speaker-segmentation-sula-hf_luganda_mental_health_dataset-v2
This model is a fine-tuned version of pyannote/speaker-diarization-3.0 on the Beijuka/luganda_mental_health_dataset_v2 dataset. It achieves the following results on the evaluation set:
- Loss: 0.5084
- Model Preparation Time: 0.0039
- Der: 0.1486
- False Alarm: 0.0406
- Missed Detection: 0.0737
- Confusion: 0.0343
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 |
|---|---|---|---|---|---|---|---|---|
| 0.3112 | 1.0 | 222 | 0.3944 | 0.0039 | 0.1642 | 0.0294 | 0.0735 | 0.0614 |
| 0.2713 | 2.0 | 444 | 0.3690 | 0.0039 | 0.1528 | 0.0362 | 0.0589 | 0.0577 |
| 0.2640 | 3.0 | 666 | 0.3789 | 0.0039 | 0.1539 | 0.0374 | 0.0577 | 0.0587 |
| 0.2409 | 4.0 | 888 | 0.3650 | 0.0039 | 0.1484 | 0.0291 | 0.0639 | 0.0554 |
| 0.2302 | 5.0 | 1110 | 0.3527 | 0.0039 | 0.1421 | 0.0311 | 0.0628 | 0.0482 |
| 0.2247 | 6.0 | 1332 | 0.3574 | 0.0039 | 0.1415 | 0.0306 | 0.0645 | 0.0465 |
| 0.2219 | 7.0 | 1554 | 0.3620 | 0.0039 | 0.1411 | 0.0340 | 0.0595 | 0.0476 |
| 0.2145 | 8.0 | 1776 | 0.3581 | 0.0039 | 0.1390 | 0.0350 | 0.0579 | 0.0461 |
Framework versions
- Transformers 5.0.0
- Pytorch 2.10.0+cu128
- Datasets 4.0.0
- Tokenizers 0.22.2