Spaces:
Running
title: README
emoji: 🐢
colorFrom: purple
colorTo: gray
sdk: static
pinned: false
Intel and Hugging Face are building powerful optimization tools to accelerate training and inference with Transformers.
Models
Check out Intel's models here on our Hugging Face page or directly through the Hugging Face Models Hub search. Here are some of Intel's models:
| Model | Type |
|---|---|
| dpt-hybrid-midas | Monocular depth estimation |
| llava-gemma-2b | Multimodal |
| gpt2 on Gaudi | Text generation |
| neural-chat-7b-v3-3-int8-ov | Text generation |
Datasets
Intel has created a number of datasets for use in fine-tuning both vision and language models. Check out the datasets below on our page, including orca_dpo_pairs for natural language processing tasks and SocialCounterfactuals for vision tasks.
Collections
Our Collections categorize models that pertain to Intel hardware and software. Here are a few:
| Collection | Description |
|---|---|
| DPT 3.1 | Monocular depth (MiDaS) models, leveraging state-of-the-art vision backbones such as BEiT and Swinv2 |
| Whisper | Whisper models for automatic speech recognition (ASR) and speech translation, quantized for faster inference speeds. |
| Intel Neural Chat | Fine-tuned 7B parameter LLM models, one of which made it to the top of the 7B HF LLM Leaderboard |
Spaces
Check out Intel's leaderboards and other demo applications from our Spaces:
| Space | Description |
|---|---|
| Powered-by-Intel LLM Leaderboard | Evaluate, score, and rank open-source LLMs that have been pre-trained or fine-tuned on Intel Hardware 🦾 |
| Intel Low-bit Quantized Open LLM Leaderboard | Evaluation leaderboard for quantized language models |
Blogs
Get started with deploying Intel's models on Intel architecture with these hands-on tutorials from blogs written by staff from Hugging Face and Intel:
| Blog | Description |
|---|---|
| Building Cost-Efficient Enterprise RAG applications with Intel Gaudi 2 and Intel Xeon | Develop and deploy RAG applications as part of OPEA, the Open Platform for Enterprise AI |
| Running Large Multimodal Models on an AI PC's NPU | Run the llava-gemma-2b model on an AI PC's NPU |
| A Chatbot on your Laptop: Phi-2 on Intel Meteor Lake | Deploy Phi-2 on your local laptop with Intel OpenVINO in the Optimum Intel library |
| Partnering to Democratize ML Hardware Acceleration | Intel and Hugging Face collaborate to build state-of-the-art hardware acceleration to train, fine-tune and predict with Transformers |
Documentation
To learn more about deploying models on Intel hardware with Transformers, visit the resources listed below.
Optimum Habana - To deploy on Intel Gaudi accelerators, check out optimum-habana, the interface between Gaudi and the 🤗 Transformers and Diffusers libraries. To install the latest stable release:
pip install --upgrade-strategy eager optimum[habana]
Optimum Intel - To deploy on all other Intel architectures, check out optimum-intel, the interface between Intel architectures and the 🤗 Transformers and Diffusers libraries. Depending on your need, you can use these backends:
| Accelerator | Installation |
|---|---|
| Intel Neural Compressor | pip install --upgrade --upgrade-strategy eager "optimum[neural-compressor]" |
| OpenVINO | pip install --upgrade --upgrade-strategy eager "optimum[openvino]" |
| Intel Extension for PyTorch | pip install --upgrade --upgrade-strategy eager "optimum[ipex]" |
Join Our Dev Community
Please join us on the Intel DevHub Discord to ask questions and interact with our AI developer community!