I forked the repo and tried to run it in Google Colab.
I did some revision on formatting and data fetching code.
My current version of Colab notebook is at https://colab.research.google.com/drive/1UxYltzePhc3MJzZhuBycBRR3ettQc8gL#revisionId=0B497FfEid-K5ZFNML3Y3SlBoZDZlT2tiUVBiUzkzU2VPaGxJPQ.
I was able to download the files. However, when I started training the sem_seg data it runs out of RAM.
When I have all batches and data loaded it has already consumed over 60% of available RAM, and when the session is initialized it takes up all the memory prior to starting the training.
I forked the repo and tried to run it in Google Colab. I did some revision on formatting and data fetching code. My current version of Colab notebook is at https://colab.research.google.com/drive/1UxYltzePhc3MJzZhuBycBRR3ettQc8gL#revisionId=0B497FfEid-K5ZFNML3Y3SlBoZDZlT2tiUVBiUzkzU2VPaGxJPQ. I was able to download the files. However, when I started training the sem_seg data it runs out of RAM. When I have all batches and data loaded it has already consumed over 60% of available RAM, and when the session is initialized it takes up all the memory prior to starting the training.