t4ai/distilbert-finetuned-t3-qa

This model is a fine-tuned version of distilbert-base-cased SQUaD Dataset (https://www.kaggle.com/datasets/stanfordu/stanford-question-answering-dataset). It achieves the following results on the evaluation set:

• Train Loss: 0.7523
• Epoch: 2

Model description

distilBERT base model fine-tuned for extractive Q&A. This model achieved an F1 score of 76.28 and EM score of 61.51 against SQUaD test set.

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• optimizer: {'inner_optimizer': {'module': 'transformers.optimization_tf', 'class_name': 'AdamWeightDecay', 'config': {'name': 'AdamWeightDecay', 'learning_rate': {'module': 'keras.optimizers.schedules', 'class_name': 'PolynomialDecay', 'config': {'initial_learning_rate': 2e-05, 'decay_steps': 16755, 'end_learning_rate': 0.0, 'power': 1.0, 'cycle': False, 'name': None}, 'registered_name': None}, 'decay': 0.0, 'beta_1': 0.8999999761581421, 'beta_2': 0.9990000128746033, 'epsilon': 1e-08, 'amsgrad': False, 'weight_decay_rate': 0.01}, 'registered_name': 'AdamWeightDecay'}, 'dynamic': True, 'initial_scale': 32768.0, 'dynamic_growth_steps': 2000}
• training_precision: mixed_float16

Training results

Train Loss	Epoch
1.5389	0
0.9645	1
0.7523	2

Framework versions

• Transformers 4.34.0
• TensorFlow 2.13.0
• Datasets 2.14.5
• Tokenizers 0.14.1