Downloads · 30 days
0
ch-w2v-conformer
emiyasstar/ch-w2v-conformer
ch-w2v-conformer is a machine learning model from emiyasstar. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
The ch-w2v-conformer model uses following datasets to pretrain:
Access
Public
Updated Mar 31, 2022
Repo size
1.4 GB
Likes
2
Public
Repo makeup
Click a slice to open those files.
.pt689 MB · 100%
At a glance
- Access
- Public
- Created
- Mar 29, 2022
- Updated
- Mar 31, 2022
- SHA
- 580221ee
Model card
From the Hugging Face model README
The ch-w2v-conformer model uses following datasets to pretrain:
ISML datasets (6 languages,70k hours): internal dataset contains 40k hours Chinese, Cantonese, Tibetan, Inner Mongolian, Inner Kazakh, Uighur.
Babel datasets (17 languages, 2k hours): Assamese, Bengali, Cantonese, Cebuano, Georgian, Haitian, Kazakh, Kurmanji, Lao, Pashto, Swahili, Tagalog, Tamil, Tok, Turkish, Vietnamese, Zulu
After pretraining, we build ASR system based on CTC-Attention structure. In very low resource task, we find that if too many initialization network structures are constructed in the upper layer of pre-training conformer encoder, the migration performance of the pre-training model will be destroyed, so we only build a single-layer transformer decoder for joint training.
pretrained model link:
constrained-plus Task Performance
- Languages: Cantonese,mongolian,kazakh
- config: conf/train_conformer_large_10h.yaml
- Feature info: using mfcc feature, with dither 1.0, without cmvn
- Training info: lr 0.001, batch size 10, 4 gpus on V100, acc_grad 1, 80 epochs
- Decoding info: ctc_weight 0.5, average_num 35
dev set results trained only with 10 hours training set
w2v-Conformer
| decoding_method | Cantonese(CER) | mongolian(WER) |
|---|---|---|
| ctc_greedy_search | 31.46 | 53.64 |
| ctc_prefix_search | 31.47 | 53.50 |
| attention_rescoring | 31.45 | 52.96 |
Conformer （train from scartch）
| decoding_method | Cantonese(CER) | mongolian(WER) |
|---|---|---|
| ctc_greedy_search | 61.43 | 89.38 |
| ctc_prefix_search | 61.37 | 89.53 |
| attention_rescoring | 60.61 | 89.60 |
- Created
- Mar 29, 2022
- Updated
- Mar 31, 2022