whisper-small-10DB-3r6

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

• Cer: 27.9030
• Loss: 0.3175
• Wer: 74.6757

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: 3e-06
• train_batch_size: 32
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• 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: 100
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Cer	Validation Loss	Wer
4.2071	1.2675	100	60.5454	1.0186	100.1235
2.3919	2.5350	200	36.2347	0.6153	85.4231
1.8048	3.8025	300	31.8374	0.4921	80.2965
1.3156	5.0637	400	31.7468	0.3902	78.5053
0.9425	6.3312	500	29.7243	0.3259	76.7758
0.8272	7.5987	600	28.2753	0.3199	75.9728
0.7388	8.8662	700	27.9030	0.3175	74.6757

Framework versions

• Transformers 5.0.0
• Pytorch 2.10.0+cu128
• Datasets 4.8.3
• Tokenizers 0.22.2