Esperanto Technologies describes its custom ONNXRuntime execution backend for running small language models on ET-SoC-1 hardware, including options for Esperanto’s EtGlowExecutionProvider.
Esperanto Technologies describes an optimized approach to preparing inputs for small language model inference on the ET-SoC-1 using ONNXRuntime IOBindings.
The company describes a keyword-augmented method for summarizing clinical notes and discusses how it may mitigate hallucinations, along with its methodologies and results.
An IEEE Spectrum article by Stacey Higginbotham says RISC-V makes it easier for companies to design and build specialized chips for graphics and the Internet of Things.
The page provides a general inquiry form and lists the phone number (650) 319-7357 and locations in Austin, Texas; Portland, Oregon; Barcelona, Spain; and Belgrade, Serbia.
Esperanto Technologies has ceased operations, and Nekko.ai acquired its intellectual property. The page also describes the Esperanto AI / HPC Compute Server for generative AI inference, HPC, and mixed-mode workloads.