
+
booktitle={Proceedings of IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
|
35 |
+
pages={7654--7658},
|
36 |
+
year={2020},
|
37 |
+
organization={IEEE}
|
38 |
+
}
|
39 |
+
```
|
40 |
+
or arXiv:
|
41 |
+
```bibtex
|
42 |
+
@misc{watanabe2018espnet,
|
43 |
+
title={ESPnet: End-to-End Speech Processing Toolkit},
|
44 |
+
author={Shinji Watanabe and Takaaki Hori and Shigeki Karita and Tomoki Hayashi and Jiro Nishitoba and Yuya Unno and Nelson Enrique Yalta Soplin and Jahn Heymann and Matthew Wiesner and Nanxin Chen and Adithya Renduchintala and Tsubasa Ochiai},
|
45 |
+
year={2018},
|
46 |
+
eprint={1804.00015},
|
47 |
+
archivePrefix={arXiv},
|
48 |
+
primaryClass={cs.CL}
|
49 |
+
}
|
50 |
+
```
|
exp/asr_stats_raw_char_sp/train/feats_stats.npz
ADDED
Binary file (1.4 kB). View file
|
|
exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/RESULTS.md
ADDED
@@ -0,0 +1,231 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
<!-- Generated by scripts/utils/show_asr_result.sh -->
|
2 |
+
# RESULTS
|
3 |
+
## Environments
|
4 |
+
- date: `Tue Nov 17 22:56:55 JST 2020`
|
5 |
+
- python version: `3.7.3 (default, Mar 27 2019, 22:11:17) [GCC 7.3.0]`
|
6 |
+
- espnet version: `espnet 0.9.5`
|
7 |
+
- pytorch version: `pytorch 1.5.1`
|
8 |
+
- Git hash: `7aad8240a2c0643289ce6ea76d6f42eb12c15674`
|
9 |
+
- Commit date: `Tue Nov 10 22:46:38 2020 +0900`
|
10 |
+
|
11 |
+
## asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp
|
12 |
+
### WER
|
13 |
+
|
14 |
+
|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
|
15 |
+
|---|---|---|---|---|---|---|---|---|
|
16 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_1ch|179|3066|91.4|6.7|1.9|0.9|9.5|55.3|
|
17 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_1ch_far|89|1463|90.8|7.4|1.8|1.4|10.5|60.7|
|
18 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_1ch_near|90|1603|92.0|6.0|2.1|0.6|8.6|50.0|
|
19 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_1ch_wpe|179|3066|91.8|6.3|1.9|0.8|9.0|54.2|
|
20 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_1ch_wpe_far|89|1463|91.4|6.8|1.8|1.1|9.7|58.4|
|
21 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_1ch_wpe_near|90|1603|92.2|5.7|2.1|0.6|8.4|50.0|
|
22 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_8ch_beamformit|179|3066|95.7|3.4|0.9|0.5|4.9|35.8|
|
23 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_8ch_beamformit_far|89|1463|95.0|4.0|1.0|0.7|5.7|38.2|
|
24 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_8ch_beamformit_near|90|1603|96.3|2.8|0.9|0.4|4.1|33.3|
|
25 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_1ch|1484|24338|96.3|2.9|0.8|0.3|4.0|33.9|
|
26 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_1ch_far|742|12169|95.5|3.5|1.0|0.4|4.9|38.9|
|
27 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_1ch_near|742|12169|97.0|2.4|0.6|0.2|3.2|28.8|
|
28 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_1ch_wpe|1484|24338|96.4|2.9|0.8|0.3|4.0|33.4|
|
29 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_1ch_wpe_far|742|12169|95.7|3.4|0.9|0.4|4.8|38.0|
|
30 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_1ch_wpe_near|742|12169|97.0|2.3|0.6|0.2|3.2|28.8|
|
31 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_8ch_beamformit|1484|24338|97.3|2.1|0.6|0.2|2.9|26.6|
|
32 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_8ch_beamformit_far|742|12169|97.2|2.2|0.6|0.2|2.9|27.5|
|
33 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_8ch_beamformit_near|742|12169|97.3|2.1|0.6|0.2|2.8|25.7|
|
34 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch|372|6093|91.6|6.9|1.5|1.0|9.4|53.0|
|
35 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_far|186|2962|90.7|7.5|1.8|1.3|10.5|54.8|
|
36 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_near|186|3131|92.4|6.3|1.3|0.8|8.3|51.1|
|
37 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|6093|92.3|6.2|1.5|1.0|8.7|53.8|
|
38 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|2962|91.4|6.8|1.8|1.1|9.8|54.8|
|
39 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|3131|93.1|5.7|1.2|0.8|7.7|52.7|
|
40 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_8ch_beamformit|372|6093|96.3|3.0|0.7|0.6|4.3|33.6|
|
41 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_8ch_beamformit_far|186|2962|96.0|3.4|0.6|0.7|4.7|33.9|
|
42 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_8ch_beamformit_near|186|3131|96.5|2.6|0.9|0.5|4.0|33.3|
|
43 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_1ch|2176|35972|96.4|2.9|0.7|0.3|3.9|35.7|
|
44 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_1ch_far|1088|17986|95.7|3.6|0.8|0.4|4.7|40.2|
|
45 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_1ch_near|1088|17986|97.2|2.3|0.6|0.3|3.1|31.3|
|
46 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_1ch_wpe|2176|35972|96.5|2.8|0.6|0.3|3.7|35.1|
