breco/whisper-tiny-1000-audios

Whisper Tiny 1000 Audios - vfranchis

This model is a fine-tuned version of openai/whisper-tiny on the 1000 audios 1.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.5691
• Wer: 30.7692

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: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 25
• training_steps: 300
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.4694	0.4	25	1.0082	38.4615
0.2677	0.8	50	0.7480	46.1538
0.1034	1.2	75	0.6340	46.1538
0.0672	1.6	100	0.6319	46.1538
0.0547	2.0	125	0.5773	30.7692
0.0299	2.4	150	0.5612	30.7692
0.022	2.8	175	0.5784	30.7692
0.0218	3.2	200	0.5702	30.7692
0.0127	3.6	225	0.5721	30.7692
0.013	4.0	250	0.5554	30.7692
0.0084	4.4	275	0.5680	30.7692
0.0102	4.8	300	0.5691	30.7692

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1