[Loader][BugFix] Fix some parameters place on CPU in PaddleOCR-VL

[SigureMo]

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)

LazyGuard 下的参数是未初始化的，而 uninit_tensor.copy_(data) 结果会是 CPU 的，这导致了 OCR 里的参数加载后基本都是 CPU 的

此前使用了 get_tensor 强制 H2D，因此即便没有初始化仍然正确放在 GPU

但是 #4532 移除了 get_tensor，这就导致这些参数最后都在 CPU，性能惨不忍睹

另外 #4532 影响面不确定，本 PR 只测了 OCR 模型，其他的建议都查查

Modifications

补全 OCR 中缺失的针对未初始化参数的 init 逻辑，并统一使用 h2d_copy

不过这个 h2d_copy 名字感觉不太合适啊，看着像强制 h2d，但是实现并不是

Usage or Command

python -m fastdeploy.entrypoints.openai.api_server \
        --model /root/paddlejob/tmpspace/MODELS/PaddlePaddle/PaddleOCR-VL \
        --port 8295 \
        --metrics-port 8296 \
        --engine-worker-queue-port 8297 \
        --cache-queue-port 55660 \
        --max-model-len 16384 \
        --max-num-batched-tokens 16384 \
        --gpu-memory-utilization 0.7 \
        --max-num-seqs 256 \
        --workers 2 \
        --graph-optimization-config '{"graph_opt_level":0, "use_cudagraph":true}'

Accuracy Tests

无

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!

[Copilot]

Pull request overview

此 PR 修复了 PaddleOCR-VL 模型加载时参数被错误放置在 CPU 上的问题。由于 PR #4532 移除了 get_tensor 的强制 H2D 调用，导致 LazyGuard 下未初始化的参数在使用 copy_ 后仍停留在 CPU，严重影响性能。

主要变更

• 在所有 weight_loader 方法中添加参数初始化检查，确保未初始化的参数先调用 initialize()
• 统一使用 h2d_copy 替代 param.copy_ 以保证正确的设备放置
• 修复了 3 个不同的 weight_loader 方法（SiglipAttention.out_proj_weight_loader、SiglipMLP.weight_loader、Projector.weight_loader）

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
fastdeploy/model_executor/models/paddleocr_vl/siglip.py	为 SiglipAttention 和 SiglipMLP 的 weight_loader 添加参数初始化逻辑，并替换为 h2d_copy；新增 get_tensor 导入但未使用
fastdeploy/model_executor/models/paddleocr_vl/projector.py	为 Projector 的 weight_loader 添加参数初始化逻辑，并替换为 h2d_copy

[Copilot]

The get_tensor import is unused in this file. Since PR #4532 removed the usage of get_tensor and this PR replaces it with h2d_copy (which internally calls get_tensor when needed), this import can be safely removed.

Suggested change

from fastdeploy.model_executor.layers.utils import get_tensor

[codecov-commenter]

Codecov Report

❌ Patch coverage is 70.00000% with 3 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (develop@f6eb4dc). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
...eploy/model_executor/models/paddleocr_vl/siglip.py	66.66%	0 Missing and 2 partials ⚠️
...oy/model_executor/models/paddleocr_vl/projector.py	75.00%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             develop    #5413   +/-   ##
==========================================
  Coverage           ?   58.26%           
==========================================
  Files              ?      327           
  Lines              ?    40566           
  Branches           ?     6157           
==========================================
  Hits               ?    23636           
  Misses             ?    15098           
  Partials           ?     1832           

Flag	Coverage Δ
GPU	58.26% <70.00%> (?)

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.