rename scheme INT8_W8A8 to INT8

[thuang6]

Description

Rename scheme INT8_W8A8 to INT8, including code, test, doc.

Use INT8 in doc, but also keep backward compatibility of using old name.

Type of Change

• Bug fix
• New feature
• Documentation update
• Performance improvement
• Code refactoring
• Other (please specify):

Related Issues

Fixes or relates to #1674

Checklist Before Submitting

• My code has been tested locally.
• Documentation has been updated as needed.
• New or updated tests are included where applicable.

[azure-pipelines]

Azure Pipelines:
Successfully started running 6 pipeline(s).
1 pipeline(s) require an authorized user to comment /azp run to run.

[azure-pipelines]

Azure Pipelines:
Successfully started running 6 pipeline(s).
1 pipeline(s) require an authorized user to comment /azp run to run.

[azure-pipelines]

Azure Pipelines:
Successfully started running 6 pipeline(s).
1 pipeline(s) require an authorized user to comment /azp run to run.

[Copilot]

Pull request overview

Renames the quantization scheme name from INT8_W8A8 to INT8 across code, tests, and documentation, while attempting to keep backward compatibility for the old scheme name.

Changes:

• Renamed the preset scheme identifier to INT8 and mapped INT8_W8A8 to the same preset for backward compatibility.
• Updated export-format handling and tests to use INT8, plus added a regression test ensuring INT8_W8A8 still works.
• Updated README documentation (EN + CN) to display INT8 instead of INT8_W8A8.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
test/test_cpu/export/test_export.py	Updates tests to use INT8 and adds explicit backward-compat coverage for INT8_W8A8.
auto_round/schemes.py	Renames the preset scheme object to INT8 and maps old preset name to the new scheme.
auto_round/formats.py	Renames the export format enum member to INT8 while keeping the underlying backend token int8_w8a8; updates llm_compressor export selection.
README.md	Updates documented supported schemes to show INT8.
README_CN.md	Mirrors README scheme rename to Chinese documentation.

Comments suppressed due to low confidence (2)

auto_round/formats.py:392

• In the is_dynamic_wint8aint8 branch, the backend is constructed with AutoRoundExportFormat.INT8.name and then output_format is manually overwritten to use .value. This indirection is compensating for the enum name/value mismatch and makes the flow harder to reason about. If you introduce an enum alias for INT8_W8A8 (or accept values in the validation), you should be able to pass the desired backend token directly and avoid mutating output_format after construction.

                    )
                    ar.act_group_size = 0

auto_round/schemes.py:313

• INT8_W8A8 was renamed to INT8, but the module-level constant name is removed entirely. If any downstream code imports INT8_W8A8 from auto_round.schemes, this becomes a breaking change even though preset-name parsing remains backward compatible. Consider keeping INT8_W8A8 = INT8 as a deprecated alias to preserve API compatibility while encouraging the new name.

INT8 = QuantizationScheme.from_dict(
    {
        "bits": 8,
        "group_size": -1,
        "data_type": "int",
        "sym": True,
        "act_bits": 8,
        "act_group_size": -1,
        "act_data_type": "int",
        "act_dynamic": True,
        "act_sym": True,
    }
)

[wenhuach21]

dupliacated?

[thuang6]

This is kept for backward compatibility since it was already in previous releases

[wenhuach21]

should change the value to INT8 as well?

[wenhuach21]

it’s probably better to deprecate it. If a user provides int8_w8a8, we can trigger a warning and automatically map it to INT8 instead

[thuang6]

make sense, updated with deprecated warning. INT8_W8A8 has mapped to INT8 in PRESET_SCHEME. Here value is used to map with compressed_tensor since its format uses this name

[azure-pipelines]

Azure Pipelines:
Successfully started running 6 pipeline(s).
1 pipeline(s) require an authorized user to comment /azp run to run.