[PLE] The pull wait/copy op is a graph splitting op

With the wait/copy inside a compiled piece, Inductor's post-grad
auto-functionalization decomposition failed on the stage-0 ranks.

Co-Authored-By: Hinata <hinata@idihore.id>
Signed-off-by: MOVZX <movzx@yahoo.com>

vllm/config/compilation.py

@@ -775,6 +775,12 @@ class CompilationConfig:
         "vllm::qwen4_exp_ple_finalize_prefetch",
         "vllm::qwen4_exp_ple_short_conv",
         "vllm::qwen4_exp_qsa_with_output",
+        # Fork (pull transport with the compile-profile guard, found on the reference host): the PLE pull transport's wait+DMA
+        # mutates a module-owned buffer; inside an Inductor graph it failed
+        # post-grad with "auto_functionalized was not removed" on the stage-0
+        # ranks. As a split point it runs between pieces, like attention.
+        "vllm::ple_offload_wait_copy",
+        "vllm::ple_offload_wait",
         "vllm::linear_attention",
         "vllm::qwen_gdn_attention_core",
         "vllm::qwen_gdn_attention_core_fused_norm_packed",

vllm/v1/e2_guards.py

@@ -61,6 +61,7 @@ REQUIRED_SPLITTING_OPS = (
     "vllm::qwen4_exp_ple_finalize_prefetch",
     "vllm::qwen_gdn_attention_core",
     "vllm::qwen_gdn_attention_core_fused_norm_packed",
+    "vllm::ple_offload_wait_copy",
 )
 
 #: Upstream v0.30.0 auto-enabled breakable cudagraphs for these; this tree