GPT-Neo_DnD_Control
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: 2.6518
- eval_runtime: 141.422
- eval_samples_per_second: 6.527
- eval_steps_per_second: 3.267
- epoch: 3.9
- step: 36000
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: 1.5e-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: 5
Framework versions
- Transformers 4.21.3
- Pytorch 1.12.1+cu113
- Datasets 2.4.0
- Tokenizers 0.12.1