Model conversion and quantization

Create hardware-specific variants and validate compatibility and quality.

load trained model→export ONNXexport TorchScript→quantize CPUquantize GPU→validate outputs→publish variants

Model conversion and quantization

Activity responsibilities

Activity	Responsibility
load-trained-model	Execute the load trained model stage and publish its declared outputs for downstream activities.
export-onnx	Execute the export ONNX stage and publish its declared outputs for downstream activities.
export-torchscript	Execute the export TorchScript stage and publish its declared outputs for downstream activities.
quantize-cpu	Execute the quantize CPU stage and publish its declared outputs for downstream activities.
quantize-gpu	Execute the quantize GPU stage and publish its declared outputs for downstream activities.
validate-outputs	Execute the validate outputs stage and publish its declared outputs for downstream activities.
publish-variants	Execute the publish variants stage and publish its declared outputs for downstream activities.

Inputs

• Canonical model
• Target formats
• Tolerance policy

Outputs

• Optimized variants
• Compatibility report
• Registry entries

Provenance to preserve

• Converter versions
• Calibration data
• Target hardware

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

Conversion branches can use different images. Validate optimized artifacts on representative target hardware before publication.

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