whisper-small-ru-v8lb

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

• Loss: 0.1316
• Wer: 10.9565

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: 32
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 400
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2765	0.1934	200	0.3496	18.6461
0.2192	0.3868	400	0.2859	17.0775
0.2197	0.5803	600	0.2025	14.9428
0.1827	0.7737	800	0.1833	13.9875
0.1563	0.9671	1000	0.1636	13.1619
0.1185	1.1605	1200	0.1614	12.5722
0.1102	1.3540	1400	0.1402	11.6995
0.109	1.5474	1600	0.1330	11.3929
0.1061	1.7408	1800	0.1310	11.0862
0.1072	1.9342	2000	0.1316	10.9565

Framework versions

• Transformers 4.49.0
• Pytorch 2.1.0+cu118
• Datasets 3.3.2
• Tokenizers 0.21.0