Dave12121/Fsentiment

This model is a fine-tuned version of distilbert-base-uncased on the financial phrasebank sentences all agree dataset.

• Train Loss: 0.0517
• Validation Loss: 0.2117
• Train Accuracy: 0.9384
• Epoch: 2

It achieves an macro f1 score on the validation set financial phrasebank sentences 75% of: 0.92

The testing data is a data subset of the finantial phrasebank in which 66% annotators agreed on the label.

The reported macro f1 score on the test set is: 0.86

Model description

More information needed

Intended uses & limitations

Model should be treated with care. A simple review showed first signs of gender bias within the model.

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• optimizer: {'name': 'Adam', 'weight_decay': None, 'clipnorm': None, 'global_clipnorm': None, 'clipvalue': None, 'use_ema': False, 'ema_momentum': 0.99, 'ema_overwrite_frequency': None, 'jit_compile': False, 'is_legacy_optimizer': False, 'learning_rate': {'class_name': 'PolynomialDecay', 'config': {'initial_learning_rate': 2e-05, 'decay_steps': 705, 'end_learning_rate': 0.0, 'power': 1.0, 'cycle': False, 'name': None}}, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-08, 'amsgrad': False}
• training_precision: float32

Training results

Train Loss	Validation Loss	Train Accuracy	Epoch
0.5028	0.3128	0.8904	0
0.1137	0.2117	0.9375	1
0.0517	0.2117	0.9384	2

Framework versions

• Transformers 4.35.0
• TensorFlow 2.11.1
• Datasets 2.14.6
• Tokenizers 0.14.1