Makkoen/whisper-large-v3-cit-do01-wd0-lr3e-06-FULL4

./4607

This model is a fine-tuned version of openai/whisper-large-v3 on the 4607 FULL-2024-09-26 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4741
• Wer Ortho: 26.0994
• Wer: 18.8648

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: 3e-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
0.7796	0.7715	200	0.5425	30.3365	22.7179
0.5403	1.5429	400	0.4978	28.0609	20.7078
0.4638	2.3144	600	0.4795	26.7206	19.8063
0.3967	3.0858	800	0.4749	26.6020	19.1786
0.3536	3.8573	1000	0.4722	26.2948	19.0317
0.3289	4.6287	1200	0.4741	26.0994	18.8648

Framework versions

• Transformers 4.45.1
• Pytorch 1.13.1+cu117
• Datasets 3.0.1
• Tokenizers 0.20.0