[Optimization] Qwen2.5-VL support multi-batch prefill

[aquagull]

Motivation

none

💡 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

1. Added conditional conversion of tasks to tensors in zmq->scheduler when FD_ENABLE_E2W_TENSOR_CONVERT is enabled.
2. Updated scheduling logic to allow multiple multimodal prefill requests when FD_ENABLE_MAX_PREFILL is enabled.
3. Modified extract_vision_features_qwen to handle batched image tensors

Usage or Command

export FD_ENABLE_E2W_TENSOR_CONVERT=1
export FD_ENABLE_MAX_PREFILL=1

Accuracy Tests

none

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

This PR adds multi-batch prefill support for Qwen2.5-VL models to improve throughput and efficiency. The implementation introduces conditional logic controlled by environment variables FD_ENABLE_MAX_PREFILL and FD_ENABLE_E2W_TENSOR_CONVERT.

• Enables multi-batch prefill scheduling for multimodal requests when FD_ENABLE_MAX_PREFILL is set
• Adds tensor conversion in the zmq→scheduler pipeline when FD_ENABLE_E2W_TENSOR_CONVERT is enabled
• Updates vision feature extraction to handle batched inputs for Qwen models

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
fastdeploy/worker/gpu_model_runner.py	Modified _apply_mm_inputs to conditionally batch vision inputs and updated extract_vision_features_qwen to handle batched image tensors
fastdeploy/entrypoints/engine_client.py	Added conditional tensor conversion in _send_task before sending tasks via ZMQ
fastdeploy/engine/sched/resource_manager_v1.py	Updated scheduling logic to allow multiple multimodal prefill requests when max prefill is enabled

[codecov-commenter]

Codecov Report

❌ Patch coverage is 16.66667% with 10 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (develop@ce9a49f). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
fastdeploy/worker/gpu_model_runner.py	22.22%	4 Missing and 3 partials ⚠️
fastdeploy/entrypoints/engine_client.py	0.00%	2 Missing ⚠️
fastdeploy/engine/sched/resource_manager_v1.py	0.00%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             develop    #5269   +/-   ##
==========================================
  Coverage           ?   60.12%           
==========================================
  Files              ?      320           
  Lines              ?    39061           
  Branches           ?     5875           
==========================================
  Hits               ?    23487           
  Misses             ?    13718           
  Partials           ?     1856           

Flag	Coverage Δ
GPU	60.12% <16.66%> (?)

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.