code for exporting onnx model to tflite

[dsbyprateekg]

Using this code, I am able to convert my fine tuned model from onnx to tflite format. I have also added the inference code. This code has been tested in google colab notebook with A100 GPU.

Description

First I converted my fine tuned model to onnx format using below code-
from rfdetr import RFDETRBase
model = RFDETRBase(pretrain_weights="rf-detr_lego_model_30epochs.pth")
model.export()

Then I use following library for converting onnx model to tflite format: https://github.com/PINTO0309/onnx2tf
I also added inference code with converted tflite model to verify if its working as expected or not.

List any dependencies that are required for this change.

code has been tested in google colab notebook with A100 GPU.

Please delete options that are not relevant.

• This change requires a documentation update

How has this change been tested, please provide a testcase or example of how you tested the change?

I have tested the inference using converted tflite model and its output is as same as original fine tuned model.

[probicheaux]

Hi, really cool, and thank you so much for this!

1. can you move the notebook to a new dir notebooks
2. would it be possible to add this new format as a keyword to the export command and do it there? no worries if not

[CLAassistant]

All committers have signed the CLA.

[dsbyprateekg]

@probicheaux I have moved the notebook to a new dir notebooks.

[SkalskiP]

Hi @dsbyprateekg 👋🏻 thanks a lot for this PR! Looks like the code worked well for @gpokat in this issue — great job!

Any chance you could migrate it from the notebook into our export module?