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: 0.5893
- Der: 0.1935
- False Alarm: 0.0760
- Missed Detection: 0.0713
- Confusion: 0.0461
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 |
|---|---|---|---|---|---|---|---|
| 0.5956 | 1.0 | 336 | 0.6028 | 0.2004 | 0.0637 | 0.0824 | 0.0543 |
| 0.5696 | 2.0 | 672 | 0.6137 | 0.2001 | 0.0760 | 0.0741 | 0.0500 |
| 0.5334 | 3.0 | 1008 | 0.5886 | 0.1941 | 0.0787 | 0.0701 | 0.0453 |
| 0.5168 | 4.0 | 1344 | 0.5914 | 0.1933 | 0.0745 | 0.0723 | 0.0465 |
| 0.5174 | 5.0 | 1680 | 0.5893 | 0.1935 | 0.0760 | 0.0713 | 0.0461 |
Framework versions
- Transformers 4.40.0
- Pytorch 2.2.1+cu121
- Datasets 2.18.0
- Tokenizers 0.19.1