[Speculative Decoding]Support multi-step mtp with cudagraph

[freeliuzc]

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.

[paddle-bot]

Thanks for your contribution!

[Copilot]

Pull request overview

This pull request adds CUDA graph support for multi-step MTP (Medusa Tree Proposal) in speculative decoding. The changes ensure proper CUDA graph capture behavior when executing multiple model steps during draft token generation.

Key Changes

• Modified _initialize_forward_meta to accept new parameters (is_dummy_run, substep) for controlling CUDA graph capture during multi-step execution
• Updated CUDA graph logic to only capture the first substep during dummy runs (graph warmup phase)
• Replaced paddle.clone() with copy_() for last_seq_lens_this_time to avoid error 700 in CUDA graph mode

[gongshaotian]

这是个新增的模型输入吗

[freeliuzc]

这里不是新增的，多步下一直有，如果用 clone 会奇怪和 cudagraph 耦合

[gongshaotian]

LGTM

[codecov-commenter]

Codecov Report

❌ Patch coverage is 75.00000% with 1 line in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (develop@2785b82). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
fastdeploy/spec_decode/mtp.py	75.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             develop    #5886   +/-   ##
==========================================
  Coverage           ?   65.96%           
==========================================
  Files              ?      347           
  Lines              ?    44410           
  Branches           ?     6818           
==========================================
  Hits               ?    29294           
  Misses             ?    12956           
  Partials           ?     2160           

Flag	Coverage Δ
GPU	65.96% <75.00%> (?)

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.