nocturneFlow/whisper-medium-ft

Whisper Medium KK - Kazakh - Fleurs - Common Voice

This model is a fine-tuned version of openai/whisper-medium on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3910
• Wer: 21.2101

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: 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
0.0045	7.5725	1000	0.3121	23.2826
0.0003	15.1507	2000	0.3523	21.3939
0.0001	22.7232	3000	0.3738	21.3661
0.0001	30.3013	4000	0.3863	21.3772
0.0001	37.8738	5000	0.3910	21.2101

Framework versions

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