metadata
library_name: transformers
language:
- bn
license: mit
base_model: pyannote/segmentation-3.0
tags:
- audio
- speaker-diarization
- bengali
- pyannote
- speech
- generated_from_trainer
datasets:
- bengali-speaker-diarization
model-index:
- name: speaker-segmentation-bengali-optimized-conservative
results: []
speaker-segmentation-bengali-optimized-conservative
This model is a fine-tuned version of pyannote/segmentation-3.0 on the bengali-speaker-diarization dataset. It achieves the following results on the evaluation set:
- Loss: 0.4778
- Model Preparation Time: 0.0043
- Der: 0.1599
- False Alarm: 0.0403
- Missed Detection: 0.0162
- Confusion: 0.1034
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: 0.0005
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 10
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Model Preparation Time | Der | False Alarm | Missed Detection | Confusion |
|---|---|---|---|---|---|---|---|---|
| 0.4209 | 1.0 | 254 | 0.4600 | 0.0043 | 0.1574 | 0.0369 | 0.0185 | 0.1019 |
| 0.391 | 2.0 | 508 | 0.4628 | 0.0043 | 0.1586 | 0.0365 | 0.0215 | 0.1007 |
| 0.4302 | 3.0 | 762 | 0.4624 | 0.0043 | 0.1579 | 0.0388 | 0.0179 | 0.1012 |
| 0.3421 | 4.0 | 1016 | 0.4778 | 0.0043 | 0.1599 | 0.0403 | 0.0162 | 0.1034 |
Framework versions
- Transformers 4.48.3
- Pytorch 2.8.0+cu126
- Datasets 3.6.0
- Tokenizers 0.21.4