whisper-tiny-v2-ta_50

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

• Loss: 0.2683
• Wer: 31.0896

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: 0.00015
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 3000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4405	0.11	300	0.4624	105.3753
0.3629	0.21	600	0.4473	40.7264
0.3163	0.32	900	0.3771	35.1090
0.279	0.42	1200	0.3547	34.9153
0.2594	0.53	1500	0.3311	31.6223
0.2284	0.63	1800	0.3131	29.3947
0.2294	0.74	2100	0.2863	27.5061
0.2022	0.84	2400	0.2735	27.2155
0.185	0.95	2700	0.2683	31.0896

Framework versions

• Transformers 4.28.0.dev0
• Pytorch 1.12.1
• Datasets 2.11.0
• Tokenizers 0.13.3