miosipof/asr2_medium_v0.7

Whisper Medium

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

• Loss: 0.3776
• Wer: 41.2607
• Cer: 30.2338
• Lr: 0.0000

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: 0.0003
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.3
• num_epochs: 12
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer	Lr
1.2755	1.0	251	1.0659	71.4900	45.8103	0.0001
0.8309	2.0	502	0.7274	63.6103	44.2868	0.0002
0.5827	3.0	753	0.5902	56.1605	38.9283	0.0002
0.3714	4.0	1004	0.5072	53.2951	38.7444	0.0003
0.1876	5.0	1255	0.4535	46.2751	32.8080	0.0003
0.1278	6.0	1506	0.3975	44.8424	33.0444	0.0002
0.0562	7.0	1757	0.3698	36.1032	26.3987	0.0002
0.0209	8.0	2008	0.4188	56.3037	46.4145	0.0001
0.0123	9.0	2259	0.3916	40.8309	29.8398	0.0001
0.005	10.0	2510	0.3819	41.5473	30.4965	0.0001
0.0031	11.0	2761	0.3779	41.8338	30.7591	0.0000
0.0018	12.0	3012	0.3776	41.2607	30.2338	0.0000

Framework versions

• PEFT 0.14.0
• Transformers 4.48.3
• Pytorch 2.2.0
• Datasets 3.2.0
• Tokenizers 0.21.0