Back to browse
GitHub Repository
5 starsPython

TiGrIS, a tiling compiler that fits ML models onto embedded devices

by asteinh·Apr 29, 2026·20 points·0 comments

AI Analysis

●●●BangerBig BrainNiche Gem

Fits 4.5MB models into 256KB SRAM via tiling with zero dynamic memory allocation.

Strengths
  • Shrinks 4.59MB peak memory to 256KB via temporal partitioning and tiling.
  • Zero dynamic allocation runtime ensures deterministic execution on bare-metal hardware.
  • Target-agnostic binary plan works across ESP32, Cortex-M, and POSIX hosts.
Weaknesses
  • Requires separate runtime repository, adding integration complexity to existing firmware builds.
  • Tiling overhead may increase latency compared to native operators on hardware.
Category
Target Audience

Embedded ML engineers, firmware developers working with Cortex-M or ESP32

Similar To

Apache TVM · TensorFlow Lite for Microcontrollers · ONNX Runtime

Similar Projects

Developer Tools●●Solid

Overlay map tiles onto Autodesk's 3D BIM Viewer

Coordinate transform pipeline from Revit to WGS84 that Autodesk's own viewer doesn't provide.

Niche GemBig Brain
gubets
402mo ago