Downloads · 30 days
18
5% of all-time downloads
whisper-small-bn-all-600
Rakib/whisper-small-bn-all-600
whisper-small-bn-all-600 is a automatic speech recognition model from Rakib. Use it when you need speech turned into text. It is set up for transformers. The card lists the license as apache-2.0.
should probably proofread and complete it, then remove this comment. --
All-time downloads
350
Public
Repo size
26.1 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.bin967 MB · 100%
At a glance
- Task
- Automatic Speech Recognition
- Library
- transformers
- License
- apache-2.0
- Model type
- whisper
- Access
- Public
- Created
- Mar 11, 2023
- Updated
- Mar 14, 2023
- SHA
- dc977b55
Spaces
Model card
From the Hugging Face model README
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->
Whisper Small - Mohammed Rakib
This model is a fine-tuned version of openai/whisper-small on the common-voice-11, the google-fleurs and the openslr53 datasets. It achieves the following results on the evaluation set:
- Loss: 0.0605
- Cer: 5.6737
- Wer: 10.5971
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 1e-05
- train_batch_size: 4
- eval_batch_size: 32
- seed: 42
- gradient_accumulation_steps: 128
- total_train_batch_size: 512
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 888
- training_steps: 3000
Training results
| Training Loss | Epoch | Step | Validation Loss | Cer | Wer |
|---|---|---|---|---|---|
| 1.4902 | 0.11 | 100 | 1.2636 | 1208.7378 | 889.9406 |
| 0.5703 | 0.23 | 200 | 0.4612 | 121.7766 | 108.6401 |
| 0.3679 | 0.34 | 300 | 0.3046 | 17.7385 | 35.2744 |
| 0.2301 | 0.45 | 400 | 0.2059 | 14.8853 | 29.7671 |
| 0.191 | 0.56 | 500 | 0.1693 | 12.7561 | 25.4528 |
| 0.1605 | 0.68 | 600 | 0.1462 | 11.5880 | 22.7845 |
| 0.146 | 0.79 | 700 | 0.1300 | 10.4321 | 20.5541 |
| 0.1296 | 0.9 | 800 | 0.1156 | 9.8572 | 19.2143 |
| 0.1212 | 1.01 | 900 | 0.1055 | 8.9462 | 17.4004 |
| 0.1072 | 1.13 | 1000 | 0.0978 | 8.2675 | 15.9234 |
| 0.1013 | 1.24 | 1100 | 0.0912 | 7.7918 | 15.0605 |
| 0.0952 | 1.35 | 1200 | 0.0854 | 7.5497 | 14.3207 |
| 0.0915 | 1.47 | 1300 | 0.0809 | 6.9833 | 13.3163 |
| 0.0843 | 1.58 | 1400 | 0.0780 | 6.6422 | 12.7179 |
| 0.0819 | 1.69 | 1500 | 0.0744 | 6.7287 | 12.6589 |
| 0.0798 | 1.8 | 1600 | 0.0718 | 6.4962 | 12.3022 |
| 0.0774 | 1.92 | 1700 | 0.0694 | 6.2198 | 11.8414 |
| 0.0695 | 2.03 | 1800 | 0.0680 | 6.1346 | 11.5683 |
| 0.0686 | 2.14 | 1900 | 0.0662 | 5.9758 | 11.2485 |
| 0.0681 | 2.25 | 2000 | 0.0647 | 6.0599 | 11.2842 |
| 0.0661 | 2.37 | 2100 | 0.0639 | 5.9500 | 11.1356 |
| 0.0653 | 2.48 | 2200 | 0.0631 | 5.8114 | 10.8952 |
| 0.0636 | 2.59 | 2300 | 0.0622 | 5.8502 | 10.9385 |
| 0.0641 | 2.71 | 2400 | 0.0615 | 5.7382 | 10.7015 |
| 0.0633 | 2.82 | 2500 | 0.0612 | 5.7038 | 10.6455 |
| 0.0626 | 2.93 | 2600 | 0.0608 | 5.8058 | 10.7597 |
| 0.06 | 3.04 | 2700 | 0.0605 | 5.7328 | 10.6374 |
| 0.0584 | 3.16 | 2800 | 0.0605 | 5.6737 | 10.5971 |
| 0.0585 | 3.27 | 2900 | 0.0604 | 5.6877 | 10.6098 |
| 0.0598 | 3.38 | 3000 | 0.0603 | 5.7075 | 10.6043 |
Framework versions
- Transformers 4.27.0.dev0
- Pytorch 1.13.1+cu117
- Datasets 2.10.1
- Tokenizers 0.13.2
- Task
- Automatic Speech Recognition
- Library
- transformers
- Type
- whisper
- License
- apache-2.0
- Languages
- bn
- Created
- Mar 11, 2023
- Updated
- Mar 14, 2023