speaker-segmentation-fine-tuned-hindi
This model is a fine-tuned version of pyannote/speaker-diarization-3.1 on the synthetic hindi dataset (https://huggingface.co/datasets/Samyak29/synthetic-speaker-diarization-dataset-hindi-large). It achieves the following results on the evaluation set:
- Loss: 0.4442
- Der: 0.1448
- False Alarm: 0.0243
- Missed Detection: 0.0280
- Confusion: 0.0925
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 |
|---|---|---|---|---|---|---|---|
| 0.477 | 1.0 | 194 | 0.4877 | 0.1651 | 0.0259 | 0.0320 | 0.1072 |
| 0.3908 | 2.0 | 388 | 0.4562 | 0.1526 | 0.0231 | 0.0315 | 0.0980 |
| 0.3708 | 3.0 | 582 | 0.4356 | 0.1451 | 0.0242 | 0.0284 | 0.0924 |
| 0.3567 | 4.0 | 776 | 0.4461 | 0.1441 | 0.0244 | 0.0280 | 0.0917 |
| 0.3447 | 5.0 | 970 | 0.4442 | 0.1448 | 0.0243 | 0.0280 | 0.0925 |
Framework versions
- Transformers 4.42.4
- Pytorch 2.3.1+cu121
- Datasets 2.21.0
- Tokenizers 0.19.1
- Downloads last month
- 19
Model tree for Samyak29/speaker-segmentation-fine-tuned-hindi
Base model
pyannote/speaker-diarization-3.1