See llms.txt for all machine-readable content.

Back to Templates

Run budget-checked scientific GPU experiments with EMT Compute

Created by

Created by: EMT Compute | EMT Aerospace Technologies || emtcompute
EMT Compute | EMT Aerospace Technologies

Last update

Last update 8 hours ago

Categories

Quick overview

Automate scientific GPU experiments with EMT Compute: control rental budgets, run simulations and collect calculation reports. Connect research workflows to a platform spanning guided AI training, one-click model deployment, robotics, drone simulation and GPU clusters.

How it works

  1. Starts manually and loads configuration for the experiment, quote requirements, polling limit, and spend ceiling.
  2. If an existing EMT Compute job ID is provided, it skips quoting and moves directly to monitoring the job.
  3. Otherwise, it requests an EMT Compute experiment quote, validates the quote details/expiry against your requirements and budget, and prepares a launch payload.
  4. If paid execution is explicitly allowed, it starts the EMT Compute experiment job using an Idempotency-Key header; otherwise it stops after outputting the validated quote details.
  5. Polls the EMT Compute job status every 30 seconds until the job reaches a terminal state and resource release is confirmed.
  6. Outputs the job details along with authenticated EMT Compute API URLs for available calculation reports and result artifacts.
  7. If the polling limit is reached before release, it requests job cancellation and fetches the latest job state to indicate whether release still needs attention.

Setup

  1. Create an EMT Compute account at https://emtcompute.com and a scoped API key with read, rent and manage permissions. Use n8n HTTP Header Auth to keep the key out of the workflow JSON.
  2. In n8n, create an HTTP Header Auth credential with header name Authorization and value Bearer <your EMT API key>, and select it in every EMT Compute HTTP request in the workflow.
  3. Start with allowPaidRun set to false. Edit Configuration to choose the experiment, GPU requirements, maxRentalUSD, quoteRequest.maxMinutes and maxPolls. Run once to inspect the quote without renting resources.
  4. For a paid experiment, fund the intended budget, persist a unique idempotencyKey and set allowPaidRun to true. Retain that same key and exact launch payload for retries of the same intended rental. Set existingJob to resume monitoring a known job without another launch.

Requirements

  • Tested with n8n 2.46.5 using built-in nodes only. Requires an EMT Compute account, a scoped API key and sufficient credit for paid GPU use. Check compatibility before using a different n8n release.

Customization

  • The included galaxy-collision experiment provides a starting point for scientific simulation. Adapt the experiment and parameter runs using https://emtcompute.com/integrations. Compatible parameter sweeps can span nodes; custom algorithms require their own distributed implementation. Add authenticated download nodes to store results in your research systems.

Additional info

FROM AN IDEA TO A COMPUTATIONAL EXPERIMENT

EMT Compute, by EMT Aerospace Technologies, brings GPU computing and guided tools into one environment for researchers, AI developers and engineering teams. Use compute on demand to develop models, explore scientific questions and test strategies in simulation.

GUIDED AI TRAINING
Configure datasets, models and training methods; estimate resources and costs; follow progress and keep checkpoints to continue your work.

ONE-CLICK MODEL DEPLOYMENT
Launch supported models on a machine sized by the platform, with an HTTPS endpoint ready to connect to your applications.

ROBOTICS, DRONES AND AUTONOMY
Train and compare control strategies in simulation, including locomotion, manipulation, drone flight and driving. Explore 3D visualizations and export supported policies in ONNX or TorchScript for further development.

SCIENTIFIC EXPERIMENTS AND SIMULATIONS
Start from available experiments or your own code. Run parameter studies and collect inputs, results and artifacts in reports that include the calculation steps recorded by the experiment.

GPU RESOURCES, NODES AND CLUSTERS
Choose resources for compatible workloads and distribute independent experiment runs across nodes. Distributed algorithms require compatible code; computation is not automatically distributed simply by selecting a cluster.

AN ECOSYSTEM THAT CONNECTS TO YOUR TOOLS
Integrate EMT Compute through its APIs, this n8n workflow and its MCP server. The workflow on this page handles scientific experiment quoting, controlled launch, monitoring and result links. The platform also provides the AI training, model deployment and robotics capabilities described above.

Explore the platform: https://emtcompute.com
Connect your research workflow: https://emtcompute.com/integrations
For a laboratory or engineering use case: [email protected]

USAGE AND VALIDATION
The template is free; GPU usage is paid. maxRentalUSD covers GPU rental only; retained storage and other services are separate. An initial quote reserve is not the total cost.

Report URLs require EMT authentication. Custom calculation steps must be recorded by the experiment code. If resource release is pending or an HTTP request fails, inspect the EMT console and confirm release before assuming billing has stopped.

Seven control-flow cases passed in n8n 2.46.5 against a local API fixture: quote-only, authorized launch, excessive budget, expired quote, lost start response, timeout and resume. Live EMT GPU execution and calculation-report validation were tested separately.