Infrastructure selection for training

Compare plans for local GPU, HPC, and cloud before executing the selected candidate.

profile workload→plan local GPUplan HPCplan cloud→compare plans→select scope→train

Infrastructure selection for training

Activity responsibilities

Activity	Responsibility
profile-workload	Execute the profile workload stage and publish its declared outputs for downstream activities.
plan-local-gpu	Execute the plan local GPU stage and publish its declared outputs for downstream activities.
plan-hpc	Execute the plan HPC stage and publish its declared outputs for downstream activities.
plan-cloud	Execute the plan cloud stage and publish its declared outputs for downstream activities.
compare-plans	Execute the compare plans stage and publish its declared outputs for downstream activities.
select-scope	Execute the select scope stage and publish its declared outputs for downstream activities.
train	Execute the train stage and publish its declared outputs for downstream activities.

Inputs

• Workflow profile
• Execution scopes
• Budget and deadline

Outputs

• Candidate plans
• Selection record
• Observed run

Provenance to preserve

• Planner version
• Predictions
• Plan-versus-observed error

Runnable example

This is the complete checked-in bundle for this pattern. Download the environment, scope, topology, workflow, input, container recipe, runner, and validation contract from this page before executing it.

Read setup instructions ↓Environment YAML ↓Execution scope YAML ↓Network topology YAML ↓Workflow YAML ↓Input data ↓Container recipe ↓Run the workflow ↓Validate outputs ↓Expected output manifest ↓

Verified local execution

These captures and the output manifest were produced by the fixture's local Docker run and validator. They are published with the same bundle as the runnable files.

Workflow evidence

Execution evidence

Output evidence

Open verified output manifest ↓

Execution considerations

Evaluate candidates with frozen inputs and one common model. Queue planning experiments so concurrent load does not distort evidence.

AkôFlow boundary: the engine schedules, deploys, executes, transfers data, and records evidence. The ML or agent framework remains an implementation choice inside each activity.