|
47 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_1ch_wpe_far|1088|17986|95.9|3.4|0.7|0.3|4.5|39.2|
|
48 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_1ch_wpe_near|1088|17986|97.2|2.3|0.5|0.2|3.0|31.1|
|
49 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_8ch_beamformit|2176|35972|97.4|2.0|0.5|0.2|2.8|29.0|
|
50 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_8ch_beamformit_far|1088|17986|97.4|2.1|0.5|0.2|2.9|29.5|
|
51 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_8ch_beamformit_near|1088|17986|97.5|2.0|0.5|0.2|2.7|28.6|
|
52 |
+
|decode_lm_weight0.5_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|6093|90.1|8.0|1.9|1.1|11.0|62.4|
|
53 |
+
|decode_lm_weight0.5_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|2962|88.8|9.0|2.2|1.4|12.6|64.0|
|
54 |
+
|decode_lm_weight0.5_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|3131|91.3|7.0|1.7|0.8|9.5|60.8|
|
55 |
+
|decode_lm_weight0.5_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|6093|90.5|8.1|1.5|1.2|10.7|61.3|
|
56 |
+
|decode_lm_weight0.5_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|2962|89.7|8.7|1.5|1.5|11.7|64.0|
|
57 |
+
|decode_lm_weight0.5_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|3131|91.2|7.4|1.4|0.9|9.7|58.6|
|
58 |
+
|decode_lm_weight0.5_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|6093|75.0|22.7|2.3|2.9|27.9|88.7|
|
59 |
+
|decode_lm_weight0.5_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|2962|74.5|23.1|2.5|3.3|28.8|89.8|
|
60 |
+
|decode_lm_weight0.5_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|3131|75.6|22.4|2.1|2.6|27.1|87.6|
|
61 |
+
|decode_lm_weight0.8_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|6093|91.1|6.6|2.3|0.9|9.7|56.5|
|
62 |
+
|decode_lm_weight0.8_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|2962|89.8|7.6|2.6|1.1|11.2|58.6|
|
63 |
+
|decode_lm_weight0.8_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|3131|92.3|5.7|2.0|0.6|8.3|54.3|
|
64 |
+
|decode_lm_weight0.8_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|6093|92.1|6.5|1.4|1.1|9.0|54.8|
|
65 |
+
|decode_lm_weight0.8_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|2962|91.3|7.1|1.6|1.4|10.1|57.5|
|
66 |
+
|decode_lm_weight0.8_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|3131|92.8|5.9|1.3|0.8|8.0|52.2|
|
67 |
+
|decode_lm_weight0.8_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|6093|81.6|16.6|1.8|2.4|20.9|79.8|
|
68 |
+
|decode_lm_weight0.8_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|2962|80.8|17.3|1.9|2.8|22.1|81.2|
|
69 |
+
|decode_lm_weight0.8_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|3131|82.3|16.0|1.7|2.0|19.7|78.5|
|
70 |
+
|decode_lm_weight1.1_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|6093|90.5|6.5|3.0|0.9|10.4|57.0|
|
71 |
+
|decode_lm_weight1.1_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|2962|89.5|7.3|3.1|1.1|11.5|57.5|
|
72 |
+
|decode_lm_weight1.1_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|3131|91.5|5.7|2.8|0.8|9.3|56.5|
|
73 |
+
|decode_lm_weight1.1_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|6093|92.3|6.2|1.5|1.0|8.7|53.8|
|
74 |
+
|decode_lm_weight1.1_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|2962|91.4|6.9|1.8|1.2|9.9|55.4|
|
75 |
+
|decode_lm_weight1.1_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|3131|93.2|5.6|1.2|0.7|7.5|52.2|
|
76 |
+
|decode_lm_weight1.1_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|6093|84.8|13.7|1.5|2.2|17.3|73.7|
|
77 |
+
|decode_lm_weight1.1_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|2962|83.6|14.9|1.5|2.7|19.1|76.9|
|
78 |
+
|decode_lm_weight1.1_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|3131|86.0|12.6|1.4|1.7|15.7|70.4|
|
79 |
+
|decode_lm_weight1.2_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|6093|90.0|6.9|3.2|0.9|10.9|57.8|
|
80 |
+
|decode_lm_weight1.2_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|2962|89.2|7.6|3.2|1.2|12.0|58.6|
|
81 |
+
|decode_lm_weight1.2_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|3131|90.7|6.2|3.1|0.7|9.9|57.0|
|
82 |
+
|decode_lm_weight1.2_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|6093|92.3|6.2|1.5|1.0|8.7|53.8|
|
83 |
+
|decode_lm_weight1.2_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|2962|91.4|6.8|1.8|1.1|9.8|54.8|
|
84 |
+
|decode_lm_weight1.2_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|3131|93.1|5.7|1.2|0.8|7.7|52.7|
|
85 |
+
|decode_lm_weight1.2_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|6093|85.3|13.2|1.5|2.1|16.8|73.7|
|
86 |
+
|decode_lm_weight1.2_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|2962|84.3|14.1|1.6|2.5|18.2|76.3|
|
87 |
+
|decode_lm_weight1.2_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|3131|86.2|12.4|1.4|1.8|15.6|71.0|
|
88 |
+
