Downloads · 30 days
0
pruned-transducer-stateless5-ukrainian-1-causal
darkproger/pruned-transducer-stateless5-ukrainian-1-causal
pruned-transducer-stateless5-ukrainian-1-causal is a automatic speech recognition model from darkproger. Use it when you need speech turned into text. The card lists the license as cc-by-nc-sa-4.0.
Online variant of prunedtransducerstateless5 for Ukrainian: https://github.com/proger/icefall/tree/uk
Access
Public
Updated Sep 21, 2022
Repo size
263 MB
Likes
1
Public
Repo makeup
Click a slice to open those files.
.pt263 MB · 96%
At a glance
- Task
- Automatic Speech Recognition
- License
- cc-by-nc-sa-4.0
- Access
- Public
- Created
- Sep 20, 2022
- Updated
- Sep 21, 2022
- SHA
- 3907b122
Model card
From the Hugging Face model README
Online variant of pruned_transducer_stateless5 for Ukrainian: https://github.com/proger/icefall/tree/uk
Decoding demo using Sherpa: https://twitter.com/darkproger/status/1570733844114046976
Trained on pseudolabels generated by darkproger/pruned-transducer-stateless5-ukrainian-1 on the noisy 1200 hours training set. Common Voice data was used only for validation.
./pruned_transducer_stateless5/train.py \
--world-size 2 \
--num-epochs 31 \
--start-epoch 1 \
--full-libri 1 \
--exp-dir pruned_transducer_stateless5/exp-uk-filtered2 \
--max-duration 600 \
--use-fp16 1 \
--num-encoder-layers 18 \
--dim-feedforward 1024 \
--nhead 4 \
--encoder-dim 256 \
--decoder-dim 512 \
--joiner-dim 512 \
--bpe-model uk/data/lang_bpe_250/bpe.model \
--causal-convolution True \
--dynamic-chunk-training True
- Task
- Automatic Speech Recognition
- License
- cc-by-nc-sa-4.0
- Languages
- uk
- Created
- Sep 20, 2022
- Updated
- Sep 21, 2022