DLL888/bert-base-uncased-squad

This model is a fine-tuned version of bert-base-uncased on SQuAD dataset. It achieves the following results on the evaluation set:

• Exact Match: 80.21759697256385
• F1: 87.77849998885436

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training Machine

Trained in Google Colab Pro with the following specs:

• A100-SXM4-40GB
• NVIDIA-SMI 460.32.03
• Driver Version: 460.32.03
• CUDA Version: 11.2

Training took about 26 minutes for two epochs.

Training hyperparameters

The following hyperparameters were used during training:

• optimizer: {'name': 'Adam', 'learning_rate': {'class_name': 'WarmUp', 'config': {'initial_learning_rate': 2e-05, 'decay_schedule_fn': {'class_name': 'PolynomialDecay', 'config': {'initial_learning_rate': 2e-05, 'decay_steps': 10564, 'end_learning_rate': 0.0, 'power': 1.0, 'cycle': False, 'name': None}, 'passive_serialization': True}, 'warmup_steps': 500, 'power': 1.0, 'name': None}}, 'decay': 0.0, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-08, 'amsgrad': False}
• training_precision: mixed_float16

Training results

Train Loss	Train End Logits Accuracy	Train Start Logits Accuracy	Validation Loss	Validation End Logits Accuracy	Validation Start Logits Accuracy	Epoch
1.4348	0.6368	0.5974	1.0155	0.7193	0.6825	0
0.8072	0.7735	0.7320	0.9990	0.7302	0.6983	1

Framework versions

• Transformers 4.24.0
• TensorFlow 2.9.2
• Datasets 2.7.1
• Tokenizers 0.13.2