lorenzoncina/whisper-small-ru

Whisper Small Russian

This model is a fine-tuned version of openai/whisper-small on the mozilla-foundation/common_voice_11_0 ru dataset. It achieves the following results on the evaluation set:

• Loss: 0.3060
• Wer: 12.2375

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: 32
• eval_batch_size: 16
• 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: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0731	1.04	1000	0.2183	13.0589
0.0194	3.02	2000	0.2390	12.8027
0.0067	4.06	3000	0.2524	12.5832
0.0025	6.04	4000	0.2725	12.3245
0.0017	8.02	5000	0.2854	12.7046
0.0009	9.06	6000	0.2915	12.5072
0.0005	11.04	7000	0.3006	12.2473
0.0004	13.02	8000	0.3060	12.2375
0.0003	14.06	9000	0.3129	12.2963
0.0003	16.04	10000	0.3157	12.2988

Framework versions

• Transformers 4.28.0.dev0
• Pytorch 2.0.0+cu117
• Datasets 2.11.1.dev0
• Tokenizers 0.13.2