FULL6
This model is a fine-tuned version of openai/whisper-large-v3 on the 970 FULL-2024-11-22 dataset. It achieves the following results on the evaluation set:
- Loss: 0.3523
- Wer Ortho: 19.9273
- Wer: 14.7053
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-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: 200
- training_steps: 500
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer Ortho | Wer |
|---|---|---|---|---|---|
| 0.7795 | 1.8265 | 100 | 0.4587 | 25.2085 | 19.0183 |
| 0.4482 | 3.6530 | 200 | 0.3704 | 21.8730 | 16.0813 |
| 0.3465 | 5.4795 | 300 | 0.3547 | 20.2694 | 14.7053 |
| 0.2946 | 7.3059 | 400 | 0.3511 | 19.7563 | 14.4383 |
| 0.2691 | 9.1324 | 500 | 0.3523 | 19.9273 | 14.7053 |
Framework versions
- Transformers 4.44.0
- Pytorch 1.13.1+cu117
- Datasets 2.21.0
- Tokenizers 0.19.1
- Downloads last month
- 6
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for Makkoen/whisper-large-v3-cit-do015-wd0-lr1e-06-steps500-PHI
Base model
openai/whisper-large-v3