OwLim/whisper-java-SLR41-SLR35

Whisper Small Java

This model is a fine-tuned version of openai/whisper-small on the SLR Javanenese 41_35 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4200
• Wer: 29.2466

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: Use OptimizerNames.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: 100
• training_steps: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4922	0.16	100	0.6047	37.4678
0.435	0.32	200	0.5572	35.9424
0.5688	0.48	300	0.5090	33.5649
0.4779	0.64	400	0.4799	31.8390
0.4247	0.8	500	0.4540	30.8364
0.42	0.96	600	0.4368	30.2492
0.2276	1.12	700	0.4330	29.6333
0.2137	1.28	800	0.4264	29.5832
0.236	1.44	900	0.4215	29.2395
0.1971	1.6	1000	0.4200	29.2466

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1