|decode_lm_weight1.3_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|6093|89.3|7.2|3.5|1.0|11.7|59.9|
|
89 |
+
|decode_lm_weight1.3_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|2962|88.1|8.2|3.7|1.2|13.1|61.3|
|
90 |
+
|decode_lm_weight1.3_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|3131|90.4|6.3|3.3|0.8|10.4|58.6|
|
91 |
+
|decode_lm_weight1.3_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|6093|92.1|6.3|1.5|1.1|9.0|54.0|
|
92 |
+
|decode_lm_weight1.3_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|2962|91.1|7.0|1.9|1.4|10.3|55.4|
|
93 |
+
|decode_lm_weight1.3_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|3131|93.1|5.7|1.2|0.9|7.8|52.7|
|
94 |
+
|decode_lm_weight1.3_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|6093|85.8|12.7|1.5|2.0|16.2|73.1|
|
95 |
+
|decode_lm_weight1.3_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|2962|85.0|13.4|1.6|2.4|17.4|75.8|
|
96 |
+
|decode_lm_weight1.3_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|3131|86.6|12.1|1.4|1.6|15.0|70.4|
|
97 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_1ch_wpe|179|3066|91.7|6.6|1.7|1.1|9.5|55.9|
|
98 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_1ch_wpe_far|89|1463|91.4|7.0|1.6|1.4|10.0|58.4|
|
99 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_1ch_wpe_near|90|1603|92.0|6.2|1.9|0.9|9.0|53.3|
|
100 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_8ch_beamformit|179|3066|95.4|3.9|0.7|1.0|5.6|36.9|
|
101 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_8ch_beamformit_far|89|1463|94.7|4.7|0.6|1.6|7.0|39.3|
|
102 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_8ch_beamformit_near|90|1603|96.0|3.2|0.8|0.4|4.4|34.4|
|
103 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_1ch_wpe|1484|24338|96.2|3.0|0.7|0.3|4.0|34.8|
|
104 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_1ch_wpe_far|742|12169|95.3|3.8|0.9|0.4|5.1|40.6|
|
105 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_1ch_wpe_near|742|12169|97.1|2.3|0.6|0.2|3.0|29.0|
|
106 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_8ch_beamformit|1484|24338|97.3|2.2|0.6|0.2|2.9|26.9|
|
107 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_8ch_beamformit_far|742|12169|97.2|2.2|0.6|0.2|3.0|27.0|
|
108 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_8ch_beamformit_near|742|12169|97.3|2.1|0.5|0.2|2.8|26.8|
|
109 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|6093|92.0|6.7|1.3|1.1|9.1|54.3|
|
110 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|2962|91.0|7.5|1.5|1.4|10.4|58.1|
|
111 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|3131|92.9|5.9|1.2|0.8|7.9|50.5|
|
112 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_8ch_beamformit|372|6093|96.2|3.2|0.6|0.7|4.4|34.4|
|
113 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_8ch_beamformit_far|186|2962|96.0|3.5|0.4|0.8|4.8|34.4|
|
114 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_8ch_beamformit_near|186|3131|96.4|2.9|0.7|0.5|4.1|34.4|
|
115 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_1ch_wpe|2176|35972|96.4|3.0|0.6|0.3|3.9|36.2|
|
116 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_1ch_wpe_far|1088|17986|95.8|3.5|0.7|0.3|4.6|40.4|
|
117 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_1ch_wpe_near|1088|17986|97.0|2.5|0.6|0.3|3.3|31.9|
|
118 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_8ch_beamformit|2176|35972|97.5|2.1|0.5|0.2|2.7|29.5|
|
119 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_8ch_beamformit_far|1088|17986|97.4|2.1|0.5|0.2|2.8|29.4|
|
120 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_8ch_beamformit_near|1088|17986|97.5|2.0|0.4|0.2|2.7|29.5|
|
121 |
+
|
122 |
+
### CER
|
123 |
+
|
124 |
+
|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
|
125 |
+
|---|---|---|---|---|---|---|---|---|
|
126 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_1ch|179|18181|95.8|2.0|2.2|1.1|5.3|55.3|
|
127 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_1ch_far|89|8845|95.5|2.2|2.2|1.2|5.7|60.7|
|
128 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_1ch_near|90|9336|96.1|1.8|2.1|0.9|4.8|50.0|
|
129 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_1ch_wpe|179|18181|96.1|1.9|2.1|1.0|4.9|54.2|
|
130 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_1ch_wpe_far|89|8845|95.9|1.9|2.2|1.2|5.3|58.4|
|
131 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_1ch_wpe_near|90|9336|96.3|1.8|1.9|0.8|4.5|50.0|
|
132 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_8ch_beamformit|179|18181|98.3|0.8|0.9|0.6|2.3|35.8|
|
133 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_8ch_beamformit_far|89|8845|98.0|1.0|1.1|0.7|2.8|38.2|
|
134 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_8ch_beamformit_near|90|9336|98.6|0.6|0.8|0.5|1.9|33.3|
|
135 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_1ch|1484|143048|98.4|0.7|0.8|0.3|1.9|34.2|
