[XPU] opt FD_XPU_MOE_FFN_QUANT_TYPE_MAP

[zhupengyang]

Motivation

💡 If this PR is a Cherry Pick, the PR title needs to follow the format by adding the [Cherry-Pick] label at the very beginning and appending the original PR ID at the end. For example, [Cherry-Pick][CI] Add check trigger and logic(#5191)

💡 如若此PR是Cherry Pick，PR标题需遵循格式，在最开始加上[Cherry-Pick]标签，以及最后面加上原PR ID，例如[Cherry-Pick][CI] Add check trigger and logic(#5191)

Modifications

Usage or Command

Accuracy Tests

Checklist

• Add at least a tag in the PR title.
  • Tag list: [[FDConfig],[APIServer],[Engine], [Scheduler], [PD Disaggregation], [Executor], [Graph Optimization], [Speculative Decoding], [RL], [Models], [Quantization], [Loader], [OP], [KVCache], [DataProcessor], [BugFix], [Docs], [CI], [Optimization], [Feature], [Benchmark], [Others], [XPU], [HPU], [GCU], [DCU], [Iluvatar], [Metax]]
  • You can add new tags based on the PR content, but the semantics must be clear.
• Format your code, run pre-commit before commit.
• Add unit tests. Please write the reason in this PR if no unit tests.
• Provide accuracy results.
• If the current PR is submitting to the release branch, make sure the PR has been submitted to the develop branch, then cherry-pick it to the release branch with the [Cherry-Pick] PR tag.

[PaddlePaddle-bot]

🤖 Paddle-CI-Agent | pr_review | 2026-06-05 11:19:11

📋 Review 摘要

PR 概述：修复 _set_xpu_moe_quant_type 中对 FD_XPU_MOE_FFN_QUANT_TYPE_MAP 的解析，新增对无冒号全局默认量化类型格式的支持，并修正注释中的环境变量名称。
变更范围：fastdeploy/model_executor/layers/backends/xpu/moe/fused_moe.py
影响面 Tag：[XPU] [OP]

问题

级别	文件	概述
❓ 疑问	fused_moe.py:102-103	全局默认条目放在非首位时会静默覆盖前面所有范围条目

❓ 疑问 详情（_set_xpu_moe_quant_type，新增 len == 1 分支）

当 FD_XPU_MOE_FFN_QUANT_TYPE_MAP 中全局默认条目不在首位（如 "w_channelwise_int8_a_tokenwise_float16:3->5;w_bfloat16_a_bfloat16"）时，循环末轮会无条件执行 self.xpu_moe_quant_type = "w_bfloat16_a_bfloat16"，把之前对层 3-5 的范围设置全部静默覆盖。注释已说明「should be placed first」，但代码未产生任何警告，用户误操作时难以排查。

建议在检测到非首位全局默认条目时输出 warning 日志，示例：

if len(quant_type_info) == 1:
    if quant_type_map != xpu_moe_ffn_quant_type_map.split(";")[0]:
        logger.warning(
            f"FD_XPU_MOE_FFN_QUANT_TYPE_MAP: global default entry "
            f"'{quant_type_info[0]}' is not placed first and will "
            f"override previous range-specific entries."
        )
    self.xpu_moe_quant_type = quant_type_info[0]

历史 Findings 修复情况

Finding	问题	状态
F1（上次 Review）	PR 描述模板各段为空，Checklist 未勾选	⚠️ 仍存在

📝 PR 规范检查

PR 标题包含官方 Tag [XPU] ✓，但描述模板各段（Motivation / Modifications / Usage or Command / Accuracy Tests）均为空，Checklist 未勾选。

标题建议（可直接复制）：

• [XPU][BugFix] Fix empty-string edge case in FD_XPU_MOE_FFN_QUANT_TYPE_MAP parsing

PR 描述建议（点击展开，可直接复制）

## Motivation
`_set_xpu_moe_quant_type` 在解析 `FD_XPU_MOE_FFN_QUANT_TYPE_MAP` 时，旧逻辑（`if len != 2: continue`）无法支持不含层索引范围的全局量化类型格式（如 `FD_XPU_MOE_FFN_QUANT_TYPE_MAP="w_bfloat16_a_bfloat16"`）。本 PR 新增对此格式的支持，并同步将注释中错误的变量名 `XPU_MOE_FFN_QUANT_TYPE_MAP` 修正为 `FD_XPU_MOE_FFN_QUANT_TYPE_MAP`。

## Modifications
- `fastdeploy/model_executor/layers/backends/xpu/moe/fused_moe.py`：
  - 修正 docstring 中环境变量名称（`XPU_MOE_FFN_QUANT_TYPE_MAP` → `FD_XPU_MOE_FFN_QUANT_TYPE_MAP`）
  - `_set_xpu_moe_quant_type`：新增 `len(quant_type_info) == 1` 分支，支持不含 `:` 分隔符的全局量化类型格式

## Usage or Command
```bash
# 新格式：对所有层统一设置量化类型（无需指定层索引范围）
export FD_XPU_MOE_FFN_QUANT_TYPE_MAP="w_bfloat16_a_bfloat16"

# 原格式（层范围）仍然兼容
export FD_XPU_MOE_FFN_QUANT_TYPE_MAP="w_channelwise_int8_a_tokenwise_float16:3->5,7->9"
```

## Accuracy Tests
N/A

## Checklist

- [x] Add at least a tag in the PR title.
  - Tag list: [`[FDConfig]`,`[APIServer]`,`[Engine]`, `[Scheduler]`, `[PD Disaggregation]`, `[Executor]`, `[Graph Optimization]`, `[Speculative Decoding]`, `[RL]`, `[Models]`, `[Quantization]`, `[Loader]`, `[OP]`, `[KVCache]`, `[DataProcessor]`, `[BugFix]`, `[Docs]`, `[CI]`, `[Optimization]`, `[Feature]`, `[Benchmark]`, `[Others]`, `[XPU]`, `[HPU]`, `[GCU]`, `[DCU]`, `[Iluvatar]`, `[Metax]`]
  - You can add new tags based on the PR content, but the semantics must be clear.
- [x] Format your code, run `pre-commit` before commit.
- [ ] Add unit tests. Please write the reason in this PR if no unit tests.
- [ ] Provide accuracy results.
- [ ] If the current PR is submitting to the `release` branch, make sure the PR has been submitted to the `develop` branch, then cherry-pick it to the `release` branch with the `[Cherry-Pick]` PR tag.

总体评价

变更逻辑正确，成功修复了旧代码无法支持全局默认量化类型的问题，同时新增的空字符串守卫（if not quant_type_map: continue）也处理了 "" 环境变量边界情况。建议对全局默认条目非首位的场景补充 warning 日志，以提升可调试性；PR 描述仍需完善。

[codecov-commenter]

Codecov Report

❌ Patch coverage is 0% with 10 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (develop@fab344e). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
...odel_executor/layers/backends/xpu/moe/fused_moe.py	0.00%	10 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             develop    #8008   +/-   ##
==========================================
  Coverage           ?   66.38%           
==========================================
  Files              ?      468           
  Lines              ?    65927           
  Branches           ?    10168           
==========================================
  Hits               ?    43764           
  Misses             ?    19385           
  Partials           ?     2778           

Flag	Coverage Δ
GPU	76.31% <ø> (?)
XPU	7.02% <0.00%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[hong19860320]

LGTM

[paddle-bot]

Thanks for your contribution!