Record 042 · Runtimes · 2017
67%17.42.U
TensorFlow Lite
“The mobile/edge runtime that put detection and speech on phones and Coral boards.”
TFLite (now LiteRT) was how Android, Coral, and microcontrollers ran models before ExecuTorch and ONNX ate some of the lunch.
Opened on-phone and on-dongle inference as a standard.
- FlatBuffer mobile models
- NNAPI/GPU delegates
- Microcontroller port
Models in a mobile binary.
