update readme with usage caveats and calls for research

This write-up was loosely inspired in part by Mitchell et al.’s work on
[Model Cards for Model Reporting](https://arxiv.org/abs/1810.03993).
Adding such model usage sections could be good practice in general for
open source research projects with potentially broad applications.
This commit is contained in:
Jeff Wu
2019-03-06 11:30:53 -08:00
parent ed0dedcd55
commit 953530fc24

View File

@ -6,6 +6,22 @@ For now, we have only released a smaller (117M parameter) version of GPT-2.
See more details in our [blog post](https://blog.openai.com/better-language-models/).
## Usage
This repository is meant to be a starting point for researchers and engineers to experiment with GPT-2-117M. While GPT-2-117M is less proficient than GPT-2-1.5B, it is useful for a wide range of research and applications which could also apply to larger models.
### Some caveats
- GPT-2-117M robustness and worst case behaviors are not well-understood. As with any machine-learned model, carefully evaluate GPT-2-117M for your use case, especially if used without fine-tuning or in safety-critical applications where reliability is important.
- The dataset our GPT-2-117M was trained on contains many texts with [biases](https://twitter.com/TomerUllman/status/1101485289720242177) and factual inaccuracies, and thus GPT-2-117M is 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](mailto:languagequestions@openai.com) if youre doing interesting research with or working on applications of GPT-2-117M! Were 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
## Installation
Git clone this repository, and `cd` into directory for remaining commands
@ -53,7 +69,7 @@ and a valid install of [nvidia-docker 2.0](https://github.com/nvidia/nvidia-dock
docker run --runtime=nvidia -it gpt-2 bash
```
## Usage
## Sampling scripts
| WARNING: Samples are unfiltered and may contain offensive content. |
| --- |
@ -120,4 +136,4 @@ We are still considering release of the larger models.
## License
MIT
[MIT](./LICENSE)