How to configure/use Tensorboard when ssh'ing into a remote machine?

So I guess I’m slow to the party cuz I’ve never used Tensorboard (though I’ve heard rumors it is integrated into v1 and maybe v2). Anyhow, would love to try and use it with code running on a remote machine I ssh into like this:

ssh <username>@<ip to machine> -L 8889:127.0.0.1:9999

This gives me everything I need to run my notebooks on localhost:8889.

What do I need to do in order to run/view tensorboard as well from localhost on my local machine?

You will need another forward port for tensoboard, to see the results. I can’t remember on which port it runs (but it tells you when you launch it), so forward that port to one you like on your local machine.

Thanks.

Is the syntax as such:

ssh <username>@<ip to machine> -L 8889:127.0.0.1:9999 8890:<tfboard_ip+port_running_on>

I think the syntax is

ssh <username>@<ip to machine> -L 8889:127.0.0.1:9999 -L 8890:127.0.0.1:<tfboard_ip+port_running_on>

but as you can see it starts to get pretty long. This is when having a .ssh/config file in your home directory is useful. It sould look like this:

Host badass_name
    Hostname 127.0.0.1
    User username
    LocalForward 8889 127.0.0.1:9999
    LocalForward 8890 127.0.0.1:tboard_port

then you just go ssh badass_name

3 Likes

Ah never thought of doing that … is the badass_ prefix the correct one for fastai? I was going to use bitchin_ but I think that may be a give away that I’m from socal.

Thanks!