Hi there! 😃 Like you, I'm an NLP researcher looking for a solution to run GPT-2 project on Tensorflow 2.x. As you may already know, GPT-2 has been developed by OpenAI team and use Tensorflow 1.x as a base framework and unfortunately, Tensorflow 1.x and 2.x are very different and slightly hard to upgrade from 1.x version to 2.x. Tensorflow 2.x removed some libraries / modules that are frequently used in version 1.x such as tf.contrib and also moved / modified a few other ones such as "hparams". In GPT-2 original source code, some parts were broken because of these problems. To use GPT-2 for my private project, I planned to rewrite all source code by Tensorflow 2.0 to fit my old codes and to better understand GPT-2 model. But I realized that my deadlines were killing me 😛 and in fact, maybe a new implementation could not achieve the best performance of GPT-2. So, I decided to change the GPT-2 source code with some of the smallest possible differences to keep its capabilities. To do it, I cloned a project from a super nice guy: Awesome GPT-2 with training script and do some improvements. This is what I have changed:

• Add Hparams class to replace "tf.contrib.training.HParams", you can find it in src/hparams.py file.
• Add "graph_def_editor" module to replace the "graph_editor" module. This is an awesome project written by CODAIT. I will place the path of this project in here.
• I have added a new option for the training script of nshepperd to choose GPU or CPU device. You can find more details in train.py file
• And a bunch of other minor changes ...

To test this project, you can use pip or whatever you want to set up your environment. If you like me, an Anaconda user 😃, I prepared for you an environment file to make your life easier. You can follow these steps below:

1. Install anaconda
2. cd ./src
3. conda env create -f ./environment.yml -p ./.env
4. conda activate ./.env
5. To train and test, you can follow instructions from nshepperd project I added below 😃. Remember, you can choose which device your model runs on by '--device' flag 😛.

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

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

We have currently released small (117M 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 117M 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

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.

MIT