Using the GUI with Loihi - abr/telluride-nengo GitHub Wiki

For the GUI to run models on Loihi, the server has to run on a machine connected to the hardware. If you're not physically connected to the hardware, or the hardware is connected to a virtual machine, you can set up an SSH tunnel to view the GUI locally despite not being connected to the hardware. Please ensure you have the latest commit on the master branch of nengo-loihi.

Make sure you've installed the GUI (be sure to source activate loihi) using:

cd ~
git clone https://github.com/nengo/nengo_gui
cd nengo_gui
git checkout loihi
pip install -e .
cd ~

Assuming you've already installed the nengo_loihi backend, you need to:

cd ~/nengo_loihi
git checkout gui-integration

In the example below, we'll start the GUI on the super host, port 8888. If the port is being used you can try a different one nearby.

Starting from your local machine, do:

ssh -L 8888:localhost:8888 <username>@192.168.167.211

This sets up the tunneland starts a session on super host. Once you're on super host:

source activate loihi
nengo --port 8888 --no-browser --auto-shutdown 0 --backend nengo_loihi

The GUI server should start on the remote. You can copy and paste the provided URL into your local machine web browser and use the GUI normally.

If you're running on the Intel cloud, please ensure you have the latest commit on the master branch of nengo-loihi.
To do that, update your local repo and rsync again (see here for more info). You then need to install a library:

sudo apt install libc6-dev-i386

Finally, you can execute:

SLURM=1 nengo --port 8888 --no-browser --auto-shutdown 0 --backend nengo_loihi

Then go to your local machine and open a browser using the provided URL after the above command runs. Note that the loihi will only be used for 1.0s of simulated time by default before it exits. If you want more time, you can use max_loihi_time in your script. (e.g. max_loihi_time = 2.0 will run for 2.0 simulated seconds, i.e., 2000 Loihi time steps).