francescocarpanese/code_acme

Visualization of results after training within docker.

francescocarpanese opened this issue · 0 comments

To visualize plot within docker image several solutions can be considered.

  1. X11-unix
    Explanation link. However when distributing the package not all OS has X11.

  2. jupyter notebook, exporting port, and open in browser.

jupyter notebook --port=8888 --allow-root --no-browser --ip=0.0.0.0

This solution might be ok but:

  • Currently needs to run as root. This is related to #9
  • Needs to provide scripts for visualization as jupyter notebook which for the pure purpose of visualizing the results it's an overkill.
  1. Generate files in image format
    Not that much convenient for a quick visualization

  2. bokeh plot
    link
    Export port. Need to install bokeh in the docker image.

This tutorial for visualization in AWS with docker link might come handy.

  1. Store as part of tensorboard events? See how to deal with loggers