This will give you step-by-step instructions for starting an Atmosphere instance with the kaldi_instructional resources already loaded.
Go to the cyverse login: https://user.cyverse.org/services/mine and, after logging in, click LAUNCH on Atmosphere.
You'll arrive at your Dashboard. Here you can see information about your allocations (how much you are using and how much you "have left"). Select Launch New Instance.
Enter kaldi into the Image Search, and select the kaldi_instructional_2017 image.
You'll see some details about this image. In the upper right, select Launch.
You'll now be asked to set up the instance.
a. Give an Instance Name.
b. Keep Base Image Version at 1.0.
c. You may need to name the Project (e.g. the collection of images and volumes).
d. Select your Instance Size. The code has been built under the assumption that you will be using medium2. No guarantees can be made that anything smaller will work. And while you're welcome to choose a larger instance, know that your Allocation Units are calculated based on the number of CPUs you are running, so you will likely need to Shutdown your instance when it's not in use to ensure you don't run our of monthly Allocation Units.
Click Launch Instance.
When your instance is built, you will see it shown with a Status of Active. You'll also see an IP Address to access the instance.
It's also recommended that you add a Volume (e.g. external storage drive that is separate from your instance) so that you can occasionally backup the files you build throughout the course. In case of a problem with your instance, you can always start a new instance and transfer those files from the volume. Click New-Volume.
You'll now set up the volume.
a. Give a Volume Name.
b. Select the Volume Size. 100GB should be plenty of space, but you are free to choose whatever size fits your needs and current allocations.
Click Create Volume
You will see that its Status is Unattached. Click on the volume link.
Click Attach.
You'll be asked to select an instance to attach it to.
Now when you return to your Project, you'll see that the Status of your volume is Attached to [instance_name].
The volume will be mounted at /vol_c/ in your instance.
You can now access your image by sshing into it with the following command:
ssh [your_cyverse_id]@[instance_ip]
For example:
ssh mcapizzi@123.145.125.53
However, if you intend to utilize jupyter, you will need to use the following command:
ssh -L [port]:localhost:[port] [your_cyverse_id]@[instance_ip]
The default port that is used by our scripts is 8880, and unless you have a reason to change it, that should be used for simplicity.
The command will then look like:
ssh -L 8880:localhost:8880 mcapizzi@123.145.125.53
If you do not add the -L argument, you will still be able to access your instance but not jupyter.
Note: You'll be asked for a password. This is your cyverse account password.
The kaldi_instructional_2017 image that you used already has the docker container installed along with all dependencies and compiled kaldi code.
However, you will need to do a manual clone of the repository onto your instance and checkout of our branch the first time you use your instance.
cd /scratch # the location of storage on the Atmosphere instances
sudo git clone https://github.com/michaelcapizzi/kaldi.git # clone the repository
cd kaldi # move into the newly cloned directory
sudo git fetch # fetch the other repository branches
sudo git checkout kaldi_instructional # check out the INSTRUCTIONAL branch
Now you can follow the instructions in the main README for Running docker container and Running jupyter.