roberta-base-finetuned-squad

This model is a fine-tuned version of ozgurkk/roberta-base-finetuned-squad on the SQuAD 2.0 dataset.

Model description

roberta-base model fine-tuned on SQuAD 2.0

• Language model: roberta-base
• Language: English
• Downstream-task: Extractive QA
• Training data: SQuAD 2.0
• Eval data: SQuAD 2.0

Intended uses & limitations

Question-Answering

Training and evaluation data

https://rajpurkar.github.io/SQuAD-explorer/

Evaluation Results

• "exact": 77.89943569443275
• "f1": 81.32467973552446
• "HasAns_exact": 77.09176788124157
• "HasAns_f1": 83.95207869431226
• "NoAns_exact": 78.70479394449117
• "NoAns_f1": 78.70479394449117
• "best_exact": 79.07858165585783
• "best_f1": 82.15878087260025

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3
• mixed_precision_training: Native AMP

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1