Remotely use tensorboard to train the model in Docker on the local visualization server

1. When connecting to ssh, redirect port 6006 of the server to your own machine:

 ssh -L 16006:127.0.0.1:6006 username@remote_server_ip -p docker port

Or: ssh -L 8008:localhost:6006 username@remote server ip
where: 16006:127.0.0.1 represents port 16006 on your own machine, and 6006 is the port used by tensorboard on the server.

2. Use port 6006 on the server to start tensorboard normally:

tensorboard --logdir=summary_dir --port=6006

3. Enter the address in the local browser:

  127.0.0.1:16006