dropout_gpt2

This model is a fine-tuned version of gpt2 on a English Wikipedia dataset.

Model description

Dropout debiased gpt2 using the hyperparameters specified in Measuring and Reducing Gendered Correlations in Pre-trained Models (Webster et al. 2021) and the code used in An Empirical Survey of the Effectiveness of Debiasing Techniques for Pre-trained Language Models (Meade et al. 2022).

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 4
• eval_batch_size: 8
• seed: 0
• gradient_accumulation_steps: 32
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• training_steps: 4000

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0