Your promoted model, exported and smoke-tested.
What it does
Export a promoted model for its deployment target and provide a smoke-test comparison of pre-export and post-export outputs.
When to use it
Use after checkpoint promotion, when selecting a quantization format for a target device, or when troubleshooting an exported model's smoke-test failure.
How to use it
Give it a promoted checkpoint or LoRA adapter and the target hardware, serving stack, and workload scope; it returns an exported model and a smoke-test diff report.
What you provide
- target GPU class
- serving stack
- workload types
- Your files
What you need · 3
Requires Python to run the mandatory smoke-test gate; no version is specified.
Requires a PROMOTE verdict from checkpoint-promotion before export can begin.
Access · 3
This skill
vllm
Starts a vLLM serving process
gguf-py
Installs the gguf-py package