Governed Inference Meter
Inspect inference metering through the retained kernel and its active Energy Attest source package.
Artifact: Inference-metering software kernel · Stage: Retained compatibility contract
Explore in Command Lab · Build · Evidence
Before you use it
- Use the exact Kernel Hub revision in the retained contract below. A card update does not change that code pin.
- Loading the kernel executes repository code and requires explicit trust. Inspect the immutable revision before enabling
trust_remote_code=True. - Existing builds and qualification evidence stay unchanged; this presentation adds no runtime, energy or performance measurement.
Technical details and original loading contract
The complete original source card follows. Its immutable loading pin and explicit trust requirement remain in force.
Governed inference meter — immutable loading contract
This card is the canonical loading contract for the retained Hugging Face
compatibility kernel. The implementation is maintained in
szl_energy_attest.inference_meter.
Quickstart
from kernels import get_kernel
gim = get_kernel(
"SZLHOLDINGS/governed-inference-meter",
revision="6d546bfc6591b44ae5eb57d1209678454e52a3f5",
trust_remote_code=True,
)
print(gim.selfcheck())
The full commit pin makes the loaded code immutable. The explicit trust flag is required because loading a remote organization kernel executes repository code. Advance the pin only after a new Kernel Hub revision is published and verified.
- Downloads last month
- -
kernel
szl-holdings
apache-2.0