metadata
language:
- jpn
license: apache-2.0
base_model: openai/whisper-small
tags:
- speaker-diarization
- speaker-segmentation
- generated_from_trainer
datasets:
- diarizers-community/callhome
model-index:
- name: speaker-segmentation-fine-tuned-callhome-jpn
results: []
speaker-segmentation-fine-tuned-callhome-jpn
This model is a fine-tuned version of openai/whisper-small on the diarizers-community/callhome dataset. It achieves the following results on the evaluation set:
- Loss: 0.7483
- Der: 0.2246
- False Alarm: 0.0483
- Missed Detection: 0.1328
- Confusion: 0.0435
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 |
|---|---|---|---|---|---|---|---|
| 0.5757 | 1.0 | 328 | 0.7538 | 0.2333 | 0.0470 | 0.1387 | 0.0477 |
| 0.5269 | 2.0 | 656 | 0.7608 | 0.2275 | 0.0474 | 0.1361 | 0.0439 |
| 0.5004 | 3.0 | 984 | 0.7516 | 0.2267 | 0.0463 | 0.1369 | 0.0435 |
| 0.4892 | 4.0 | 1312 | 0.7440 | 0.2241 | 0.0498 | 0.1317 | 0.0426 |
| 0.5122 | 5.0 | 1640 | 0.7483 | 0.2246 | 0.0483 | 0.1328 | 0.0435 |
Framework versions
- Transformers 4.40.0
- Pytorch 2.2.1+cu121
- Datasets 2.19.0
- Tokenizers 0.19.1