Skip to content
Train 774M, 1.5B models with the latest optimizer
Branch: finetuning
Clone or download
Pull request Compare This branch is 26 commits ahead of nshepperd:finetuning.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.

Train GPT-2

Fine tuning on custom datasets

To retrain GPT-2 small (124M parameter), medium (355M parameter), large (774M parameter), and (upcoming) mega (1.5B) parameter models on a custom text dataset:


We use Google AI's latest SM3 Optimizer: Memory-Efficient Adaptive Optimization

PYTHONPATH=src ./ --dataset <file|directory|glob>

If you want to precompute the dataset's encoding for multiple runs, you can instead use:

PYTHONPATH=src ./ <file|directory|glob> /path/to/encoded.npz
PYTHONPATH=src ./ --dataset /path/to/encoded.npz

Make sure cudnn is installed. Some have reported that runs without it but has worse memory usage and might OOM.

Gradient Checkpointing is included to reduce the memory requirements of the model, and can be enabled by --memory_saving_gradients. The checkpoints are currently chosen manually (poorly) by just adding layer 10 to the 'checkpoints' collection in --memory_saving_gradients is enabled by default for training the 345M model.

Validation loss

Set --val_every to a number of steps N > 0, and "validation" loss against a fixed sample of the dataset will be calculated every N steps to get a better sense of training progress. N around 200 suggested. You can set --val_dataset to choose a separate validation dataset, otherwise it defaults to a sample from the train dataset (so not a real cross-validation loss!).


You can use SGD instead of Adam with --optimizer sgd. This also helps conserve memory when training the 345M model. Note: the learning rate needs to be adjusted for SGD, due to not having Adam's gradient normalization (0.0006 seems to be a good number from some experiments).

Multi gpu (not supported, PR is welcome)

Code from the paper "Language Models are Unsupervised Multitask Learners".

We have currently released small (124M parameter), medium (355M parameter), and large (774M parameter) versions of GPT-2*, with only the full model as of yet unreleased. We have also released a dataset for researchers to study their behaviors.

You can read about GPT-2 and release decisions in our original blog post and 6 month follow-up post.

* Note that our original parameter counts were wrong due to an error (in our previous blog posts and paper). Thus you may have seen small referred to as 117M and medium referred to as 345M.


This repository is meant to be a starting point for researchers and engineers to experiment with GPT-2.

For basic information, see our model card.

Some caveats

  • GPT-2 models' robustness and worst case behaviors are not well-understood. As with any machine-learned model, carefully evaluate GPT-2 for your use case, especially if used without fine-tuning or in safety-critical applications where reliability is important.
  • The dataset our GPT-2 models were trained on contains many texts with biases and factual inaccuracies, and thus GPT-2 models are likely to be biased and inaccurate as well.
  • To avoid having samples mistaken as human-written, we recommend clearly labeling samples as synthetic before wide dissemination. Our models are often incoherent or inaccurate in subtle ways, which takes more than a quick read for a human to notice.

Work with us

Please let us know if you’re doing interesting research with or working on applications of GPT-2! We’re especially interested in hearing from and potentially working with those who are studying

  • Potential malicious use cases and defenses against them (e.g. the detectability of synthetic text)
  • The extent of problematic content (e.g. bias) being baked into the models and effective mitigations





GPT-2 samples

WARNING: Samples are unfiltered and may contain offensive content.

While we have not yet released GPT-2 itself, you can see some samples from it in the gpt-2-samples folder. We show unconditional samples with default settings (temperature 1 and no truncation), with temperature 0.7, and with truncation with top_k 40. We show conditional samples, with contexts drawn from WebText's test set, with default settings (temperature 1 and no truncation), with temperature 0.7, and with truncation with top_k 40.


Please use the following bibtex entry:

  title={Language Models are Unsupervised Multitask Learners},
  author={Radford, Alec and Wu, Jeff and Child, Rewon and Luan, David and Amodei, Dario and Sutskever, Ilya},

Reference: "Beginner’s Guide to Retrain GPT-2 (117M) to Generate Custom Text Content"

Future work

We may release code for evaluating the models on various benchmarks.

We are still considering release of the larger models.



You can’t perform that action at this time.