support fleet-gqa-latent

[chang-wenbin]

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!

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
⚠️ Please upload report for BASE (develop@60e6223). Learn more about missing BASE report.

Additional details and impacted files

@@            Coverage Diff             @@
##             develop    #7955   +/-   ##
==========================================
  Coverage           ?   67.38%           
==========================================
  Files              ?      467           
  Lines              ?    65183           
  Branches           ?    10007           
==========================================
  Hits               ?    43921           
  Misses             ?    18443           
  Partials           ?     2819           

Flag	Coverage Δ
GPU	77.60% <100.00%> (?)
XPU	7.08% <0.00%> (?)

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.

[PaddlePaddle-bot]

🤖 Paddle-CI-Agent | ci_status_monitor | 2026-05-29 02:30:31

CI报告基于以下代码生成（30分钟更新一次）:

• PR commit: 5cf4206
• Merge base: 60e6223 (branch: develop)
• 查看完整 Diff
• CI 详情

---

1 任务总览

Required 任务 9/10 通过，当前有 1 个 required 失败任务，会阻塞合并；无 required 运行中/等待中。Optional 任务存在失败/等待/取消，仅供参考。

总执行（rerun次数）	总任务	✅ 通过	❌ 失败	⏳ 运行中	⏸️ 等待中	跳过	取消
41(0)	41	36	3	0	1	0	1

2 任务状态汇总

2.1 Required任务 : 9/10 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
❌	Run FastDeploy Unit Tests and Coverage / run_tests_with_coverage	1h26m	PR问题：新增 rotary_dim 计算未兼容 Mock 配置	兼容缺失 head_dim 或补 Mock 字段	Job	-
✅	其余 9 个必选任务通过	-	-	-	-	-

2.2 可选任务 — 27/31 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	22s	Job	-
❌	Trigger Jenkins for PR	17s	Job	-
⏸️	CI_HPU	-	Workflow	-
🚫	Run iluvatar Tests / run_iluvatar_cases	-	Workflow	-
✅	其余 27 个可选任务通过	-	-	-

---

3 失败详情（仅 required）

Run FastDeploy Unit Tests and Coverage / run_tests_with_coverage — 测试失败（置信度: 高）

Run FastDeploy Unit Tests and Coverage / run_tests_with_coverage

• 状态: ❌ 失败
• 错误类型: 测试失败
• 置信度: 高
• 根因摘要: 新增 rotary_dim 计算未兼容 Mock 配置
• 分析器: ci_analyze_unittest_fastdeploy

失败用例:

测试	错误	根因
tests/worker/test_gpu_model_runner.py::TestProcessMMFeatures::*	TypeError: unsupported operand type(s) for *: 'Mock' and 'Mock'	7 个用例在 setUp() 创建 InputBatch(Mock) 时触发新增 rotary_dim 计算，rotary_percent/head_dim 都是 Mock
tests/worker/test_gpu_prompt_logprobs.py::TestGPUPromptLogprobs::test_prompt_logprobs	AttributeError: 'ModelConfig' object has no attribute 'head_dim'	测试的 MockConfig.ModelConfig 未提供 head_dim，新增代码无条件访问该字段
tests/e2e/test_Qwen2_5_VL_serving.py::test_profile_reset_block_num	Failed: 日志中没有Reset block num信息	config.log 中未找到 Reset block num，与本 PR 的 RoPE 维度修改暂无直接证据关联，建议修复主因后复查

根因详情:
本 PR 在 InputBatch.__init__ 中新增：rotary_percent = getattr(self.model_config, "rotary_percent", 1) 和 self.rotary_dim = int(rotary_percent * self.model_config.head_dim)，并将 get_rope(rotary_dim=...) 改为使用 self.rotary_dim。但现有 worker 单测使用精简 Mock/MockConfig 构造 InputBatch，其中 rotary_percent 可能是 Mock、head_dim 可能缺失；因此对象初始化阶段即失败，导致多个用例无法进入原本被测逻辑。ProposerInputBatch 中也有同样新增计算，建议一并处理。

关键日志:

fastdeploy/worker/input_batch.py:113
> self.rotary_dim = int(rotary_percent * self.model_config.head_dim)
E TypeError: unsupported operand type(s) for *: 'Mock' and 'Mock'

fastdeploy/worker/input_batch.py:113
> self.rotary_dim = int(rotary_percent * self.model_config.head_dim)
E AttributeError: 'ModelConfig' object has no attribute 'head_dim'

FAILED tests/worker/test_gpu_model_runner.py::TestProcessMMFeatures::* - TypeError
FAILED tests/worker/test_gpu_prompt_logprobs.py::TestGPUPromptLogprobs::test_prompt_logprobs - AttributeError

修复建议:

1. fastdeploy/worker/input_batch.py L111-L113：将 rotary_dim 计算封装成兼容逻辑；当 rotary_percent 缺失/非数值时使用 1，当 head_dim 缺失但 hidden_size/num_attention_heads 可用时派生，否则只在实际创建 rope_emb 前校验并报出清晰错误。
2. fastdeploy/worker/input_batch.py L768-L770：ProposerInputBatch 中同样的 rotary_dim 计算建议复用同一个 helper，避免后续 speculative 路径出现相同问题。
3. 如果 head_dim 已成为必需字段，也需同步补齐测试 Mock：tests/worker/test_gpu_model_runner.py 的 self.mock_model_config 以及 tests/worker/test_gpu_prompt_logprobs.py 的 MockConfig.ModelConfig 至少设置 head_dim 与 rotary_percent=1。
4. tests/e2e/test_Qwen2_5_VL_serving.py:450 的 Reset block num 缺失失败目前不像由新增 rotary_dim 行直接触发；建议修复上述初始化失败后 rerun，若仍失败再检查启动时 config.log 是否仍应输出该字段。

修复建议摘要: 兼容缺失 head_dim 或补 Mock 字段

关联变更: fastdeploy/worker/input_batch.py L111-L113、L243-L249、L768-L770、L827-L833

链接: 查看日志