|
136 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_1ch_far|742|71524|98.1|0.9|1.0|0.4|2.3|39.2|
|
137 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_1ch_near|742|71524|98.8|0.5|0.7|0.2|1.4|29.1|
|
138 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_1ch_wpe|1484|143048|98.5|0.7|0.8|0.3|1.8|33.7|
|
139 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_1ch_wpe_far|742|71524|98.1|0.9|1.0|0.4|2.3|38.3|
|
140 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_1ch_wpe_near|742|71524|98.8|0.5|0.7|0.2|1.4|29.1|
|
141 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_8ch_beamformit|1484|143048|99.0|0.4|0.6|0.2|1.2|27.0|
|
142 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_8ch_beamformit_far|742|71524|99.0|0.4|0.6|0.2|1.2|27.9|
|
143 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_8ch_beamformit_near|742|71524|99.0|0.4|0.6|0.2|1.2|26.1|
|
144 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch|372|35694|96.0|2.1|1.9|1.0|5.0|53.0|
|
145 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_far|186|17261|95.5|2.3|2.2|1.3|5.8|54.8|
|
146 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_near|186|18433|96.5|1.9|1.6|0.8|4.3|51.1|
|
147 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|35694|96.4|1.9|1.6|1.0|4.5|53.8|
|
148 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|17261|95.8|2.2|2.0|1.1|5.3|54.8|
|
149 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|18433|97.0|1.7|1.3|0.8|3.8|52.7|
|
150 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_8ch_beamformit|372|35694|98.5|0.8|0.7|0.7|2.2|33.6|
|
151 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_8ch_beamformit_far|186|17261|98.4|0.9|0.7|0.8|2.4|33.9|
|
152 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_8ch_beamformit_near|186|18433|98.6|0.6|0.8|0.6|1.9|33.3|
|
153 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_1ch|2176|210960|98.6|0.7|0.8|0.3|1.8|36.4|
|
154 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_1ch_far|1088|105480|98.2|0.9|0.9|0.4|2.2|40.8|
|
155 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_1ch_near|1088|105480|98.9|0.5|0.6|0.2|1.3|32.0|
|
156 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_1ch_wpe|2176|210960|98.6|0.6|0.7|0.3|1.7|35.8|
|
157 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_1ch_wpe_far|1088|105480|98.4|0.8|0.9|0.4|2.0|39.8|
|
158 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_1ch_wpe_near|1088|105480|98.9|0.5|0.6|0.2|1.3|31.9|
|
159 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_8ch_beamformit|2176|210960|99.0|0.4|0.6|0.2|1.2|29.9|
|
160 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_8ch_beamformit_far|1088|105480|99.0|0.4|0.6|0.2|1.2|30.1|
|
161 |
+
|decode_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_8ch_beamformit_near|1088|105480|99.0|0.4|0.6|0.2|1.2|29.6|
|
162 |
+
|decode_lm_weight0.5_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|35694|95.6|2.3|2.1|1.2|5.6|62.4|
|
163 |
+
|decode_lm_weight0.5_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|17261|94.9|2.7|2.4|1.6|6.7|64.0|
|
164 |
+
|decode_lm_weight0.5_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|18433|96.2|2.0|1.8|0.8|4.6|60.8|
|
165 |
+
|decode_lm_weight0.5_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|35694|95.9|2.2|1.8|1.1|5.2|61.3|
|
166 |
+
|decode_lm_weight0.5_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|17261|95.5|2.5|2.0|1.3|5.8|64.0|
|
167 |
+
|decode_lm_weight0.5_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|18433|96.3|2.0|1.7|0.9|4.6|58.6|
|
168 |
+
|decode_lm_weight0.5_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|35694|91.6|4.6|3.9|2.6|11.1|87.6|
|
169 |
+
|decode_lm_weight0.5_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|17261|91.0|4.8|4.2|2.8|11.7|88.2|
|
170 |
+
|decode_lm_weight0.5_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|18433|92.1|4.4|3.6|2.5|10.4|87.1|
|
171 |
+
|decode_lm_weight0.8_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|35694|95.6|2.0|2.3|1.0|5.4|56.5|
|
172 |
+
|decode_lm_weight0.8_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|17261|95.0|2.4|2.7|1.3|6.3|58.6|
|
173 |
+
|decode_lm_weight0.8_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|18433|96.2|1.7|2.1|0.7|4.5|54.3|
|
174 |
+
|decode_lm_weight0.8_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|35694|96.4|2.0|1.6|1.0|4.6|54.8|
|
175 |
+
|decode_lm_weight0.8_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|17261|95.9|2.2|1.8|1.2|5.2|57.5|
|
176 |
+
|decode_lm_weight0.8_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|18433|96.9|1.8|1.4|0.8|3.9|52.2|
|
177 |
+
|decode_lm_weight0.8_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|35694|92.9|4.0|3.1|2.0|9.0|79.8|
|
178 |
+
|decode_lm_weight0.8_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|17261|92.5|4.3|3.2|2.2|9.7|81.2|
