We recently announced Inference for PROs, our new offering that makes larger models accessible to a broader audience. This opportunity opens up new possibilities for running end-user applications using Hugging Face as a platform. An example of such an application is the AI Comic Factory – a Space that has proved incredibly popular. Thousands of users have tried it to    

A spectre is haunting chat models – the spectre of incorrect formatting! tl;dr Chat models have been trained with very different formats for converting conversations into a single tokenizable string. Using a format different from the format a model    

Generative AI models, such as Stable Diffusion XL (SDXL), enable the creation of high-quality, realistic content with wide-ranging applications. However, harnessing the power of such models presents significant challenges and computational costs. SDXL is a large image generation model whose UNet component is about three times as large as the one in the previous version of the model. Deploying a model like this in production is challenging due to the increased memory requirements, as well as increased inference times. Today, […]

Gradio is a popular Python library for creating interactive machine learning apps. Traditionally, Gradio applications have relied on server-side infrastructure to run, which can be a hurdle for developers who need to host their applications. Enter Gradio-lite (@gradio/lite): a library that leverages Pyodide to bring Gradio directly to your browser. In this blog post, we’ll explore what @gradio/lite is, go over example code, and discuss the benefits it offers for running Gradio applications. What is @gradio/lite?    

Stable Diffusion XL (SDXL) is the latest latent diffusion model by Stability AI for generating high-quality super realistic images. It overcomes challenges of previous Stable Diffusion models like getting hands    

RLHF / ChatGPT has been a popular research topic these days. In our quest to research more on RLHF, this blog post attempts to do a reproduction of OpenAI’s 2019 original RLHF codebase at openai/lm-human-preferences. Despite its “tensorflow-1.x-ness,” OpenAI’s original codebase is very well-evaluated and benchmarked, making it a good place to study RLHF implementation engineering details. We aim to: reproduce OAI’s results in stylistic tasks and match the learning curves of openai/lm-human-preferences. present a checklist of implementation details, similar […]