wh_med_cv_trial

This model is a fine-tuned version of openai/whisper-medium on the common_voice_22_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.1752
• Global Wer: 26.2986

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: 5e-06
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Use adamw_torch_fused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 50
• training_steps: 500

Training results

Training Loss	Epoch	Step	Validation Loss	Global Wer
3.1441	0.1642	50	0.6716	49.9325
0.9852	0.3284	100	0.2411	32.9473
0.8377	0.4926	150	0.2149	31.2274
0.7306	0.6568	200	0.1980	28.9720
0.6913	0.8210	250	0.1904	28.1796
0.7020	0.9852	300	0.1796	27.5134
0.3592	1.1478	350	0.1803	26.3596
0.4114	1.3120	400	0.1799	27.1128
0.3974	1.4762	450	0.1736	26.6818
0.3723	1.6404	500	0.1752	26.2986

Framework versions

• Transformers 5.0.0.dev0
• Pytorch 2.9.0+cu126
• Datasets 3.6.0
• Tokenizers 0.22.2