|
179 |
+
|decode_lm_weight0.8_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|18433|93.4|3.7|2.9|1.8|8.4|78.5|
|
180 |
+
|decode_lm_weight1.1_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|35694|94.9|2.2|3.0|1.1|6.2|57.0|
|
181 |
+
|decode_lm_weight1.1_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|17261|94.3|2.6|3.1|1.4|7.1|57.5|
|
182 |
+
|decode_lm_weight1.1_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|18433|95.4|1.8|2.8|0.9|5.5|56.5|
|
183 |
+
|decode_lm_weight1.1_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|35694|96.5|1.9|1.6|1.0|4.5|53.8|
|
184 |
+
|decode_lm_weight1.1_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|17261|96.0|2.1|1.9|1.2|5.2|55.4|
|
185 |
+
|decode_lm_weight1.1_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|18433|97.0|1.7|1.3|0.8|3.8|52.2|
|
186 |
+
|decode_lm_weight1.1_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|35694|93.7|3.7|2.7|1.6|8.0|73.7|
|
187 |
+
|decode_lm_weight1.1_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|17261|93.0|4.0|3.0|1.8|8.8|76.9|
|
188 |
+
|decode_lm_weight1.1_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|18433|94.3|3.3|2.4|1.4|7.2|70.4|
|
189 |
+
|decode_lm_weight1.2_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|35694|94.6|2.3|3.1|1.2|6.6|57.8|
|
190 |
+
|decode_lm_weight1.2_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|17261|94.2|2.7|3.2|1.4|7.3|58.6|
|
191 |
+
|decode_lm_weight1.2_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|18433|95.0|2.0|3.0|1.0|5.9|57.0|
|
192 |
+
|decode_lm_weight1.2_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|35694|96.4|1.9|1.6|1.0|4.5|53.8|
|
193 |
+
|decode_lm_weight1.2_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|17261|95.8|2.2|2.0|1.1|5.3|54.8|
|
194 |
+
|decode_lm_weight1.2_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|18433|97.0|1.7|1.3|0.8|3.8|52.7|
|
195 |
+
|decode_lm_weight1.2_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|35694|93.9|3.5|2.6|1.6|7.8|73.7|
|
196 |
+
|decode_lm_weight1.2_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|17261|93.2|3.8|2.9|1.9|8.6|76.3|
|
197 |
+
|decode_lm_weight1.2_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|18433|94.4|3.3|2.3|1.4|7.0|71.0|
|
198 |
+
|decode_lm_weight1.3_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|35694|94.1|2.5|3.4|1.3|7.1|59.9|
|
199 |
+
|decode_lm_weight1.3_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|17261|93.5|2.9|3.6|1.5|8.1|61.3|
|
200 |
+
|decode_lm_weight1.3_ctc_weight0.1_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|18433|94.8|2.1|3.2|1.0|6.2|58.6|
|
201 |
+
|decode_lm_weight1.3_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|35694|96.3|2.0|1.7|1.1|4.7|54.0|
|
202 |
+
|decode_lm_weight1.3_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|17261|95.7|2.3|2.0|1.3|5.6|55.4|
|
203 |
+
|decode_lm_weight1.3_ctc_weight0.3_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|18433|96.9|1.7|1.4|0.9|4.0|52.7|
|
204 |
+
|decode_lm_weight1.3_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|35694|94.0|3.5|2.5|1.6|7.6|73.1|
|
205 |
+
|decode_lm_weight1.3_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|17261|93.4|3.7|2.8|1.8|8.4|75.8|
|
206 |
+
|decode_lm_weight1.3_ctc_weight1.5_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|18433|94.5|3.2|2.3|1.4|6.9|70.4|
|
207 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_1ch_wpe|179|18181|96.2|2.0|1.8|1.0|4.8|55.9|
|
208 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_1ch_wpe_far|89|8845|96.1|2.1|1.8|1.0|5.0|58.4|
|
209 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_1ch_wpe_near|90|9336|96.3|1.8|1.8|1.0|4.6|53.3|
|
210 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_8ch_beamformit|179|18181|98.3|0.9|0.9|0.9|2.6|36.9|
|
211 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_8ch_beamformit_far|89|8845|98.0|1.2|0.9|1.2|3.3|39.3|
|
212 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_real_8ch_beamformit_near|90|9336|98.5|0.6|0.9|0.5|2.0|34.4|
|
213 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_1ch_wpe|1484|143048|98.5|0.7|0.8|0.3|1.8|35.0|
|
214 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_1ch_wpe_far|742|71524|98.0|0.9|1.0|0.4|2.3|40.8|
|
215 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_1ch_wpe_near|742|71524|98.9|0.4|0.6|0.2|1.3|29.2|
|
216 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_8ch_beamformit|1484|143048|99.0|0.4|0.6|0.2|1.2|27.2|
|
217 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_8ch_beamformit_far|742|71524|99.0|0.4|0.6|0.2|1.2|27.2|
|
218 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/dt_simu_8ch_beamformit_near|742|71524|99.0|0.4|0.6|0.2|1.1|27.2|
|
219 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe|372|35694|96.3|2.1|1.6|1.0|4.7|54.3|
|
