whisper-small-ru-v16tsb

This model is a fine-tuned version of constantinedivis/whisper-small-ru-v15tsb on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0567
• Wer: 5.2954

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
0.0197	0.1916	200	0.0884	8.2321
0.0089	0.3831	400	0.0991	8.8572
0.027	0.5747	600	0.0998	8.7628
0.0534	0.7663	800	0.0842	7.6070
0.0875	0.9579	1000	0.0712	7.1353
0.0225	1.1494	1200	0.0668	6.8286
0.0224	1.3410	1400	0.0634	6.1918
0.0197	1.5326	1600	0.0612	5.8969
0.0211	1.7241	1800	0.0579	5.5077
0.0191	1.9157	2000	0.0567	5.2954

Framework versions

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