reward / README.md
claudiubarbu's picture
HW2-reward
572f948 verified
metadata
library_name: transformers
license: mit
base_model: openai-community/gpt2
tags:
  - trl
  - reward-trainer
  - generated_from_trainer
datasets:
  - piqa
model-index:
  - name: HW2-reward
    results: []

HW2-reward

This model is a fine-tuned version of openai-community/gpt2 on the piqa dataset.

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: 5e-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: 3

Training results

Framework versions

  • Transformers 4.44.2
  • Pytorch 2.0.0
  • Datasets 2.16.1
  • Tokenizers 0.19.1