dominguesm/whisper-tiny-pt

Whisper Tiny PT

This model is a fine-tuned version of openai/whisper-tiny on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.6077
• Wer: 29.9844

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: 64
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• 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
0.4143	1.04	500	0.5325	32.7399
0.2693	3.03	1000	0.4718	29.4867
0.1724	5.01	1500	0.4758	28.7218
0.0849	7.0	2000	0.5070	29.2211
0.0659	8.04	2500	0.5223	29.3169
0.0539	10.03	3000	0.5402	30.1458
0.0376	12.02	3500	0.5755	29.9995
0.0217	14.0	4000	0.6067	29.6565
0.0168	15.04	4500	0.6082	29.8162
0.0205	17.03	5000	0.6077	29.9844

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.0+cu116
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2