SPEAK-ASR/whisper-si-exp-3

SPEAK-ASR/speak-whisper-small-si-full-dataset

This model is a fine-tuned version of openai/whisper-small on the Whisper Small - Sinhala ASR Fine-Tuned dataset. It achieves the following results on the evaluation set:

• Loss: 0.1504

Model description

• WandB ID: azure-river-47
• Used full OpenSLR dataset
• Used LoRA adapters
• Previous name was SPEAK-ASR/speak-whisper-small-si-full-dataset

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: 0.0001
• train_batch_size: 64
• eval_batch_size: 32
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2372	0.4322	1000	0.2510	28.1934
0.2014	0.8643	2000	0.1958	22.7679
0.1824	1.2965	3000	0.1746	20.6799
0.1595	1.7286	4000	0.1651
0.1741	2.1608	5000	0.1560
0.1744	2.5929	6000	0.1504

Framework versions

• PEFT 0.18.1
• Transformers 5.0.0
• Pytorch 2.10.0+cu128
• Datasets 4.5.0
• Tokenizers 0.22.2