octava/optimized-sm-whisper-id

Optimized Whisper Small Id for Inspirasi

This model is a fine-tuned version of openai/whisper-small on the Extracted Youtube with CommonVoice11, Fleurs, OpenSLR, and MagicData dataset. It achieves the following results on the evaluation set:

• Loss: 0.3376
• Wer: 19.9620

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: 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: 500
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4122	0.1686	500	0.3999	24.8908
0.2737	0.3373	1000	0.3655	22.4691
0.2311	0.5059	1500	0.3491	21.5195
0.1947	0.6745	2000	0.3339	21.5100
0.169	0.8432	2500	0.3408	20.6363
0.0875	1.0118	3000	0.3429	21.2726
0.0877	1.1804	3500	0.3430	20.4748
0.0726	1.3491	4000	0.3396	20.2469
0.0741	1.5177	4500	0.3378	20.2754
0.0675	1.6863	5000	0.3376	19.9620

Framework versions

• Transformers 4.49.0
• Pytorch 2.2.0a0+81ea7a4
• Datasets 3.3.2
• Tokenizers 0.21.0