metadata
language:
- ru
license: apache-2.0
base_model: openai/whisper-small
tags:
- generated_from_trainer
datasets:
- bond005/sberdevices_golos_10h_crowd
metrics:
- wer
model-index:
- name: my_model - Val123val
results:
- task:
name: Automatic Speech Recognition
type: automatic-speech-recognition
dataset:
name: Sberdevices_golos_10h_crowd
type: bond005/sberdevices_golos_10h_crowd
args: 'split: test'
metrics:
- name: Wer
type: wer
value: 42.241139818232334
my_model - Val123val
This model is a fine-tuned version of openai/whisper-small on the Sberdevices_golos_10h_crowd dataset. It achieves the following results on the evaluation set:
- Loss: 0.1761
- Wer: 42.2411
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: 16
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- training_steps: 5000
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.1521 | 0.91 | 500 | 0.1824 | 29.3408 |
| 0.0824 | 1.82 | 1000 | 0.1702 | 27.5291 |
| 0.0304 | 2.73 | 1500 | 0.1726 | 45.1046 |
| 0.0114 | 3.64 | 2000 | 0.1704 | 40.1238 |
| 0.0039 | 4.55 | 2500 | 0.1692 | 32.1903 |
| 0.0013 | 5.45 | 3000 | 0.1704 | 34.0298 |
| 0.0029 | 6.36 | 3500 | 0.1712 | 39.8976 |
| 0.0007 | 7.27 | 4000 | 0.1738 | 39.4273 |
| 0.0006 | 8.18 | 4500 | 0.1755 | 41.0664 |
| 0.0005 | 9.09 | 5000 | 0.1761 | 42.2411 |
Framework versions
- Transformers 4.36.2
- Pytorch 2.1.2+cpu
- Datasets 2.16.0
- Tokenizers 0.15.0