SPEAK-ASR/mBART-large-50-si-2

This model is a fine-tuned version of facebook/mbart-large-50 on the SPEAK-ASR/akura-sinhala-dyslexia-corrected dataset. It achieves the following results on the evaluation set:

• Loss: 0.0369

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: 5e-05
• train_batch_size: 32
• 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: 5

Training results

Training Loss	Epoch	Step	Validation Loss
0.0383	0.9921	500	0.0384
0.0608	1.9841	1000	0.0316
0.0205	2.9762	1500	0.0311
0.023	3.9683	2000	0.0335
0.0115	4.9603	2500	0.0369

Framework versions

• Transformers 4.57.3
• Pytorch 2.9.0+cu126
• Datasets 4.4.2
• Tokenizers 0.22.2