Hugging Face Expands Optimum Graphcore Model Lineup
Hugging Face has once again worked with Graphcore, bringing transformer models to optimal performance on Graphcore IPUs. Nine new models have been added to the program.
Created on May 26|Last edited on May 27
Comment
Hugging Face partnered with Graphcore in 2021 through the Hardware Partner Program, their partnership supporting the increasing accessibility of using models on Graphcore IPU processors. This partnership comes to a head with Optimum Graphcore models via the Hugging Face Optimum platform, the first of which being an IPU-optimized implementation of the classic model BERT.
Today they've announced a huge expansion to the lineup of IPU-ready Graphcore Optimum models available for users to try out. These models are all optimized to run the best they can on Graphcore IPUs, all through the easy-to-use Hugging Face Optimum library.
New Optimum Graphcore Models
Here's the quick list of all the newly added models optimized for Graphcore IPUs.
New Optimum Graphcore Computer Vision Models:
- ViT
New Optimum Graphcore NLP Models:
- GPT-2
- RoBERTa
- DeBERTa
- BART
- LXMERT
- T5
New Optimum Graphcore Speech Models:
- HuBERT
- Wav2Vec2
This brings the total number of models available for optimal performance on Graphcore IPUS to 10, including the original BERT model already available.
Find out more
Add a comment
Tags: ML News
Iterate on AI agents and models faster. Try Weights & Biases today.