I tried the NVIDIA GPU Cloud and used an nvidia-docker container with DIGITS 6. This allowed me to successfully train a DetectNet model with a dog image as explained here: https://github.com/dusty-nv/jetson-inference. However, once I downloaded the model and tried to run inference on the Jetson TX2, I ran into compatibility problems with Caffe (it seems the unit is expecting Caffe 0.15, while I had trained with Caffe 0.16.4 as included in DIGITS 6).
Is there any way to specify the use of Caffe 0.15 when using DIGITS on the NVIDIA GPU Cloud? Or is this older version of Caffe not supported? Thanks for your help!