Compatibility considerations for Hardware and OS when running Tensorflow Lite Model maker

TensorFlow General Discussion
, ,
1

Hi all,

I am currently running Tensorflow lite model maker for my object detection projects.

I am planning a hardware and OS upgrade. The hardware specs I am look at are:

  • i7-13700KF 3.40GHz 8+8CORE/24THREAD
  • DDR5 6000MHz 32GB(16GBx2)
  • 2TB GEN4 NVME M.2 SSD
  • MSI RTX4080

The OS will be Win 11 pro.

My first question is 1) are there any potential compatibility issues (eg, tensorflow stable release, CUDA ) with the hardware specs I am looking for. Am I overspending (eg. cannot reap the full benefit) or underspending (eg. bottleneck for the model training) on any component.

I am current running my model maker on MacBook Air Ventura:

  • i7 1.2Ghz Quad-core
  • 16GB 3733MHz LPDDR4
  • Mac Harddisk 500GB
  • Intel Iris Plus Graphics 1536 MB

My 2nd question is, how much of a performance improvement can I expect when I make the above upgrade. Currently, I have training that runs for 48hours.

3

Hi @Raymond_Wong ,

There shouldn’t be any major compatibility issues with the specifications you mentioned. But there is an issue with tflite-model-maker and team is working to resolve, in the mean time please try to use mediapipe-model-maker instead: here is an example gist.

Thank You