xlm-roberta-base-finetuned-pquad-squad

This model is a version of xlm-roberta-base fine-tuned jointly on PQuAD and SQuAD_v2 datasets.

Results

Trained only for 3000/12000 steps (1/4th of an epoch) due to computational restrictions.

{'exact': 66.57766134314072, 'f1': 73.79323871739385, 'total': 19849, 'HasAns_exact': 64.51396460622327, 'HasAns_f1': 76.52620945245035, 'HasAns_total': 11923, 'NoAns_exact': 69.68205904617714, 'NoAns_f1': 69.68205904617714, 'NoAns_total': 7926, 'best_exact': 66.57766134314072, 'best_exact_thresh': 0.0, 'best_f1': 73.79323871739267, 'best_f1_thresh': 0.0}

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 1

Framework versions

• Transformers 4.26.1
• Pytorch 1.13.1+cu116
• Datasets 2.9.0
• Tokenizers 0.13.2