HAAG HPC Training

Implementation and Testing Record

Procedure implemented

No procedure received a formal implementation or broader HAAG rollout during Summer 2026. The team internally self-tested parts of the authored curriculum, most notably Procedure 205 - Submitting Your First Job, and used that run to create its Scribe walkthrough and command reference. This was author verification, not independent learner validation.

Context and participants

Implementation approach

The planned implementation was a structured pilot of the 100-300 learning path. Instead, the team:

  1. organized the repository and documentation formats while waiting for ICE access;
  2. completed the core authored materials, prioritizing Modules 200 and 300;
  3. ran internal author checks, including the first-job submission workflow;
  4. corrected technical guidance against dated official PACE sources; and
  5. deferred independent testing, broader rollout, the Microsoft Forms workflows, and Module 400.

No adaptations were observed in learner use because no independent learner executed the procedures.

Outcomes and evidence

Evidence available this semester is limited to the repository artifacts, the working author-tested Procedure 205 example, and the team’s self-testing observations. The team did not collect independent completion rates, elapsed time, learner questions, error counts, adoption, support-request reduction, or broader HAAG feedback. No claim of learner effectiveness or time saved can therefore be made.

Current status: prepared for structured learner testing, not implemented across HAAG.

Effectiveness

Internal testing showed that the authors could execute the documented first-job workflow and generate the expected output. It also helped identify technical details that needed correction. However, authors already knew the intended path, so this evidence cannot show whether a beginner will understand the vocabulary, navigation, resource choices, or missing assumptions. The original goal - independent onboarding and first-job completion with fewer repeated questions - remains unvalidated.

Challenges and limitations

Next steps

1. Polish and review

Finish the remaining Module 100 review, verify all current claims against official PACE sources, and obtain subject-matter review where the team is not authoritative.

2. Pilot with 2-3 new researchers

Select researchers who have never used ICE. Confirm that credentials are provisioned before timing begins. Ask each person to follow the 100-300 path without coaching unless safety or policy requires intervention.

Record:

The presentation discussion suggested testing whether the 100-series can be completed in under 45 minutes and whether a credentialed learner can reach a first job in roughly an hour. These are pilot hypotheses, not current performance claims.

3. Create an HPC Help & Guide Feedback form

Provide two routes: ask a PACE/ICE onboarding question, or report broken, unclear, outdated, or incorrect training content. Capture module/page, issue description, screenshot, urgency, contact, assigned owner, response status, and repository correction.

4. Create a Research Compute Needs Intake form

Collect team, project, expected users, workload type, software dependencies, CPU, GPU/VRAM, RAM, storage, duration, frequency, data sensitivity, transfer, and retention needs before cluster access.

5. Revise and expand adoption

Revise the procedures from pilot evidence, repeat any failed scenarios, and only then distribute the curriculum through HAAG project leads and onboarding channels. Track repeated support questions before and after distribution where feasible.

6. Build Module 400 carefully

Add specialized topics only after each has an owner, a domain/PACE reviewer, a reproducible example, and a learner-validation plan. The initial backlog is Ollama, staying current, distributed training, and bioinformatics; other tracks such as computer vision may be added when demand is demonstrated.

Contributors