aslon1213/whisper-small-uz-with-uzbekvoice

Whisper Small Uz - Aslon Khamidov -- with Uzbek Voice dataset

This model is a fine-tuned version of openai/whisper-small on the Common Voice 16.1 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3052
• Wer: 30.2049

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: 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: 500
• training_steps: 15000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4689	0.0210	1000	0.5616	48.2462
0.3234	0.0420	2000	0.4695	44.8210
0.3078	0.0630	3000	0.4184	38.8747
0.2845	0.0840	4000	0.3955	36.2861
0.2771	0.1050	5000	0.3720	35.5344
0.2459	0.1260	6000	0.3649	35.9415
0.2482	0.1470	7000	0.3499	34.3993
0.26	0.1680	8000	0.3389	32.9183
0.2128	0.1891	9000	0.3321	33.2493
0.2092	0.2101	10000	0.3215	31.4973
0.1942	0.2311	11000	0.3194	31.0465
0.1912	0.2521	12000	0.3184	31.2850
0.2199	0.2731	13000	0.3100	30.6395
0.1861	0.2941	14000	0.3059	30.8667
0.2344	0.3151	15000	0.3052	30.2049

Framework versions

• Transformers 4.41.2
• Pytorch 2.2.0
• Datasets 2.19.2
• Tokenizers 0.19.1