finetune-medium-1210225-exp1

This model is a fine-tuned version of openai/whisper-medium on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1607
• Wer: 4.8403
• Cer: 2.4738

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: 0.0001
• train_batch_size: 64
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 128
• 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: 115
• training_steps: 575

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.3327	1.0	58	0.2997	10.1162	5.1950
0.1398	2.0	116	0.1622	6.1471	2.7872
0.0657	3.0	174	0.1670	8.6641	4.9641
0.0275	4.0	232	0.1507	5.9535	3.1253
0.0107	5.0	290	0.1564	5.5663	2.8037
0.0033	6.0	348	0.1573	4.6467	2.2759
0.0018	7.0	406	0.1594	4.8403	2.4738
0.0017	8.0	464	0.1607	4.8403	2.4738
0.0016	9.0	522	0.1607	4.8403	2.4738

Framework versions

• Transformers 4.51.1
• Pytorch 2.8.0+cu128
• Datasets 4.0.0
• Tokenizers 0.21.4