AutoRound Refactor

[Kaihui-intel]

User description

Type of Change

feature

Description

Directly pass parameters to autoround

Expected Behavior & Potential Risk

the expected behavior that triggered by this PR

How has this PR been tested?

how to reproduce the test (including hardware information)

Dependency Change?

any library dependency introduced or removed

---

PR Type

Enhancement

---

Description

• Refactor get_model_info method to return None

• Update to_config_mapping to directly set configuration

• Add new tests for static AFP8 export

---

Diagram Walkthrough

flowchart LR
  A["Refactor get_model_info"] -- "Return None" --> B["Update to_config_mapping"]
  B -- "Set configuration directly" --> C["Add new tests"]

Loading

File Walkthrough

	Relevant files
Enhancement	config.py Refactor model info and config mapping                                      neural_compressor/torch/quantization/config.py Changed get_model_info to return None Modified to_config_mapping to set configuration directly +3/-11

---

[PRAgent4INC]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 3 🔵🔵🔵⚪⚪

🧪 PR contains tests

🔒 No security concerns identified

⚡ Recommended focus areas for review Possible Issue The get_model_info method now returns None, which may break any code relying on the previous behavior of returning a list of tuples containing the name and type of each module in the model. return None Logic Change The to_config_mapping method no longer filters modules based on WOQ_WHITE_LIST and logs the model info. This change may affect the configuration mapping logic. config_mapping = OrderedDict({self.name: self})

[PRAgent4INC]

PR Code Suggestions ✨

[xin3he]

@Kaihui-intel Please help move the autoround.py from weight_only to a separate folder as we discussed before.