220 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_far|186|17261|95.8|2.3|1.9|1.2|5.5|58.1|
|
221 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_1ch_wpe_near|186|18433|96.8|1.8|1.4|0.8|3.9|50.5|
|
222 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_8ch_beamformit|372|35694|98.6|0.8|0.6|0.7|2.1|34.4|
|
223 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_8ch_beamformit_far|186|17261|98.4|0.9|0.7|0.8|2.4|34.4|
|
224 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_real_8ch_beamformit_near|186|18433|98.7|0.6|0.6|0.5|1.8|34.4|
|
225 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_1ch_wpe|2176|210960|98.6|0.6|0.8|0.3|1.7|36.8|
|
226 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_1ch_wpe_far|1088|105480|98.3|0.8|0.9|0.4|2.1|41.0|
|
227 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_1ch_wpe_near|1088|105480|98.8|0.5|0.7|0.2|1.4|32.6|
|
228 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_8ch_beamformit|2176|210960|99.1|0.4|0.5|0.2|1.1|30.2|
|
229 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_8ch_beamformit_far|1088|105480|99.1|0.4|0.5|0.2|1.1|30.1|
|
230 |
+
|inference_lm_lm_train_lm_transformer_char_valid.loss.ave_asr_model_valid.acc.ave/et_simu_8ch_beamformit_near|1088|105480|99.1|0.4|0.5|0.2|1.1|30.3|
|
231 |
+
|
exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/config.yaml
ADDED
@@ -0,0 +1,224 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
config: ./conf/tuning/train_asr_transformer4.yaml
|
2 |
+
print_config: false
|
3 |
+
log_level: INFO
|
4 |
+
dry_run: false
|
5 |
+
iterator_type: sequence
|
6 |
+
output_dir: exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp
|
7 |
+
ngpu: 1
|
8 |
+
seed: 0
|
9 |
+
num_workers: 1
|
10 |
+
num_att_plot: 3
|
11 |
+
dist_backend: nccl
|
12 |
+
dist_init_method: env://
|
13 |
+
dist_world_size: 4
|
14 |
+
dist_rank: 0
|
15 |
+
local_rank: 0
|
16 |
+
dist_master_addr: localhost
|
17 |
+
dist_master_port: 39192
|
18 |
+
dist_launcher: null
|
19 |
+
multiprocessing_distributed: true
|
20 |
+
cudnn_enabled: true
|
21 |
+
cudnn_benchmark: false
|
22 |
+
cudnn_deterministic: true
|
23 |
+
collect_stats: false
|
24 |
+
write_collected_feats: false
|
25 |
+
max_epoch: 50
|
26 |
+
patience: null
|
27 |
+
val_scheduler_criterion:
|
28 |
+
- valid
|
29 |
+
- loss
|
30 |
+
early_stopping_criterion:
|
31 |
+
- valid
|
32 |
+
- loss
|
33 |
+
- min
|
34 |
+
best_model_criterion:
|
35 |
+
- - valid
|
36 |
+
- acc
|
37 |
+
- max
|
38 |
+
keep_nbest_models: 10
|
39 |
+
grad_clip: 5.0
|
40 |
+
grad_clip_type: 2.0
|
41 |
+
grad_noise: false
|
42 |
+
accum_grad: 1
|
43 |
+
no_forward_run: false
|
44 |
+
resume: true
|
45 |
+
train_dtype: float32
|
46 |
+
use_amp: false
|
47 |
+
log_interval: null
|
48 |
+
pretrain_path: []
|
49 |
+
pretrain_key: []
|
50 |
+
num_iters_per_epoch: null
|
51 |
+
batch_size: 20
|
52 |
+
valid_batch_size: null
|
53 |
+
batch_bins: 16000000
|
54 |
+
valid_batch_bins: null
|
55 |
+
train_shape_file:
|
56 |
+
- exp/asr_stats_raw_char_sp/train/speech_shape
|
57 |
+
- exp/asr_stats_raw_char_sp/train/text_shape.char
|
58 |
+
valid_shape_file:
|
59 |
+
- exp/asr_stats_raw_char_sp/valid/speech_shape
|
60 |
+
- exp/asr_stats_raw_char_sp/valid/text_shape.char
|
61 |
+
batch_type: numel
|
62 |
+
valid_batch_type: null
|
63 |
+
fold_length:
|
64 |
+
- 80000
|
65 |
+
- 150
|
66 |
+
sort_in_batch: descending
|
67 |
+
sort_batch: descending
|
68 |
+
multiple_iterator: false
|
69 |
+
chunk_length: 500
|
70 |
+
chunk_shift_ratio: 0.5
|
71 |
+
num_cache_chunks: 1024
|
72 |
+
train_data_path_and_name_and_type:
|
73 |
+
- - dump/raw/tr_simu_8ch_si284_sp/wav.scp
|
74 |
+
- speech
|
75 |
+
- kaldi_ark
|
76 |
+
- - dump/raw/tr_simu_8ch_si284_sp/text
|
77 |
+
- text
|
78 |
+
- text
|
79 |
+
valid_data_path_and_name_and_type:
|
80 |
+
- - dump/raw/dt_mult_1ch/wav.scp
|
81 |
+
- speech
|
82 |
+
- kaldi_ark
|
83 |
+
- - dump/raw/dt_mult_1ch/text
|
84 |
+
- text
|
85 |
+
- text
|
86 |
+
allow_variable_data_keys: false
|
87 |
+
max_cache_size: 0.0
|
88 |
+
max_cache_fd: 32
|
89 |
+
valid_max_cache_size: null
|
90 |
+
optim: adam
|
91 |
+
optim_conf:
|
92 |
+
lr: 0.005
|
93 |
+
scheduler: warmuplr
|
94 |
+
scheduler_conf:
|
95 |
+
warmup_steps: 30000
|
96 |
+
token_list:
|
97 |
+
- <blank>
|
98 |
+
- <unk>
|
99 |
+
- <space>
|
100 |
+
- E
|
101 |
+
- T
|
102 |
+
- A
|
103 |
+
- N
|
104 |
+
- I
|
105 |
+
- O
|
106 |
+
- S
|
107 |
+
- R
|
108 |
+
- H
|
109 |
+
- L
|
110 |
+
- D
|
111 |
+
- C
|
112 |
+
- U
|
113 |
+
- M
|
114 |
+
- P
|
115 |
+
- F
|
116 |
+
- G
|
117 |
+
- Y
|
118 |
+
- W
|
119 |
+
- B
|
120 |
+
- V
|
121 |
+
- K
|
122 |
+
- .
|
123 |
+
- X
|
124 |
+
- ''''
|
125 |
+
- J
|
126 |
+
- Q
|
127 |
+
- Z
|
128 |
+
- <NOISE>
|
129 |
+
- ','
|
130 |
+
- '-'
|
131 |
+
- '"'
|
132 |
+
- '*'
|
133 |
+
- ':'
|
134 |
+
- (
|
135 |
+
- )
|
136 |
+
- '?'
|
137 |
+
- '!'
