Hi, I am trying to use MediaPipe to construct on-device LLM driven agents workflow for academic research. I wonder is it possible to use QNN TFLite Delegate to run LLM like gemma by Java API? Is it possible to delegate a “calculator” node to NPU within a MediaPipe Graph?
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Help converting tflite models with mediapipe | 2 | 550 | January 22, 2025 | |
| Asking about coral-tflite-delegate in raspberry | 1 | 1030 | February 13, 2024 | |
| TFLITE to mediapipe | 1 | 230 | June 26, 2024 | |
| May I ask how I can use tflite/sensor flow more effectively on Snapdragon processors? | 3 | 515 | November 17, 2023 | |
| Help Needed: Converting Fine-Tuned Gemma 3 Vision Model to TensorFlow Lite (TFLite) | 4 | 627 | August 15, 2025 |