Tensorflow Version in Clara Training

Hi,
I am trying to look at the documentation of Clara training. Unfortunately, for me it looks to be all over the place,and sometimes cyclical; not easy to navigate or maybe I am missing something. What i understood is that “The Clara Train 3.1 release is based on NVIDIA’s container for TensorFlow release 20.08 with support for NVIDIA Ampere GPUs.” Looking at the release notes of

I could see that both tensorflow 1.15.3 and 2.2.0 are supported. Upone launching the Clara docker image in an interactive session, I see that the tensorflow version is 1.15.3. A lot of work which I do uses tf 2.0. Is there any version/way where the clara train can use tf2.0?
Sorry if the question is elementary. We are looking for a way to standardise clara as the only framework for training, these things would help.
Regards
krishnan

Hi
Thanks for your intertest in Clara Train SDK.
Sorry for the confusion. You are correct regarding Clara train 3.1 based on Nvidia tensor flow and support for Ampere GPUs. However the sdk uses tensor flow 1.x and does NOT use tf2.x.
We have just released clara train V4 in early access that is based on Monai that uses PyTorch. You can sign up for early access https://developer.nvidia.com/clara-medical-imaging