Lightning Talk: Accelerating Experimentation...
summary
Lightning’s platform plus Infineon toolchain trains and deploys tiny language models (25M, 13M, 8M parameters) to Cortex‑M55/U55 microcontrollers with 6.5 MB SRAM, using hundreds of parallel data and training sweeps and reproducible pipelines. Result: sub 30MB language models running fully on‑device with inference under 100ms per token and reported 98% energy savings versus cloud, achieved by quantization and microcontroller conversion pipelines and validated end‑to‑end in live demos.
AI-written summary · about
chapters (6)
- 0:00Introduction
- 0:42Lightning AI platform overview
- 5:14Edge language models on Pox Edge
- 8:24Live demo of edge inference
- 9:29Platform benefits and call to action
- 10:33Q&A
related talks
10mLightning Talk: Debugging Edge AI on Zephyr and Lessons Learned
14mLightning Talk: Every Millisecond Counts: The Fine-tuning Journey of an Ultra-Eff...
01h 53mTinyML: Machine Learning’s Future is Tiny and Bright
27mWhat’s TinyML good for
10mExecuTorch on Microcontrollers: Deploying PyTorch To...
10mLightning Talk: Software Development for ML and RISC-V Vector Acceler...
13mAI inference for the smallest processors
14mFlexible Deployment of PyTorch Models on MCU-Class...