JSWOOK/pyannote_finetuning
This model is a fine-tuned version of pyannote/speaker-diarization-3.1 on the diarizers-community/voxconverse dataset. It achieves the following results on the evaluation set:
- Loss: 0.1283
- Model Preparation Time: 0.0036
- Der: 0.0490
- False Alarm: 0.0309
- Missed Detection: 0.0091
- Confusion: 0.0090
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 | Model Preparation Time | Der | False Alarm | Missed Detection | Confusion |
|---|---|---|---|---|---|---|---|---|
| No log | 1.0 | 21 | 0.1258 | 0.0036 | 0.0485 | 0.0287 | 0.0105 | 0.0093 |
| 0.228 | 2.0 | 42 | 0.1327 | 0.0036 | 0.0509 | 0.0300 | 0.0098 | 0.0112 |
| 0.1873 | 3.0 | 63 | 0.1280 | 0.0036 | 0.0496 | 0.0307 | 0.0092 | 0.0097 |
| 0.166 | 4.0 | 84 | 0.1280 | 0.0036 | 0.0487 | 0.0307 | 0.0091 | 0.0090 |
| 0.152 | 5.0 | 105 | 0.1283 | 0.0036 | 0.0490 | 0.0309 | 0.0091 | 0.0090 |
Framework versions
- Transformers 4.44.2
- Pytorch 2.5.0+cu121
- Datasets 3.1.0
- Tokenizers 0.19.1