metadata
license: mit
base_model: pyannote/segmentation-3.0
tags:
- speaker-diarization
- speaker-segmentation
- generated_from_trainer
datasets:
- ArtFair/diarizers_dataset_70-15-15
model-index:
- name: fine_tuned_segmentation-3.0_1e-5_32
results: []
fine_tuned_segmentation-3.0_1e-5_32
This model is a fine-tuned version of pyannote/segmentation-3.0 on the ArtFair/diarizers_dataset_70-15-15 default dataset. It achieves the following results on the evaluation set:
- Loss: 0.4908
- Der: 0.3585
- False Alarm: 0.2041
- Missed Detection: 0.1214
- Confusion: 0.0330
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: 1e-05
- 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 |
|---|---|---|---|---|---|---|---|
| 0.621 | 1.0 | 929 | 0.5601 | 0.3774 | 0.2133 | 0.1279 | 0.0361 |
| 0.5472 | 2.0 | 1858 | 0.5149 | 0.3680 | 0.2084 | 0.1254 | 0.0342 |
| 0.5353 | 3.0 | 2787 | 0.4969 | 0.3615 | 0.2049 | 0.1234 | 0.0333 |
| 0.517 | 4.0 | 3716 | 0.4919 | 0.3590 | 0.2042 | 0.1216 | 0.0331 |
| 0.5217 | 5.0 | 4645 | 0.4908 | 0.3585 | 0.2041 | 0.1214 | 0.0330 |
Framework versions
- Transformers 4.37.2
- Pytorch 2.4.1+cu121
- Datasets 2.17.0
- Tokenizers 0.15.2