speaker-segmentation-fine-tuned-callhome-eng-4
This model is a fine-tuned version of pyannote/segmentation-3.0 on the diarizers-community/callhome eng dataset. It achieves the following results on the evaluation set:
- Loss: 0.4660
- Der: 0.1806
- False Alarm: 0.0592
- Missed Detection: 0.0714
- Confusion: 0.0501
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: 10.0
Training results
| Training Loss | Epoch | Step | Validation Loss | Der | False Alarm | Missed Detection | Confusion |
|---|---|---|---|---|---|---|---|
| 0.4104 | 1.0 | 362 | 0.4742 | 0.1920 | 0.0615 | 0.0742 | 0.0562 |
| 0.4041 | 2.0 | 724 | 0.4738 | 0.1868 | 0.0620 | 0.0714 | 0.0534 |
| 0.3741 | 3.0 | 1086 | 0.4695 | 0.1851 | 0.0625 | 0.0705 | 0.0521 |
| 0.3612 | 4.0 | 1448 | 0.4689 | 0.1814 | 0.0588 | 0.0707 | 0.0519 |
| 0.3404 | 5.0 | 1810 | 0.4649 | 0.1792 | 0.0580 | 0.0720 | 0.0492 |
| 0.3462 | 6.0 | 2172 | 0.4620 | 0.1812 | 0.0615 | 0.0692 | 0.0505 |
| 0.3296 | 7.0 | 2534 | 0.4631 | 0.1800 | 0.0582 | 0.0713 | 0.0506 |
| 0.3261 | 8.0 | 2896 | 0.4731 | 0.1820 | 0.0586 | 0.0733 | 0.0501 |
| 0.3251 | 9.0 | 3258 | 0.4663 | 0.1811 | 0.0579 | 0.0727 | 0.0506 |
| 0.3154 | 10.0 | 3620 | 0.4660 | 0.1806 | 0.0592 | 0.0714 | 0.0501 |
Framework versions
- Transformers 4.40.1
- Pytorch 2.2.0+cu121
- Datasets 2.17.0
- Tokenizers 0.19.1