\n\nGPT-2 is a Natural Language Processing model [developed by OpenAI](https://openai.com/blog/better-language-models/) for text generation. It is the successor to the GPT (Generative Pre-trained Transformer) model trained on 40GB of text from the internet. It features a Transformer model that was brought to light by the [Attention Is All You Need](https://arxiv.org/abs/1706.03762) paper in 2017. The model has two versions - `117M` and `345M` - that differ based on the amount of training data fed to it and the number of parameters they contain. \n \n \nThe `345M` model is currently the largest one available while the 1.5B model is being vetted for release with selected partners. Only recently has OpenAI decided to release its training weights as part of its Staged Release plan. There have been several implications and debates over their release plan regarding misuse.\n \n \nFinally, `gpt2-client` is a wrapper around the original [`gpt-2` repository](https://github.com/openai/gpt-2) that features the same functionality but with more accessiblity, comprehensibility, and utilty. You can play around both GPT-2 models with less than five lines of code.\n\n> ***Note**: This client wrapper is in no way liable to any damage caused directly or indirectly. Any names, places, and objects referenced by the model are fictional and seek no resemblance to real life entities or organisations. Samples are unfiltered and may contain offensive content. User discretion advised.*\n\n
Installation
\n\nInstall client via `pip`. Ideally, `gpt2-client` is well supported for Python >= 3.5 and TensorFlow >= 1.X. Some libraries may need to be reinstalled or upgraded using the `--upgrade` flag via `pip` if *Python 2.X* is used.\n\n```bash\npip install gpt2-client\n```\n\n> ***Note:*** `gpt2-client` is **not** compatible with TensorFlow 2.0\n\n
Getting started
\n\n**1. Download the model weights and checkpoints**\n\n```python\nfrom gpt2_client import GPT2Client\n\ngpt2 = GPT2Client('117M', save_dir='models') # This could also be `345M` or `774M`. Rename `save_dir` to anything.\ngpt2.load_model(force_download=False) # Use cached versions if available.\n```\n\nThis creates a directory called `models` in the current working directory and downloads the weights, checkpoints, model JSON, and hyper-parameters required by the model. Once you have called the `load_model()` function, you need not call it again assuming that the files have finished downloading in the `models` directory.\n\n> ***Note:*** Set `force_download=True` to overwrite the existing cached model weights and checkpoints\n\n**2. Start generating text!**\n\n```python\nfrom gpt2_client import GPT2Client\n\ngpt2 = GPT2Client('117M') # This could also be `345M` or `774M`\n\ngpt2.generate(interactive=True) # Asks user for prompt\ngpt2.generate(n_samples=4) # Generates 4 pieces of text\ntext = gpt.generate(return_text=True) # Generates text and returns it in an array\ngpt2.generate(interactive=True, n_samples=3) # A different prompt each time\n```\n\nYou can see from the aforementioned sample that the generation options are highly flexible. You can mix and match based on what kind of text you need generated, be it multiple chunks or one at a time with prompts.\n\n**3. Generating text from batch of prompts**\n\n```python\nfrom gpt2_client import GPT2Client\n\ngpt2 = GPT2Client('117M') # This could also be `345M` or `774M`\n\nprompts = [\n \"This is a prompt 1\",\n \"This is a prompt 2\",\n \"This is a prompt 3\",\n \"This is a prompt 4\"\n]\n\ntext = gpt2.generate_batch_from_prompts(prompts) # returns an array of generated text\n```\n\n**4. Fine-tuning GPT-2 to custom datasets**\n\n```python\nfrom gpt2_client import GPT2Client\n\ngpt2 = GPT2Client('117M') # This could also be `345M` or `774M`\n\nmy_corpus = './data/shakespeare.txt' # path to corpus\ncustom_text = gpt2.finetune(my_corpus, return_text=True) # Load your custom dataset\n```\n\nIn order to fine-tune GPT-2 to your custom corpus or dataset, it's ideal to have a GPU or TPU at hand. [Google Colab](http://colab.research.google.com) is one such tool you can make use of to re-train/fine-tune your custom model.\n\n
Contributing
\n\nSuggestions, improvements, and enhancements are always welcome! If you have any issues, please do raise one in the Issues section. If you have an improvement, do file an issue to discuss the suggestion before creating a PR.\n\nAll ideas \u2013 no matter how outrageous \u2013 welcome!\n\n
![\"Pypi](\"https://img.shields.io/pypi/v/gpt2-client?color=%231dd1a1&logo=%231dd1a1&logoColor=%231dd1a1\")
\n\n\n![\"GitHub](\"https://img.shields.io/badge/license-MIT-%23feca57\"){kind=icon}
\n\n\n![](\"https://colab.research.google.com/assets/colab-badge.svg\"){kind=icon}
\n\n\n![](\"https://github.com/rish-16/gpt2client/raw/master/assets/demo.png\")