Skip to main content

Troubleshooting

Back to top

Technical Requirments

Verifying technical requirements

All you need to take a DLI course is a web browser. DLI currently supports the most recent versions of Firefox or Chrome.

DLI courses offer the exciting oppurtunity to interact directly with a GPU for accelerated computing, deep learning, and more. To accomplish this many of the courses rely on a python technology called Jupyter Notebooks.

In order for Jupyter Notebooks to run properly you need to ensure that websockets are enabled on your network over port 80.

To test for websockets, go to websocketstest.com and verify the WebSockets (Port 80) has four green checkmarks.

If websockets do not work, try disabling your antivirus software, using a different web browser, disabling any privacy browser plugins, or using your browser in private browsing mode. VPNs have often been known to interfere with websockets. If you are still having issues try disabling/enabling your VPN.

Please note, that certain public WiFI access spots, companies, or institutions may have firewall rules in place that block websockets.

Instructor-Led Events

Event codes and where to get them

An event code is string that is used during Instructor-Led workshops and Instructor-Led conference sessions. Your instructor will give you this code at the beginning of your event. Entering an event code on the event page will enroll you into the course being taught by your instructor. Before using an event code you must first create an account and log in.

Event code not working

Event codes are only valid until the end of an Instructor-Led event. If your event has ended the code will no longer be valid. If the event has not ended ensure there are no trailing/leading spaces and verify you are using proper capitilzation.

An invalid event code may bring you to an error page.

Starting or continuing courses after entering event code

After entering your event code you may be brought to the login page. You must be logged in before using the event code.

After you are logged in entering an event code should bring you to your dashboard. This dashboard will display all the courses you are currently enrolled in. Once you are enrolled in a course you retain access to the materials. After the Instructor-Led event you can come back to your dashboard at any time and continue or retake the courses.

Back to top

Jupyter Notebook Issues

Jupyter is not responding and cells are not running

This will happen for a few reason. The first reason is because the instance may be dead. Return to the page where you launched the task and verify the counter has not reached zero. If it has you will need to restart the task.

This can also happen if your browser is having issues and does not support Jupyter. Certain versions of browsers can cause incompatibilities, try switching between Firefox and Chrome. If changing the web browser does not work verify that you do not have any privacy or security plugins that are blocking websockets or scripts. Look at the url bar for any red X you can click to allow scripts, open your browser in private or incognito mode, and disable any non-standard plugins that you have installed.

If the issue still persists it could be a networking firewall issue. Jupyter requires websockets to be enabled over port 80 which can be verified here. If that tests passes please reach out to dli support for additional assistance along with a screenshot of the report and the Jupyter Notebook. If that report fails try disabling the firewall on your computer.

If you have disabled firewalls on your computer and are still having issues, there may be a firewall on your network that is causing issues. Try using a VPN or disabling the VPN you are using. Try switching to a different wireless connection or leaving your office. Many corporate or institutional networks block websockets for security reasons, so you may have to contact your local IT.

Jupyter keeps crashing

Sometimes when you are practicing deep learning you may accidentally adjust parameters that cause a GPU or system to run out of memory or other resources. If you are experiencing frequent crashes in Jupyter after modifying some parameters the system may be running out of memory, try switching the parameters back.

Back to top

Labs Not Starting

Start button keeps spinning

Some tasks can take up to 15 minutes to load. If you click start and it takes more than 20 minutes try clicking the stop button and restarting the task. If this issues persists more than once and you have waited 20 minutes please contact dli support for assistance with a link to the task and a screen shot.

Page does not load after clicking launch button

If this occurs try refreshing the page, clicking start again, and then clicking the launch button. If the page still fails to load after 10 seconds try stopping and restarting the task.

Page loads with 502 error

A small number of GPU tasks may launch with a 502 error due to intermittent networking issues. If this occurs simply stop and restart the task.

Back to top

Login Issues

No username field on login page

If you find yourself on a page where the buttons looks different or there is no place to enter a username you may have accidentally landed on our legacy login page. Please click the login or create account buttons on the top of the page and continue from there.

Cannot create account

If you cannot create an account it is possible you have already registered your email with NVIDIA. Try resetting your password or creating an account with a different email.

No confirmation emails

If you have not recieved a confirmation email after 15 minutes, please try again. If the issue persists verify that you are using the correct email and that you have previously created an NVIDIA developer account. This may be different from the NVIDIA account you have created for other NVIDIA services.

Cannot log in

If you were previously logged in and attempting to log back in results in an error, there may be a cookie issue. Try vistiging the logout page and then attempting logging in again.

If that does not resolve your problem, there may be an SSO issue. Try clicking the first create account issue and then logging in one the second page. If that login also fails try creating a new account with your email address to verify you are using the correct address.

If you have entered your password incorrectly three times your account will be locked. You can either wait 1 hour for your account to be unlocked or reset your password.

Back to top

Common Deep Learning & Accelerated Computing Questions

DIGITS is not reading my data or model directory correctly

The model directory and the data directory paths both look similar. Make sure you are copying the right path to the correlating variable name. When you are in DIGITS double check whether you are in the data tab or the models tab.

DIGITS is asking for a username

DIGITS is a tool for collaborative deep learning work. It makes it easy for teams to work together and track who is creating which models or importing which datasets. Because we are using DIGITS here as a teaching tool you will have a dedicated instance and the username does not matter. Simply enter your first name.

Back to top