Model distillation

Generate teacher targets and train a smaller student with measured quality and latency.

load teacher→generate soft labels→train student→evaluate qualitybenchmark latency→publish student

Model distillation

Activity responsibilities

Activity	Responsibility
load-teacher	Execute the load teacher stage and publish its declared outputs for downstream activities.
generate-soft-labels	Execute the generate soft labels stage and publish its declared outputs for downstream activities.
train-student	Execute the train student stage and publish its declared outputs for downstream activities.
evaluate-quality	Execute the evaluate quality stage and publish its declared outputs for downstream activities.
benchmark-latency	Execute the benchmark latency stage and publish its declared outputs for downstream activities.
publish-student	Execute the publish student stage and publish its declared outputs for downstream activities.

Inputs

• Teacher model
• Examples
• Student architecture

Outputs

• Soft labels
• Student model
• Quality and latency reports

Provenance to preserve

• Teacher revision
• Generated targets
• Benchmark hardware

Source bundle

The executable workflow and its implementation files are maintained in the separate AkôFlow examples repository. The documentation intentionally keeps code, container recipes, generators, and validation scripts out of the narrative page.

Open the complete source bundle ↗

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

Execution considerations

Cache generated labels and run latency tests on the intended deployment class rather than the training host.

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