sam-alavardo-1980/whisper-small-ru-v4

Whisper Small Ru - v4

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

• Loss: 0.2167
• Wer Ortho: 16.3879
• Wer: 11.9935

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: 50
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.1695	0.4921	500	0.2079	17.6749	13.3434
0.1548	0.9843	1000	0.1894	16.4416	12.2240
0.0704	1.4764	1500	0.1878	16.1107	12.0106
0.0722	1.9685	2000	0.1854	15.7395	11.7887
0.0328	2.4606	2500	0.1927	15.7822	11.6404
0.0344	2.9528	3000	0.1929	15.5746	11.6060
0.0147	3.4449	3500	0.2059	15.6992	11.5141
0.0148	3.9370	4000	0.2046	15.7859	11.5962
0.0067	4.4291	4500	0.2169	16.0374	11.6784
0.0078	4.9213	5000	0.2167	16.3879	11.9935

Framework versions

• Transformers 4.42.3
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1