This model is a fine-tuned version of distilbert-base-uncased on a joke/no-joke dataset in order to detect humor.
It achieves the following results on the evaluation set:
Loss: 0.1269
F1: 0.9537
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: 2e-06
train_batch_size: 16
eval_batch_size: 16
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
