Magic Castle — Enabling Scalable HPC Training through Scalable Supporting Infrastructures
Bibliographic record
Abstract
The potential HPC community grows ever wider as methodologies such as AI and big data analytics push the computational needs of more and more researchers into the HPC space.As a result, requirements for training are exploding as HPC adoption continues to gather pace.However, the number of topics that can be thoroughly addressed without providing access to actual HPC resources is very limited, even at the introductory level.In cases where access to production HPC resources is available, security concerns and the typical overhead of arranging for account provision and training reservations make the scalability of this approach challenging.Magic Castle aims to recreate the supercomputer user experience in public or private clouds.To define the virtual machines, volumes, and networks that are required in a cloud-provider agnostic way, it uses the open-source software Terraform and HashiCorp Language (HCL).These resources are then configured using the configuration management and deployment tool Puppet to replicate a virtual HPC infrastructure with a full scientific software stack, and including a feature-rich JupyterHub environment.The final resource is accessible both through a web browser and via SSH, making it trivially OS-agnostic for the trainees.Through the use of Magic Castle, we demonstrate that it is possible to dynamically provision virtual HPC system(s) in public or private cloud infrastructure easily, quickly, and cheaply.We also show that such infrastructures can support accelerators and fast interconnects, meaning that they can still be considered "true" HPC resources.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.004 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.005 |
| Open science | 0.003 | 0.008 |
| Research integrity | 0.001 | 0.003 |
| Insufficient payload (model declined to judge) | 0.026 | 0.008 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".