misterkissi/whisper-small-vai

wisper-small-vai

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

• Loss: 0.4601
• Wer: 0.2969

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: 128
• eval_batch_size: 16
• seed: 42
• 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
0.0656	7.1429	500	0.3198	0.3513
0.0136	14.2857	1000	0.3727	0.4512
0.0033	21.4286	1500	0.3980	0.3258
0.0019	28.5714	2000	0.4147	0.3070
0.0004	35.7143	2500	0.4326	0.2999
0.0001	42.8571	3000	0.4428	0.3053
0.0001	50.0	3500	0.4502	0.3037
0.0001	57.1429	4000	0.4554	0.3049
0.0001	64.2857	4500	0.4587	0.2975
0.0001	71.4286	5000	0.4601	0.2969

Framework versions

• Transformers 4.52.4
• Pytorch 2.6.0+cu124
• Datasets 2.14.4
• Tokenizers 0.21.2