KannaBERT-xl

This model is a pretrained model on the Kannada IndicCorp dataset. It achieves the following results on the evaluation set:

• Loss: 3.8383

Model description

General pretrained TinyBERT model for Kannada (kn).

Training and evaluation data

Used 4.9GB IndicCorp data for pretraining. Model is not finetuned.

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 with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 1.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
4.0098	1.0	590626	3.8378

Framework versions

• Transformers 4.49.0
• Pytorch 2.5.1
• Datasets 2.19.1
• Tokenizers 0.21.0