Alif Semiconductor announced that developers can now use the ExecuTorch Runtime, a quantization extension of the popular PyTorch ML framework, for AI applications built to run on its Ensemble E4/E6/E8 series of MCUs and fusion processors.
The ExecuTorch extension enables developers of AI models aimed at resource-constrained hardware targets such as MCUs to use the PyTorch framework to create lightweight models that support low-latency, high-accuracy inference. Since the Ensemble E4, E6 and E8 products include the Arm Ethos-U85 NPU, which supports transformer-based ML networks, developers can now use PyTorch and the ExecuTorch Runtime to build generative AI applications to run locally in battery-powered endpoint devices such as smart glasses, human-to-computer interfacing, healthcare and diagnostics, robotics, transportation, toys and education, smart homes, and smart city equipment.
At the PyTorch Conference (Moscone West, San Francisco CA, 22-23 October), Arm demonstrated examples of generative AI applications compiled with ExecuTorch and running on Alif’s Ensemble E8 fusion processor, including: A small language model running on the Ensemble E8, capable of generating stories suitable for children in response to visual prompts, and extendable to verbal prompts. Real-time on-device speech-to-text models, suitable for integration into wearables such as smart glasses. The demo performs real-time transcription of speech to enable live captioning.
The Alif DK-E8 development board used in the PyTorch Conference demonstrations, which supports development on the entire E4/E6/E8 series, is available to purchase now.
Leave a Reply