finetune-medium-1210225-exp2

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: 0.3188
• Wer: 9.6805
• Cer: 4.2055

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-05
• 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
1.0774	1.0	58	1.2712	24.7822	15.4696
0.6718	2.0	116	0.7613	15.7793	8.7738
0.5073	3.0	174	0.5796	10.4066	4.3869
0.4199	4.0	232	0.4781	10.4550	4.3869
0.3605	5.0	290	0.4094	10.1646	4.3622
0.3237	6.0	348	0.3673	9.9710	4.4116
0.2981	7.0	406	0.3387	9.6805	4.2385
0.285	8.0	464	0.3245	9.7773	4.2550
0.28	9.0	522	0.3188	9.6805	4.2055

Framework versions

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