ModernBERT-base-finetuned-squad

This model is a fine-tuned version of answerdotai/ModernBERT-base on rajpurkar/squad dataset.

Model description

ModernBERT-base finetuned on rajpurkar/squad dataset for Extractive-QA. It performs better than the bert-base-uncased model on the same task & dataset.

Test results

Performance of the bert-base-uncased model on the same test data was:

{'exact_match': 81.18259224219489, 'f1': 88.67381321905516}

Quite a significant improvement right !!

Training and evaluation data

rajpurkar/squad

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 8
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• 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
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Framework versions

• Transformers 5.0.0
• Pytorch 2.10.0+cu128
• Datasets 5.0.0
• Tokenizers 0.22.2