Downloads · 30 days
17
30% of all-time downloads
PA-RNN-T-Vietnamese-vlsp-100h-v2
MinhLe999/PA-RNN-T-Vietnamese-vlsp-100h-v2
PA-RNN-T-Vietnamese-vlsp-100h-v2 is a automatic speech recognition model from MinhLe999. Use it when you need speech turned into text. It is set up for k2. The card lists the license as apache-2.0.
Custom PA-RNN-T checkpoint trained on MinhLe999/forked-doofferb-vlsp2020-vinai-100h-v2.
All-time downloads
56
Public
Repo size
1.7 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.pt1.7 GB · 100%
At a glance
- Task
- Automatic Speech Recognition
- Library
- k2
- License
- apache-2.0
- Access
- Public
- Created
- Jun 19, 2026
- Updated
- Jun 19, 2026
- SHA
- 3968ca0b
Model card
From the Hugging Face model README
PA-RNN-T Vietnamese VLSP 100h v2
Custom PA-RNN-T checkpoint trained on MinhLe999/forked-doofferb-vlsp2020-vinai-100h-v2.
Checkpoint
epoch-3.pt
SSL backbone
nguyenvulebinh/wav2vec2-base-vi
Training setup
- batch size: 4
- epochs: 3
- SSL frozen until step 7250, then unfrozen
- RNN-T loss via k2
- decode setting used for reported eval:
max_symbols_per_frame=3,blank_penalty=2.0
Included files
epoch-3.pt: PyTorch checkpoint with model state dict and optimizer stateconfig.json: training/model configvocab.txt: phone vocabularycode/: architecture and decode code snapshot
- Task
- Automatic Speech Recognition
- Library
- k2
- License
- apache-2.0
- Created
- Jun 19, 2026
- Updated
- Jun 19, 2026