GPT-Neo_DnD

This model is a fine-tuned version of PatrickTyBrown/GPT-Neo_DnD on the None dataset. It achieves the following results on the evaluation set:

• eval_loss: 3.2754
• eval_runtime: 88.78
• eval_samples_per_second: 6.51
• eval_steps_per_second: 3.255
• epoch: 2.9
• step: 16800

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: 2.5e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3

Framework versions

• Transformers 4.21.3
• Pytorch 1.12.1+cu113
• Datasets 2.4.0
• Tokenizers 0.12.1