Hello I had some questions regarding transfer learning with 0.6.0 release,
I want to use the checkpoint version 0.6.0 and tune with my own data.
Which branch should i use for continuing training from a release model : transfer-learning2 or the master branch.
This topic is raised because its written that the latest model 0.6.0 is not backward compatible with 0.5.1.
Will the model exported with transfer-learning2 branch be compatible with 0.6.0?