How to use or convert my none tlt feature extraction model in the tlt platform?

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

Hi all.
my system info: (GPU , dockerfile of tlt version 2)
I have a feature extraction model which is based on Tensorflow. I want to use this model in tlt and then prune and quantize it.

(for example we can convert our pytorch model into tensorflow model by using ONNX, I’m looking for something like that to convert my model into tlt)

  • Is it possible to use our none tlt model in tlt platform (v 2.0 or 3.0)?
    None tlt model such as :
    1. classification model
    2. feature extraction model
  • for this purpose what can I do (which steps should be taken)?

any idea will be appreciated

2

See Frequently Asked Questions — Transfer Learning Toolkit 2.0 documentation

Q:Instead of NVIDIA provided pre-trained models, can I use TLT with my own or any open source pre-trained models?

A: No third party pre-trained models are supported by TLT.

1 Like
3

Thanks alot @Morganh