Help on imatrix quantization algorithm for inference

[TheTinyTeddy]

For a quantized model involving imatrix, does the quantization/de-quantization process explicitly use importance matrix for computation during inference? Is the algorithm for the inference process the same as those without imatrix?

[CISC]

No, imatrix is only used during quantization, it's not used for inference.

[TheTinyTeddy]

So am I right to understand the inference process is the same as those model without imatrix quantization?

[CISC]

Yes, it's the same, it wouldn't even know there was an imatrix applied.