bert-small-pretrained-finetuned-squad

This model is a fine-tuned version of anas-awadalla/bert-small-pretrained-on-squad on the squad dataset.

• "exact_match": 72.20435193945127
• "f1": 81.31832229156294

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: 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: 3.0

Training results

Framework versions

• Transformers 4.16.0.dev0
• Pytorch 1.10.1+cu102
• Datasets 1.17.0
• Tokenizers 0.10.3