whisper-tiny-en-US

This model is a fine-tuned version of openai/whisper-tiny on the PolyAI/minds14 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4990
• Wer Ortho: 0.2965
• Wer: 0.2960

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-07
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 100
• training_steps: 500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.0244	0.8969	50	0.5282	0.3012	0.3005
0.0178	1.7937	100	0.5213	0.2985	0.2986
0.0171	2.6906	150	0.5147	0.2979	0.2967
0.0121	3.5874	200	0.5092	0.2925	0.2915
0.0071	4.4843	250	0.5057	0.3072	0.3069
0.0073	5.3812	300	0.5034	0.2945	0.2941
0.003	6.2780	350	0.5014	0.2945	0.2934
0.0036	7.1749	400	0.5003	0.2972	0.2967
0.0034	8.0717	450	0.4997	0.2965	0.2960
0.0034	8.9686	500	0.4990	0.2965	0.2960

Framework versions

• Transformers 4.44.0
• Pytorch 2.4.0
• Datasets 2.21.0
• Tokenizers 0.19.1