# Ultralytics version for voxl-tflite-server yolov8

Source: https://forum.modalai.com/topic/3888/ultralytics-version-for-voxl-tflite-server-yolov8
Category: VOXL SDK and Software (https://forum.modalai.com/category/47/voxl-sdk-and-software)
Tags: voxl-sdk
Posted: 2024-10-25 22:23:10 UTC by atomsmasher9
Replies: 1 · Views: 875

## atomsmasher9 · 2024-10-25 22:23:10 UTC

Hello!

I am trying to export a custom yolov8n model from Ultralytics to use on the VOXL2.  I incorrectly developed it on a newer version of TensorFlow and have now downgraded to TensorFlow 2.8.0 to produce a correct model.  Upon trying to convert even a default, pre-trained yolov8n model I am getting export errors (for example not having dtensors available).

My question is where was the model loaded in the voxl-tflite-server repo produced from? Was Ultralytics used, if so what version? It not Ultralytics can the source training code for the uploaded weights be provided?

## Reply by Zachary Lowell 0 · 2024-10-28 14:33:07 UTC

@atomsmasher9 said in [Ultralytics version for voxl\-tflite\-server yolov8](/post/19445):
> 
> Reply

Hi @atomsmasher9  - so ultralytics is leveraged for training - have you followed the instructions within this repository?

https://gitlab.com/voxl-public/support/voxl-train-yolov8/-/tree/master?ref_type=heads

Currently I have some updated written for this as well - so for example in export.py, this line:     model.export("tflite") -->     model.export(model = "tflite"), but I have yet to make the PR or commit for it - the README is relatively straight forward for how you can build a model leveraging docker, an nvidia gpu, and ultralytics to create the custom model. I was able to do these below and have a successful model created and uploaded to the voxl2 - ensure you place the model in /usr/bin/dnn directory alongside the labels.txt file to ensure you have the right labels - also you can manually update the file path in the conf file in /etc/modalai/voxl-tflite-server.conf.