|
138 |
+
- '&'
|
139 |
+
- ;
|
140 |
+
- '1'
|
141 |
+
- '2'
|
142 |
+
- '0'
|
143 |
+
- /
|
144 |
+
- $
|
145 |
+
- '{'
|
146 |
+
- '}'
|
147 |
+
- '8'
|
148 |
+
- '9'
|
149 |
+
- '6'
|
150 |
+
- '3'
|
151 |
+
- '5'
|
152 |
+
- '7'
|
153 |
+
- '4'
|
154 |
+
- '~'
|
155 |
+
- '`'
|
156 |
+
- _
|
157 |
+
- <*IN*>
|
158 |
+
- <*MR.*>
|
159 |
+
- \
|
160 |
+
- ^
|
161 |
+
- <sos/eos>
|
162 |
+
init: xavier_uniform
|
163 |
+
input_size: null
|
164 |
+
ctc_conf:
|
165 |
+
dropout_rate: 0.0
|
166 |
+
ctc_type: builtin
|
167 |
+
reduce: true
|
168 |
+
model_conf:
|
169 |
+
ctc_weight: 0.3
|
170 |
+
lsm_weight: 0.1
|
171 |
+
length_normalized_loss: false
|
172 |
+
use_preprocessor: true
|
173 |
+
token_type: char
|
174 |
+
bpemodel: null
|
175 |
+
non_linguistic_symbols: data/nlsyms.txt
|
176 |
+
cleaner: null
|
177 |
+
g2p: null
|
178 |
+
waveaug: null
|
179 |
+
waveaug_conf: {}
|
180 |
+
frontend: default
|
181 |
+
frontend_conf:
|
182 |
+
fs: 16k
|
183 |
+
specaug: specaug
|
184 |
+
specaug_conf:
|
185 |
+
apply_time_warp: true
|
186 |
+
time_warp_window: 5
|
187 |
+
time_warp_mode: bicubic
|
188 |
+
apply_freq_mask: true
|
189 |
+
freq_mask_width_range:
|
190 |
+
- 0
|
191 |
+
- 30
|
192 |
+
num_freq_mask: 2
|
193 |
+
apply_time_mask: true
|
194 |
+
time_mask_width_range:
|
195 |
+
- 0
|
196 |
+
- 40
|
197 |
+
num_time_mask: 2
|
198 |
+
normalize: global_mvn
|
199 |
+
normalize_conf:
|
200 |
+
stats_file: exp/asr_stats_raw_char_sp/train/feats_stats.npz
|
201 |
+
encoder: transformer
|
202 |
+
encoder_conf:
|
203 |
+
output_size: 256
|
204 |
+
attention_heads: 4
|
205 |
+
linear_units: 2048
|
206 |
+
num_blocks: 12
|
207 |
+
dropout_rate: 0.1
|
208 |
+
positional_dropout_rate: 0.1
|
209 |
+
attention_dropout_rate: 0.0
|
210 |
+
input_layer: conv2d
|
211 |
+
normalize_before: true
|
212 |
+
decoder: transformer
|
213 |
+
decoder_conf:
|
214 |
+
attention_heads: 4
|
215 |
+
linear_units: 2048
|
216 |
+
num_blocks: 6
|
217 |
+
dropout_rate: 0.1
|
218 |
+
positional_dropout_rate: 0.1
|
219 |
+
self_attention_dropout_rate: 0.0
|
220 |
+
src_attention_dropout_rate: 0.0
|
221 |
+
required:
|
222 |
+
- output_dir
|
223 |
+
- token_list
|
224 |
+
distributed: true
|
exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/acc.png
ADDED
exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/backward_time.png
ADDED
exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/cer.png
ADDED
exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/cer_ctc.png
ADDED
exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/forward_time.png
ADDED
exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/iter_time.png
ADDED
exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/loss.png
ADDED
exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/loss_att.png
ADDED
exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/loss_ctc.png
ADDED
exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/lr_0.png
ADDED
exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/optim_step_time.png
ADDED
exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/train_time.png
ADDED
exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/images/wer.png
ADDED
exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/valid.acc.ave_10best.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:eb8240351c0e6b53f28a18a0bf87fffdac9537796ca7ab8f516b7c3d810c976b
|
3 |
+
size 108733058
|
exp/lm_train_lm_transformer_char/config.yaml
ADDED
@@ -0,0 +1,174 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
config: ./conf/tuning/train_lm_transformer.yaml
|
2 |
+
print_config: false
|
3 |
+
log_level: INFO
|
4 |
+
dry_run: false
|
5 |
+
iterator_type: sequence
|
6 |
+
output_dir: exp/lm_train_lm_transformer_char
|
7 |
+
ngpu: 1
|
8 |
+
seed: 0
|
9 |
+
num_workers: 1
|
10 |
+
num_att_plot: 3
|
11 |
+
dist_backend: nccl
|
12 |
+
dist_init_method: env://
|
13 |
+
dist_world_size: 4
|
14 |
+
dist_rank: 0
|
15 |
+
local_rank: 0
|
16 |
+
dist_master_addr: localhost
|
17 |
+
dist_master_port: 37417
|
18 |
+
dist_launcher: null
|
19 |
+
multiprocessing_distributed: true
|
20 |
+
cudnn_enabled: true
|
21 |
+
cudnn_benchmark: false
|
22 |
+
cudnn_deterministic: true
|
23 |
+
collect_stats: false
|
24 |
+
write_collected_feats: false
|
25 |
+
max_epoch: 25
|
26 |
+
patience: null
|
27 |
+
val_scheduler_criterion:
|
28 |
+
- valid
|
29 |
+
- loss
|
30 |
+
early_stopping_criterion:
|
31 |
+
- valid
|
32 |
+
