[BugFix] fix nvfp4

[lizexu123]

Motivation

修复因为fp4模型多模权重没加载跳过了，导致GLM错误的bug

💡 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.

[paddle-bot]

Thanks for your contribution!

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

root seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
⚠️ Please upload report for BASE (develop@4402396). Learn more about missing BASE report.

Additional details and impacted files

@@            Coverage Diff             @@
##             develop    #7879   +/-   ##
==========================================
  Coverage           ?   63.87%           
==========================================
  Files              ?      462           
  Lines              ?    64487           
  Branches           ?     9886           
==========================================
  Hits               ?    41188           
  Misses             ?    20499           
  Partials           ?     2800           

Flag	Coverage Δ
GPU	73.04% <ø> (?)
XPU	7.11% <ø> (?)

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

☔ View full report in Codecov by Sentry.
📢 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.

[PaddlePaddle-bot]

🤖 Paddle-CI-Agent | ci_status_monitor | 2026-05-21 16:59:00

CI报告基于以下代码生成（30分钟更新一次）:

• PR commit: 68b901d
• Merge base: 4402396 (branch: develop)
• 查看完整 Diff
• CI 详情

---

1 任务总览

所有 Required 任务均已通过，PR 可合并（2 个可选任务失败，不阻塞合并）。

总执行（rerun次数）	总任务	✅ 通过	❌ 失败	⏳ 运行中	⏸️ 等待中	跳过
59(18)	41	38	2	1	0	0

2 任务状态汇总

2.1 Required任务 : 10/10 通过

必选任务阻塞合并，失败需优先处理。

状态	任务	耗时	根因	修复建议	日志	重跑
✅	其余 10 个必选任务通过	-	-	-	-	-

2.2 可选任务 — 28/31 通过

可选任务不阻塞合并，失败仅供参考。

状态	任务	耗时	日志	重跑
❌	Check PR Template	12s	Job	-
❌	Trigger Jenkins for PR	7m54s	Job	-
⏳	CI_HPU	-	Job	-
✅	其余 28 个可选任务通过	-	-	-

---

3 失败详情（仅 required）

无

[PaddlePaddle-bot]

🤖 Paddle-CI-Agent | pr_review | 2026-05-21 15:50:47

📋 Review 摘要

PR 概述：移除 process_weight_transpose 中对权重 _is_initialized() 的检查，修复 fp4 多模权重因被误判为未初始化而跳过 transpose 导致 GLM 推理错误的 Bug。
变更范围：fastdeploy/model_executor/utils.py
影响面 Tag：[Loader] [Quantization]

问题

级别	文件	概述
❓ 疑问	utils.py:135~138（已删除行）	全量删除 _is_initialized() 检查是否会引入新的崩溃风险？
📝 PR 规范	—	Modifications 章节为空；Checklist 第一条应勾选

❓ 疑问详述

原来的逻辑：若 weight._is_initialized() == False，打印 warning 并 return（graceful skip）。此次修复将该检查全量删除，无论初始化状态如何都直接访问 weight.shape。

潜在风险：多模态模型在纯文本推理场景下，某些视觉分支权重可能从未被加载，若这些权重也会走到 process_weight_transpose，删除检查后访问 .shape 可能直接抛异常，比原来的 warning+skip 更难排查。

建议确认：fp4 多模权重 _is_initialized() 返回 False 的根因是什么？除这批权重外，是否还有其他权重也可能经过此函数但真正未初始化？如有，建议改为精准修复：

def process_weight_transpose(layer, weight_name):
    weight = getattr(layer, weight_name)
    if not weight._is_initialized():
        # fp4 packed weights may report uninitialized; check dtype before skipping
        if not hasattr(weight, 'dtype') or weight.dtype is None:
            logger.warning(f"Weight {weight_name} not initialized, skipping transpose.")
            return
    if len(weight.shape) == 2:

或者至少加注释说明为何安全删除该检查。

📝 PR 规范检查

标题格式合规（含 [BugFix] Tag）✓。Modifications 章节内容为空，Checklist 第一条应勾选（PR 已有官方 Tag）。

标题建议（可直接复制）：

• [BugFix] Fix nvfp4 multi-modal weight transpose skip bug in GLM

PR 描述建议（可直接复制，必须复刻 checklist §D2 模板的完整结构）：

## Motivation
修复 fp4 多模态模型中，权重因 `_is_initialized()` 返回 False 被误判为未初始化，导致 `process_weight_transpose` 跳过 transpose 操作，GLM 模型推理结果出现错误的 Bug。

## Modifications
- `fastdeploy/model_executor/utils.py`：移除 `process_weight_transpose` 函数中对 `weight._is_initialized()` 的检查，确保 fp4 格式的多模权重不会被错误地跳过 transpose。

## Usage or Command
N/A

## 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.
- [ ] 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

修复思路直观，删除错误的 early-return 可解决 fp4 多模权重被跳过的问题。建议作者补充说明此修改对其他"真正未初始化"权重场景的影响分析，或添加单元测试覆盖该路径，以减少回归风险。