rristo/whisper-medium-et

Whisper Medium et

This model is a fine-tuned version of openai/whisper-medium on the ERR2020, Common Voice 11.0, FLEURS dataset. It achieves the following results on the evaluation set:

• Loss: 0.4288
• Wer: 29.7203

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-06
• train_batch_size: 32
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• 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
0.4018	0.1	500	0.5518	39.3951
0.2654	0.2	1000	0.4611	34.3929
0.2121	0.3	1500	0.4346	32.0582
0.1752	0.4	2000	0.4247	31.1926
0.1337	0.5	2500	0.4216	30.3364
0.1281	0.6	3000	0.4219	30.0745
0.1127	0.7	3500	0.4252	29.7388
0.1254	0.8	4000	0.4276	29.8928
0.1035	0.9	4500	0.4292	29.7634
0.1114	1.0	5000	0.4288	29.7203

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.0+cu117
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2