add support for scheme FP8_STATIC to export llm_compressor format

[n1ck-guo]

• export llm_compressor format
• export auto_round:llm_compressor format
• extract save func from all export files, save in export/utils.py, rename to save_model

[Copilot]

Pull Request Overview

This PR adds support for the FP8_STATIC quantization scheme to export models in the llm_compressor format. The change enables static FP8 weight and activation quantization with specific configurations for compressed-tensors compatibility.

Key Changes

• Adds FP8_STATIC scheme detection and format conversion to llm_compressor
• Implements static FP8 quantization export with compressed-tensors configuration
• Consolidates common save functionality across export modules

Reviewed Changes

Copilot reviewed 10 out of 10 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
auto_round/utils.py	Modified is_static_wfp8afp8 to accept string parameters for format detection
auto_round/export/utils.py	Added shared save function to reduce code duplication across export modules
auto_round/export/export_to_llmcompressor/export_to_static_fp.py	New module implementing FP8_STATIC export with compressed-tensors configuration
auto_round/export/export_to_llmcompressor/export.py	Added FP8_STATIC support to the main export dispatcher
auto_round/autoround.py	Added FP8_STATIC format detection and validation logic
auto_round/export/export_to_awq/export.py	Refactored to use shared save function
auto_round/export/export_to_autoround/export_to_fp8.py	Renamed class and refactored to use shared save function
auto_round/export/export_to_autoround/export.py	Refactored to use shared save function
auto_round/export/export_to_autogptq/export.py	Refactored to use shared save function
test/test_cpu/test_llmcompressor.py	Added test case for FP8_STATIC export validation

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[wenhuach21]

Support it in the AutoRound format as well, and add nvfp4/fp8_static support on the vLLM side later.