speaker-segmentation-sula-hf_luganda_mental_health_dataset-v3
This model is a fine-tuned version of pyannote/speaker-diarization-3.0 on the Beijuka/luganda_mental_health_dataset_v3 dataset. It achieves the following results on the evaluation set:
- Loss: 0.4910
- Model Preparation Time: 0.0042
- Der: 0.1456
- False Alarm: 0.0473
- Missed Detection: 0.0677
- Confusion: 0.0306
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: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: cosine
- num_epochs: 15
Training results
| Training Loss | Epoch | Step | Validation Loss | Model Preparation Time | Der | False Alarm | Missed Detection | Confusion |
|---|---|---|---|---|---|---|---|---|
| 0.2372 | 1.0 | 222 | 0.3616 | 0.0042 | 0.1444 | 0.0326 | 0.0619 | 0.0498 |
| 0.2216 | 2.0 | 444 | 0.3436 | 0.0042 | 0.1400 | 0.0361 | 0.0555 | 0.0484 |
| 0.2197 | 3.0 | 666 | 0.3568 | 0.0042 | 0.1456 | 0.0413 | 0.0514 | 0.0529 |
| 0.2000 | 4.0 | 888 | 0.3516 | 0.0042 | 0.1383 | 0.0345 | 0.0595 | 0.0444 |
| 0.1946 | 5.0 | 1110 | 0.3829 | 0.0042 | 0.1405 | 0.0360 | 0.0564 | 0.0481 |
| 0.1894 | 6.0 | 1332 | 0.3932 | 0.0042 | 0.1386 | 0.0373 | 0.0559 | 0.0454 |
| 0.1953 | 7.0 | 1554 | 0.3691 | 0.0042 | 0.1369 | 0.0338 | 0.0598 | 0.0433 |
Framework versions
- Transformers 5.0.0
- Pytorch 2.10.0+cu128
- Datasets 4.0.0
- Tokenizers 0.22.2