Sleepyp00/whisper-small-sv-test

whisper-small-test

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

• Loss: 0.4846
• Wer: 30.9001

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: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 500
• training_steps: 6000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.5133	0.64	500	0.5562	38.3826
0.3764	1.28	1000	0.4754	34.1921
0.3501	1.92	1500	0.4554	32.8509
0.222	2.55	2000	0.4399	32.3844
0.1283	3.19	2500	0.4440	32.0081
0.1575	3.83	3000	0.4295	31.8225
0.1205	4.47	3500	0.4410	31.0459
0.0992	5.11	4000	0.4505	31.1042
0.0941	5.75	4500	0.4551	30.8498
0.0627	6.39	5000	0.4696	31.5999
0.0652	7.02	5500	0.4761	31.8013
0.0573	7.66	6000	0.4846	30.9001

Framework versions

• Transformers 4.36.0.dev0
• Pytorch 2.1.1+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0