CODEWITHSUNDEEP
tensorflow
schil
schil

Reputation: 322

tensorflow tools: graph_transform v/s toco

What are the differences in graph transformation when it comes to graph_transform tool and toco converter

Graph transform tool: https://github.com/tensorflow/tensorflow/tree/master/tensorflow/tools/graph_transforms

Toco: https://github.com/tensorflow/tensorflow/tree/master/tensorflow/contrib/lite/toco/

I understand that toco also converts operators to tf lite. But I constantly run into "unsupported ops" issue. So, I was wondering if I can only do graph transformations. And to do so, which tool should I use?

Upvotes: 0

Views: 451

Answers (1)

suharshs
suharshs

Reputation: 1088

The intention of TOCO is to convert to TensorFlow Lite models. Support for new ops will be added and eventually TF parity is a goal.

graph_transforms is a general tool for transformations on the TensorFlow GraphDef. Many of the existing transformations (like quantization) are a bit behind what TOCO/TFLite can do.

The decision of what to use or where to implement your graph transformations really depends on what you are trying to do. If you comment on your use case I may be able to help with that.

Upvotes: 1

Related Questions