# PR #36248 完整报告

- 仓库：`sgl-project/sglang`
- 标题：[PP] Support prefill CUDA graph proxy tensors
- 合并时间：2026-08-31 14:11
- 原文链接：http://prhub.com.cn/sgl-project/sglang/pull/36248

---

# 执行摘要

- 一句话：
- 推荐动作：

# 功能与动机



# 实现拆解

1. **核心源码路径**
- `python/sglang/srt/model_executor/runner/prefill_cuda_graph_runner.py`（data-contract）：源码主路径；涉及符号 _build_layer_model_forward_kwargs, _static_pp_proxy_tensors, _capture_pp_proxy_tensors；包含 控制流调整、配置键调整、符号定义调整；+40/-10；关键符号：_build_layer_model_forward_kwargs, _static_pp_proxy_tensors, _capture_pp_proxy_tensors
- `python/sglang/srt/model_executor/runner_utils/buffers.py`（data-contract）：源码主路径；涉及符号 _allocate_pp_proxy_tensors；包含 控制流调整、配置键调整、符号定义调整；+59/-46；关键符号：_allocate_pp_proxy_tensors
- `python/sglang/srt/arg_groups/memory_hook.py`（dependency-wiring）：源码主路径；包含 导入关系调整、控制流调整、配置键调整；+15/-1

2. **测试与验证**
- `test/registered/unit/model_executor/test_prefill_cuda_graph_runner_helpers.py`（test-coverage）：测试配套；涉及符号 _make_pp_buffers_and_registry, test_pp_proxy_stable_buffers_accept_full_and_hidden_only_contracts, test_layer_model_kwargs_bind_optional_inputs_by_signature, proxy_before_embeds；包含 测试覆盖调整、导入关系调整、配置键调整；+127/-0；关键符号：_make_pp_buffers_and_registry, test_pp_proxy_stable_buffers_accept_full_and_hidden_only_contracts, test_layer_model_kwargs_bind_optional_inputs_by_signature, proxy_before_embeds
- `test/registered/unit/server_args/test_server_args.py`（test-coverage）：测试配套；涉及符号 TestPipelineParallelPrefillCudaGraphPolicy, test_pp_prefill_graph_is_opt_in, test_pp_prefill_capture_limit_policy；包含 测试覆盖调整、导入关系调整、控制流调整；+54/-0；关键符号：TestPipelineParallelPrefillCudaGraphPolicy, test_pp_prefill_graph_is_opt_in, test_pp_prefill_capture_limit_policy
- `test/registered/unit/model_executor/model_runner_components/test_layer_setup.py`（test-coverage）：测试配套；涉及符号 test_pipeline_placeholders_preserve_global_layer_ids；包含 测试覆盖调整、配置键调整、符号定义调整；+14/-0；关键符号：test_pipeline_placeholders_preserve_global_layer_ids

关键文件：
- `python/sglang/srt/model_executor/runner/prefill_cuda_graph_runner.py`（模块 sglang/srt；类别 source；类型 data-contract；符号 _build_layer_model_forward_kwargs, _static_pp_proxy_tensors, _capture_pp_proxy_tensors）: 源码主路径；涉及符号 _build_layer_model_forward_kwargs, _static_pp_proxy_tensors, _capture_pp_proxy_tensors；包含 控制流调整、配置键调整、符号定义调整；+40/-10
- `python/sglang/srt/model_executor/runner_utils/buffers.py`（模块 sglang/srt；类别 source；类型 data-contract；符号 _allocate_pp_proxy_tensors）: 源码主路径；涉及符号 _allocate_pp_proxy_tensors；包含 控制流调整、配置键调整、符号定义调整；+59/-46
- `test/registered/unit/model_executor/test_prefill_cuda_graph_runner_helpers.py`（模块 prefill/cuda/graph/runner/helpers；类别 test；类型 test-coverage；符号 _make_pp_buffers_and_registry, test_pp_proxy_stable_buffers_accept_full_and_hidden_only_contracts, test_layer_model_kwargs_bind_optional_inputs_by_signature, proxy_before_embeds）: 测试配套；涉及符号 _make_pp_buffers_and_registry, test_pp_proxy_stable_buffers_accept_full_and_hidden_only_contracts, test_layer_model_kwargs_bind_optional_inputs_by_signature, proxy_before_embeds；包含 测试覆盖调整、导入关系调整、配置键调整；+127/-0
- `test/registered/unit/server_args/test_server_args.py`（模块 server/args；类别 test；类型 test-coverage；符号 TestPipelineParallelPrefillCudaGraphPolicy, test_pp_prefill_graph_is_opt_in, test_pp_prefill_capture_limit_policy）: 测试配套；涉及符号 TestPipelineParallelPrefillCudaGraphPolicy, test_pp_prefill_graph_is_opt_in, test_pp_prefill_capture_limit_policy；包含 测试覆盖调整、导入关系调整、控制流调整；+54/-0
- `python/sglang/srt/arg_groups/memory_hook.py`（模块 sglang/srt；类别 source；类型 dependency-wiring）: 源码主路径；包含 导入关系调整、控制流调整、配置键调整；+15/-1

关键符号：_build_layer_model_forward_kwargs, _static_pp_proxy_tensors, _capture_pp_proxy_tensors, _allocate_pp_proxy_tensors, _make_pp_buffers_and_registry, test_pp_proxy_stable_buffers_accept_full_and_hidden_only_contracts, test_layer_model_kwargs_bind_optional_inputs_by_signature, proxy_before_embeds, test_finalize_pp_proxy_trims_padded_token_rows, test_bcg_eager_tail_uses_live_multimodal_embeddings


# 评论区精华



- 暂无高价值评论线程

# 风险与影响

- 风险：
- 影响：
 - 风险标记：暂无

# 关联脉络

- 暂无明显关联 PR