TACC Analysis Portal Pilot - Student Guide
WARNING

The iSchool Knowledge Base has moved to SharePoint. Please visit and bookmark the new iSchool Knowledge Base site at: https://utexas.sharepoint.com/sites/ischool-knowledge. This information is no longer being updated and will soon be deleted.

TACC Analysis Portal Pilot - Student Guide

Getting started with Jupyter at TACC in your iSchool class

Request your TACC Account

  1. Create a TACC account with your @my.utexas.edu email address: https://accounts.tacc.utexas.edu/register

  2. Email your professor your new TACC username, so they can grant you access.

  3. Wait for TACC staff to approve your account request. This typically occurs within 24 hours. You will receive an email notification of approval.

  4. Setup Multi-Factor Authentication for your new TACC account.

    1. Sign in with your TACC account at https://www.tacc.utexas.edu/portal/account

    2. Click “Pair Device” under MFA Devices.

    3. Follow the instructions to pair an authenticator app of your choice

    4. Detailed instructions are available here: https://docs.tacc.utexas.edu/basics/mfa/#setupmfa-step2

Using the Portal

TACC is a shared High Performance Compute Cluster. There will be times when the resource you want to use is fully occupied, and you will have to wait for an available session.

Next time, you can Resubmit a past job

  1. Navigate to https://tap.tacc.utexas.edu/mfalogin/

  2. Login with your TACC username, password, and the token from your MFA app

  3. Don’t be intimidated by the long form. After filling it out once, you can resubmit the same settings.

    1. For standard compute use these settings:

      1. System: Lonestar6

      2. Application: Jupyter Notebook

      3. Project: <select the project code shared by your instructor in the drop down>

      4. Queue: vm-small

      5. Job Name: Jupyter Standard

      6. Time Limit: 1:0:0

      7. Reservation: <Leave Blank>

    2. For GPU compute, use these settings:

      1. System: Lonestar6

      2. Application: Jupyter Notebook

      3. Project: <select the project code shared by your instructor in the drop down>

      4. Queue: gpu-a100-small

      5. Job Name: Jupyter GPU

      6. Time Limit: 1:0:0

      7. Reservation: <Leave Blank>

  4. Hit Submit and wait for the next screen. It can sometimes take up to 60 seconds

  5. On the Job Status screen, click “Connect”

    1. If you don’t see Connect, wait 30 seconds and click “Check Status” till you see Connect

  6. If your class is using huggingface (your professor will let you know), run the following code the first time you connect to your notebook

!mkdir -p $WORK/.cache !ln -s $WORK/.cache ~/.cache !ln -s $WORK ~/WORK !ln -s $SCRATCH ~/SCRATCH

 

FAQ