search

am15h / tflite_flutter_plugin

TensorFlow Lite Flutter Plugin
https://pub.dev/packages/tflite_flutter
Apache License 2.0
508 stars 353 forks source link

Enable quantization flag #109

Closed mspnr closed 3 years ago

mspnr commented 3 years ago

In the last version of TFLite there is an Enable quantization flag which allow you to use GPU delegate for quantized models. Otherwise you end up loading model as without GPU delegate (at least from my experiments).

Here are the links to the TFLite:

Actually I already experimented with it and it kinda working: You can take a look into this commit https://github.com/applikationsprogramvara/tflite_flutter_plugin/commit/11f81946c17ae5c5bf6361a64b13b94abeaf11ac Though I am not sure how it will work with the previous versions, where the quantization flag is still missing.

You can consider to add the code to the main branch.

am15h commented 3 years ago

Thanks @applikationsprogramvara, Would you like to make a PR for this?