Model conversion and quantization
Create hardware-specific variants and validate compatibility and quality.
Import this workflow from AkôFlow Hub
Open the public README and measured execution results. The Hub includes the execution capture, activity metrics, logs and version history. Reading public results does not require an account.
In Desktop, choose Create workflow → Import from Hub and paste this version-specific URI:
akoflow://hub/publications/0b26ad29-87a9-457e-bc7c-d81ae79d451f/versions/1
Download the exact workflow YAML (v1). Configure the runtime prerequisites below before planning and executing it.
Run this example locally
Export a learned affine model as real ONNX and TorchScript files, generate int8 and IEEE float16 coefficient representations in parallel, reload both executable formats, and compare their numeric outputs against the original model. The float16 branch runs on CPU and does not build or claim a TensorRT engine.
Install Go and Python 3.12 or newer.
git clone https://github.com/UFFeScience/akoflow-examples.git &&
cd akoflow-examples &&
git checkout --detach d890c3171e3661890305b763b8fc3ede48bb61b4 &&
cd machine-learning/model-conversion &&
./prepare.sh &&
./run.sh
Run the block from the directory where you keep your repositories. If you already cloned the repository, enter this example's directory and run ./prepare.sh && ./run.sh.
Preparation downloads the Go module and any declared Python packages once. The workload then runs locally with offline input data. Every meaningful step is a separate process; independent activities run in parallel after their dependencies complete.
This example installs its Python packages in a dedicated .venv: numpy==2.5.4, onnx==1.23.2, torch==2.14.1.
Read outputs/manifest.json for the generated artifacts, outputs/dag-validation.json for dependency waves, and outputs/stage-<activity>.json for each activity's observed result. A failed stage stops the run and prevents a successful final manifest.
Expected outputs: model.onnx, model.torchscript.pt, model-int8.json, model-fp16.bin, validation.json.
Run the same activities in Desktop
Start your local AkôFlow daemon with Docker available and configure AKOFLOW_API_TOKEN in your shell. From this example's directory, run:
go run ../../tools/daemon-local -root ../.. -example machine-learning/model-conversion
This command validates the daemon connection, builds the declared image, registers the local environment and workflow, and submits a real run. The daemon owns execution and file transfers. Open the printed run ID in Desktop to inspect each activity.
Activity responsibilities
| Activity | Responsibility |
|---|---|
load-model | Execute the load-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
- model.onnx
- model.torchscript.pt
- model-int8.json
- model-fp16.bin
- validation.json
Provenance to preserve
- Converter versions
- Calibration data
- Target 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 ↗Execution evidence
Verified on local Docker: 7/7 activities completed. These captures belong to run showcase-model-conversion-local-f22e706408-run-20261020t191128.224915000.
Read the daemon execution record · Local output manifest
Each stage also ran through the standalone local command. Read the dependency and artifact validation.
load-model
After: workflow start.
export-onnx
After: load-model.
export-torchscript
After: load-model.
quantize-cpu
After: export-onnx, export-torchscript.
quantize-gpu
After: export-onnx, export-torchscript.
validate-outputs
After: quantize-cpu, quantize-gpu.
publish-variants
After: validate-outputs.
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.