speaker-segmentation-fine-tuned-callhome-jpn
This model is a fine-tuned version of openai/whisper-small on the diarizers-community/callhome dataset. It achieves the following results on the evaluation set:
- Loss: 0.7479
- Der: 0.2241
- False Alarm: 0.0478
- Missed Detection: 0.1332
- Confusion: 0.0431
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: 0.001
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- num_epochs: 5
Training results
| Training Loss | Epoch | Step | Validation Loss | Der | False Alarm | Missed Detection | Confusion |
|---|---|---|---|---|---|---|---|
| 0.5757 | 1.0 | 328 | 0.7460 | 0.2299 | 0.0502 | 0.1343 | 0.0454 |
| 0.5219 | 2.0 | 656 | 0.7482 | 0.2251 | 0.0486 | 0.1340 | 0.0425 |
| 0.5067 | 3.0 | 984 | 0.7539 | 0.2259 | 0.0454 | 0.1369 | 0.0435 |
| 0.4923 | 4.0 | 1312 | 0.7453 | 0.2246 | 0.0490 | 0.1320 | 0.0436 |
| 0.5157 | 5.0 | 1640 | 0.7479 | 0.2241 | 0.0478 | 0.1332 | 0.0431 |
Framework versions
- Transformers 4.40.0
- Pytorch 2.2.1+cu121
- Datasets 2.19.0
- Tokenizers 0.19.1
- Downloads last month
- 5
Model tree for heavenode/speaker-segmentation-fine-tuned-callhome-jpn
Base model
openai/whisper-small