Skip to content

How to switch calibration algorithm from 'max' to 'histogram' in INT8_WEIGHT_ONLY_CFG? #791

@Hyubo

Description

@Hyubo

My question:

Hi,I am currently using the following INT8 quantization configuration:

INT8_WEIGHT_ONLY_CFG = { "quant_cfg": { "*weight_quantizer": {"num_bits": 8, "axis": 0}, "*input_quantizer": {"enable": False}, **_default_disabled_quantizer_cfg, }, "algorithm": "max", }

I would like to use the histogram calibration algorithm instead of max. Could you please advise on how to correctly modify this configuration to enable histogram-based calibration?

Specifically, should I simply change the "algorithm" field, or are there additional parameters required within "quant_cfg"?

Thanks!

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions