tensorflow / tensorflow/models

TF2 Object Detection Model Post training Quantization Error

Open
#10,427 1 comment 0 reactions 3 assignees View on GitHub

@pkulzc is already working on this.

Since Dec 17, 2021.

models:research:odapi type:support
Dominant language
Python
Stars
77.7k
Forks
44.8k
PR merge metrics
No merged PRs in 30d

Description

After training an Object detection model with SSD MobileNet v2_fpnlite_320x320_coco17_tpu-8, I used the exporter_main_v2.py, and then used the 'export_tflite_graph_tf2.py' to export the graph before converting it to tflite. Now I am trying to do the post training quantization using the code provided in this #9033 , and in the convert step I am getting an error : 'Given shapes, [1,20,20,128] and [1,19,19,128], are not broadcastable.Node number 68 (ADD) failed to prepare.'

I first did the above using Tensorflow 2.6.2, and after that tried in tf-nightly '2.8.0-dev20211216'.

I am not much familiar with version controls of Tensorflow so could you please help with telling me what is the issue going on here ?

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.