whisper-tiny-PolyAI-minds14

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: 0.5565
• Wer Ortho: 0.5120
• Wer: 0.3607

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-07
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 50
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
2.3523	71.43	500	2.3552	0.6089	0.4067
1.1267	142.86	1000	1.2038	0.5922	0.4132
0.5363	214.29	1500	0.7055	0.5694	0.4014
0.3846	285.71	2000	0.6171	0.5490	0.4008
0.304	357.14	2500	0.5816	0.5379	0.3890
0.2428	428.57	3000	0.5644	0.5182	0.3713
0.1922	500.0	3500	0.5570	0.5139	0.3666
0.1499	571.43	4000	0.5565	0.5120	0.3607

Framework versions

• Transformers 4.35.0
• Pytorch 2.0.0+cu117
• Datasets 2.14.6
• Tokenizers 0.14.1