Welcome to the official Google organization on Hugging Face!

Google collaborates with Hugging Face across open science, open source, cloud, and hardware to enable companies to innovate with AI on Google Cloud AI services and infrastructure with the Hugging Face ecosystem.

Featured Models and Tools

• Gemma Family of Open Multimodal Models
  • Gemma is a family of lightweight, state-of-the-art open models from Google, built from the same research and technology used to create the Gemini models
  • PaliGemma is a versatile and lightweight vision-language model (VLM)
  • CodeGemma is a collection of lightweight open code models built on top of Gemma
  • RecurrentGemma is a family of open language models built on a novel recurrent architecture developed at Google
  • ShieldGemma is a series of safety content moderation models built upon Gemma 2 that target four harm categories
• BERT, T5, and TimesFM Model Families
• Author ML models with MaxText, JAX, Keras, Tensorflow, and PyTorch/XLA

Open Research and Community Resources

• Google Blogs:
  • https://blog.google/
  • https://cloud.google.com/blog/
  • https://deepmind.google/discover/blog/
  • https://developers.google.com/learn?category=aiandmachinelearning
• Notable GitHub Repositories:
  • https://github.com/google/jax is a Python library for high-performance numerical computing and machine learning
  • https://github.com/huggingface/Google-Cloud-Containers facilitate the training and deployment of Hugging Face models on Google Cloud
  • https://github.com/pytorch/xla enables PyTorch on XLA Devices (e.g. Google TPU)
  • https://github.com/huggingface/optimum-tpu brings the power of TPUs to your training and inference stack
  • https://github.com/openxla/xla is a machine learning compiler for GPUs, CPUs, and ML accelerators
  • https://github.com/google/JetStream (and https://github.com/google/jetstream-pytorch) is a throughput and memory optimized engine for large language model (LLM) inference on XLA devices
  • https://github.com/google/flax is a neural network library for JAX that is designed for flexibility
  • https://github.com/kubernetes-sigs/lws facilitates Kubernetes deployment patterns for AI/ML inference workloads, especially multi-host inference workloads
  • https://github.com/GoogleCloudPlatform/ai-on-gke is a collection of AI examples, best-practices, and prebuilt solutions
• Google AI Research Papers: https://research.google/

On-device ML using Google AI Edge

• Customize and run common ML Tasks with low-code MediaPipe Solutions
• Run pretrained or custom models on-device with Lite RT (previously known as TensorFlow Lite)
• Convert TensorFlow and JAX models to LiteRT
• Convert PyTorch models to LiteRT and author high performance on-device LLMs with AI Edge Torch
• Visualize and debug models with Model Explorer (🤗 Space)

Partnership Highlights and Resources

• Select Google Cloud CPU, GPU, or TPU options when setting up your Hugging Face Inference Endpoints and Spaces
• Train and Deploy Hugging Face models on Google Kubernetes Engine (GKE) and Vertex AI directly from Hugging Face model landing pages or from Google Cloud Model Garden
• Integrate Colab notebooks with Hugging Face Hub via the HF_TOKEN secret manager integration and transformers/huggingface_hub pre-installs
• Leverage Hugging Face Deep Learning Containers (DLCs) for easy training and deployment of Hugging Face models on Google Cloud infrastructure

Read about our principles for responsible AI at https://ai.google/responsibility/principles