trl_rm_tldr_gptj

This model is a fine-tuned version of EleutherAI/gpt-j-6b on the TL;DR dataset. It achieves the following results on the evaluation set:

• Loss: 0.6624
• Accuracy: 0.6857

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: 1.41e-05
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.5633	1.0	22660	0.6624	0.6857

Framework versions

• PEFT 0.7.1.dev0
• Transformers 4.36.2
• Pytorch 2.1.2
• Datasets 2.15.0
• Tokenizers 0.15.0

BibTex Citation

If you would like to cite our paper when using the model, please use

@article{sun2024supervised,
  title={Supervised Fine-Tuning as Inverse Reinforcement Learning},
  author={Sun, Hao},
  journal={arXiv preprint arXiv:2403.12017},
  year={2024}
}