whisper-large-v3-turbo-half

This model is a fine-tuned version of openai/whisper-large-v3-turbo on the common_voice_16_1 dataset. It achieves the following results on the evaluation set:

• Loss: 0.7088
• Wer: 28.4350

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: 0.0002
• train_batch_size: 16
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 5000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0	0	8.8155	100.0
0.9071	0.1	500	1.5140	64.0547
0.7138	0.2	1000	1.1375	49.9023
0.5078	0.3	1500	1.0159	41.3067
0.4833	0.4	2000	0.9379	34.7081
0.4164	0.5	2500	0.8927	30.9746
0.517	0.6	3000	0.8473	31.0397
0.33	0.7	3500	0.7714	27.1326
0.364	0.8	4000	0.7508	25.6132
0.3728	0.9	4500	0.7091	24.4628
0.4321	1.0	5000	0.7088	28.4350

Framework versions

• Transformers 4.54.0
• Pytorch 2.8.0.dev20250319+cu128
• Datasets 3.6.0
• Tokenizers 0.21.2