Reference: "Beginner’s Guide to Retrain GPT-2 (124M) to Generate Custom Text Content"
Code from the paper "Language Models are Unsupervised Multitask Learners".
We have currently released small (124M parameter) and medium (345M parameter) versions of GPT-2. While we have not released the larger models, we have released a dataset for researchers to study their behaviors.
See more details in our blog post.
This repository is meant to be a starting point for researchers and engineers to experiment with GPT-2.
- 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.
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
See DEVELOPERS.md
See CONTRIBUTORS.md
To retrain GPT-2 124M model on a custom text dataset:
PYTHONPATH=src ./train.py --dataset <file|directory|glob>
If you want to precompute the dataset's encoding for multiple runs, you can instead use:
PYTHONPATH=src ./encode.py <file|directory|glob> /path/to/encoded.npz
PYTHONPATH=src ./train.py --dataset /path/to/encoded.npz
Make sure cudnn
is installed. Some have reported that train.py
runs without it but has worse memory usage and might OOM.
https://github.com/openai/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 model.py. --memory_saving_gradients
is enabled by default for training the 345M model.
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).
To do distributed on multiple GPUs or machines using Horovod:
mpirun -np 4 \
-H localhost:4 \
-bind-to none -map-by slot \
-x NCCL_DEBUG=INFO -x LD_LIBRARY_PATH -x PATH \
-x PYTHONPATH=src \
-mca pml ob1 -mca btl ^openib \
/home/jovyan/gpt-2/train-horovod.py --dataset encoded.npz
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:
@article{radford2019language,
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},
year={2019}
}
We may release code for evaluating the models on various benchmarks.
We are still considering release of the larger models.