nocturneFlow/whisper-medium-ft-issai

Whisper Medium KK - Kazakh - Fleurs - Common Voice

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

• Loss: 0.0778
• Wer: 10.2803

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.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_steps: 500
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1487	0.2173	1000	0.1408	17.6987
0.1137	0.4347	2000	0.1059	13.7526
0.092	0.6520	3000	0.0924	12.0276
0.0814	0.8693	4000	0.0814	10.6566
0.047	1.0865	5000	0.0778	10.2803

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu118
• Datasets 3.6.0
• Tokenizers 0.21.0