New quantization submission

Upload the model and calibration data required for this submission.

Secure by designFiles are used only for this quantization submission.
Model file *
Calibration data *
Product line

The platform applies the compatible processing profile automatically.

Optional — recommended values are preselected. Change them only for a specific target or with guidance from your platform team.

Samples per calibration step. For fixed-shape models, use a value compatible with the model input shape.

Controls validation coverage. More vectors may improve confidence but can increase processing time and generated artifacts.

Recommended: greedy. Use naive mainly when investigating a conversion or memory-planning issue.

preload prepares parameters before execution; oneshot uses the alternate fetch strategy. The trade-off depends on the target runtime.

axi is the default for this product line. Other modes are available only where the selected target supports them.