metadata
library_name: transformers
language:
- en
license: apache-2.0
base_model: openai/whisper-large-v3
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: FULL6
results: []
FULL6
This model is a fine-tuned version of openai/whisper-large-v3 on the 9712 FULL-2024-10-24 dataset. It achieves the following results on the evaluation set:
- Loss: 0.3669
- Wer Ortho: 20.8879
- Wer: 15.0446
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: 5e-06
- train_batch_size: 4
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- gradient_accumulation_steps: 4
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 300
- training_steps: 1200
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer Ortho | Wer |
|---|---|---|---|---|---|
| 0.6532 | 0.3661 | 200 | 0.4502 | 25.2247 | 18.8841 |
| 0.5127 | 0.7323 | 400 | 0.4048 | 22.5978 | 16.4300 |
| 0.4408 | 1.0984 | 600 | 0.3845 | 21.8987 | 16.3138 |
| 0.3665 | 1.4645 | 800 | 0.3757 | 21.8443 | 16.0030 |
| 0.3589 | 1.8307 | 1000 | 0.3684 | 20.9727 | 15.1085 |
| 0.336 | 2.1968 | 1200 | 0.3669 | 20.8879 | 15.0446 |
Framework versions
- Transformers 4.45.1
- Pytorch 1.13.1+cu117
- Datasets 3.0.1
- Tokenizers 0.20.0