speaker-segmentation-fine-tuned-merged-backup-uganda-callhome-eng
This model is a fine-tuned version of pyannote/segmentation-3.0 on the KMayanja/backup_and_callhome default dataset. It achieves the following results on the evaluation set:
- Loss: 0.3085
- Der: 0.1123
- False Alarm: 0.0384
- Missed Detection: 0.0378
- Confusion: 0.0361
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: 5.0
Training results
| Training Loss | Epoch | Step | Validation Loss | Der | False Alarm | Missed Detection | Confusion |
|---|---|---|---|---|---|---|---|
| 0.3367 | 1.0 | 605 | 0.3336 | 0.1237 | 0.0481 | 0.0369 | 0.0387 |
| 0.3267 | 2.0 | 1210 | 0.3148 | 0.1155 | 0.0416 | 0.0353 | 0.0386 |
| 0.302 | 3.0 | 1815 | 0.3119 | 0.1124 | 0.0394 | 0.0379 | 0.0351 |
| 0.29 | 4.0 | 2420 | 0.3088 | 0.1125 | 0.0393 | 0.0370 | 0.0361 |
| 0.288 | 5.0 | 3025 | 0.3085 | 0.1123 | 0.0384 | 0.0378 | 0.0361 |
Framework versions
- Transformers 4.42.4
- Pytorch 2.3.1+cu121
- Datasets 2.20.0
- Tokenizers 0.19.1
- Downloads last month
- 4
Model tree for KMayanja/speaker-segmentation-fine-tuned-merged-backup-uganda-callhome-eng
Base model
pyannote/segmentation-3.0