OwLim/whisper-java-SLR41

Whisper Small Java

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

• Loss: 0.2729
• Wer: 26.7732

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
0.8934	0.3003	100	0.8003	52.4086
0.4852	0.6006	200	0.5305	39.4578
0.4111	0.9009	300	0.4214	32.8250
0.2101	1.2012	400	0.3655	30.4527
0.1803	1.5015	500	0.3257	29.1939
0.1845	1.8018	600	0.3072	27.4752
0.0899	2.1021	700	0.2997	26.4585
0.0816	2.4024	800	0.2850	26.3617
0.078	2.7027	900	0.2755	26.7248
0.0769	3.0030	1000	0.2729	26.7732

Framework versions

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