whisper-small-fine_tuned-ru

This model is a fine-tuned version of openai/whisper-small on the Mozilla common_voice_13_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2290
• Wer: 17.6336

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: 5e-07
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 250
• training_steps: 5000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0286	0.22	500	0.2225	18.0199
0.0287	0.44	1000	0.2235	18.0455
0.0334	0.66	1500	0.2243	18.1956
0.0373	0.88	2000	0.2239	17.9803
0.0261	1.1	2500	0.2268	17.7674
0.0252	1.32	3000	0.2277	17.8221
0.0265	1.54	3500	0.2290	17.6336
0.0271	1.76	4000	0.2293	17.8279
0.0252	1.97	4500	0.2293	17.8744
0.023	2.19	5000	0.2298	17.8930

Framework versions

• Transformers 4.30.2
• Pytorch 2.0.1+cu117
• Datasets 2.13.1
• Tokenizers 0.13.3