- loss
|
33 |
+
- min
|
34 |
+
best_model_criterion:
|
35 |
+
- - valid
|
36 |
+
- loss
|
37 |
+
- min
|
38 |
+
keep_nbest_models: 10
|
39 |
+
grad_clip: 5.0
|
40 |
+
grad_clip_type: 2.0
|
41 |
+
grad_noise: false
|
42 |
+
accum_grad: 2
|
43 |
+
no_forward_run: false
|
44 |
+
resume: true
|
45 |
+
train_dtype: float32
|
46 |
+
use_amp: false
|
47 |
+
log_interval: null
|
48 |
+
pretrain_path: []
|
49 |
+
pretrain_key: []
|
50 |
+
num_iters_per_epoch: null
|
51 |
+
batch_size: 20
|
52 |
+
valid_batch_size: null
|
53 |
+
batch_bins: 350000
|
54 |
+
valid_batch_bins: null
|
55 |
+
train_shape_file:
|
56 |
+
- exp/lm_stats_char/train/text_shape.char
|
57 |
+
valid_shape_file:
|
58 |
+
- exp/lm_stats_char/valid/text_shape.char
|
59 |
+
batch_type: numel
|
60 |
+
valid_batch_type: null
|
61 |
+
fold_length:
|
62 |
+
- 150
|
63 |
+
sort_in_batch: descending
|
64 |
+
sort_batch: descending
|
65 |
+
multiple_iterator: false
|
66 |
+
chunk_length: 500
|
67 |
+
chunk_shift_ratio: 0.5
|
68 |
+
num_cache_chunks: 1024
|
69 |
+
train_data_path_and_name_and_type:
|
70 |
+
- - dump/raw/lm_train.txt
|
71 |
+
- text
|
72 |
+
- text
|
73 |
+
valid_data_path_and_name_and_type:
|
74 |
+
- - dump/raw/dt_mult_1ch/text
|
75 |
+
- text
|
76 |
+
- text
|
77 |
+
allow_variable_data_keys: false
|
78 |
+
max_cache_size: 0.0
|
79 |
+
max_cache_fd: 32
|
80 |
+
valid_max_cache_size: null
|
81 |
+
optim: adam
|
82 |
+
optim_conf:
|
83 |
+
lr: 0.001
|
84 |
+
scheduler: warmuplr
|
85 |
+
scheduler_conf:
|
86 |
+
warmup_steps: 25000
|
87 |
+
token_list:
|
88 |
+
- <blank>
|
89 |
+
- <unk>
|
90 |
+
- <space>
|
91 |
+
- E
|
92 |
+
- T
|
93 |
+
- A
|
94 |
+
- N
|
95 |
+
- I
|
96 |
+
- O
|
97 |
+
- S
|
98 |
+
- R
|
99 |
+
- H
|
100 |
+
- L
|
101 |
+
- D
|
102 |
+
- C
|
103 |
+
- U
|
104 |
+
- M
|
105 |
+
- P
|
106 |
+
- F
|
107 |
+
- G
|
108 |
+
- Y
|
109 |
+
- W
|
110 |
+
- B
|
111 |
+
- V
|
112 |
+
- K
|
113 |
+
- .
|
114 |
+
- X
|
115 |
+
- ''''
|
116 |
+
- J
|
117 |
+
- Q
|
118 |
+
- Z
|
119 |
+
- <NOISE>
|
120 |
+
- ','
|
121 |
+
- '-'
|
122 |
+
- '"'
|
123 |
+
- '*'
|
124 |
+
- ':'
|
125 |
+
- (
|
126 |
+
- )
|
127 |
+
- '?'
|
128 |
+
- '!'
|
129 |
+
- '&'
|
130 |
+
- ;
|
131 |
+
- '1'
|
132 |
+
- '2'
|
133 |
+
- '0'
|
134 |
+
- /
|
135 |
+
- $
|
136 |
+
- '{'
|
137 |
+
- '}'
|
138 |
+
- '8'
|
139 |
+
- '9'
|
140 |
+
- '6'
|
141 |
+
- '3'
|
142 |
+
- '5'
|
143 |
+
- '7'
|
144 |
+
- '4'
|
145 |
+
- '~'
|
146 |
+
- '`'
|
147 |
+
- _
|
148 |
+
- <*IN*>
|
149 |
+
- <*MR.*>
|
150 |
+
- \
|
151 |
+
- ^
|
152 |
+
- <sos/eos>
|
153 |
+
init: null
|
154 |
+
model_conf:
|
155 |
+
ignore_id: 0
|
156 |
+
use_preprocessor: true
|
157 |
+
token_type: char
|
158 |
+
bpemodel: null
|
159 |
+
non_linguistic_symbols: data/nlsyms.txt
|
160 |
+
cleaner: null
|
161 |
+
g2p: null
|
162 |
+
lm: transformer
|
163 |
+
lm_conf:
|
164 |
+
pos_enc: null
|
165 |
+
embed_unit: 128
|
166 |
+
att_unit: 512
|
167 |
+
head: 8
|
168 |
+
unit: 2048
|
169 |
+
layer: 16
|
170 |
+
dropout_rate: 0.1
|
171 |
+
required:
|
172 |
+
- output_dir
|
173 |
+
- token_list
|
174 |
+
distributed: true
|
exp/lm_train_lm_transformer_char/images/backward_time.png
ADDED
exp/lm_train_lm_transformer_char/images/forward_time.png
ADDED
exp/lm_train_lm_transformer_char/images/iter_time.png
ADDED
exp/lm_train_lm_transformer_char/images/loss.png
ADDED
exp/lm_train_lm_transformer_char/images/lr_0.png
ADDED
exp/lm_train_lm_transformer_char/images/optim_step_time.png
ADDED
exp/lm_train_lm_transformer_char/images/train_time.png
ADDED
exp/lm_train_lm_transformer_char/perplexity_test/ppl
ADDED
@@ -0,0 +1 @@
|
|
|
|
|
1 |
+
1.7382500336543565
|
exp/lm_train_lm_transformer_char/valid.loss.ave_10best.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b4e174e16b3b6e784d5c428a881faae5816836843ac60b8df589022e59291105
|
3 |
+
size 202251148
|
meta.yaml
ADDED
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
espnet: 0.9.5
|
2 |
+
files:
|
3 |
+
asr_model_file: exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/valid.acc.ave_10best.pth
|
4 |
+
lm_file: exp/lm_train_lm_transformer_char/valid.loss.ave_10best.pth
|
5 |
+
python: "3.7.3 (default, Mar 27 2019, 22:11:17) \n[GCC 7.3.0]"
|
6 |
+
timestamp: 1605672767.746709
|
7 |
+
torch: 1.5.1
|
8 |
+
yaml_files:
|
9 |
+
asr_train_config: exp/asr_train_asr_transformer4_raw_char_batch_bins16000000_accum_grad1_sp/config.yaml
|
10 |
+
lm_train_config: exp/lm_train_lm_transformer_char/config.yaml
|