Faster inference at the edge
Deploying models in edge environments is complex. The remote nature of edge devices and their inherent constraints, such as battery consumption, make edge deployments a complex puzzle. OctoML solves that puzzle by deploying your model once across all your edge devices with a few simple lines of code.
Do more with less
Maximize model performance for your specific hardware. OctoML supports devices using chips manufactured by ARM, Intel, NVIDIA, Qualcomm, and Xilinx.
Build once, deploy anywhere
Build and train your model once, and OctoML will convert your model into an efficient, common format that can be executed on a number of devices.
Reduce hardware costs
See how your model performs across different hardware and pick the one that is the most optimal for the job.
We are excited about our collaboration with OctoML on Apache TVM – one of the most promising technologies that enables data scientists to run their ML models on a diverse range of Arm devices. The OctoML Platform is one of the preferred ML acceleration stacks for Arm hardware.
Sr. Manager, AI Ecosystem, Arm