patches: glm53-dflash-01 v2 -- satisfy SupportsEagle3 by inheriting the Protocol

v1 hand-wrote set_aux_hidden_state_layers/get_eagle3_default_aux_hidden_state_layers
and the boot reproduced the identical 'Model does not support EAGLE3 interface'.
Reason: supports_eagle3() is an isinstance check against a runtime-checkable
Protocol whose members include a required ClassVar flag
(interfaces.py:1663 supports_eagle3: ClassVar[Literal[True]] = True), so methods
alone do not satisfy it. Inheriting SupportsEagle3 -- the idiom every model in
this image uses, including the fork's own DFlashLagunaForCausalLM -- supplies the
flag, SupportsEagleBase's has_own_lm_head/has_own_embed_tokens, and working
default implementations of both methods; the hand-written methods are deleted.

base 008eeaa2... -> patched 5eea89c5..., apply-verified against the pinned image.
Runbook: corrected method name, recorded the Protocol trap as the generalizable
lesson for future spec-decode ports.

scripts/patches/glm53-dflash-01-aux-hidden-states.patch

@@ -17,15 +17,20 @@
 # `glm5next` implements neither half of the contract.
 #
 # THE CONTRACT (read out of this image, not assumed):
-#   * interfaces.py:1694 -- `holder = getattr(parent_ref, "model", parent_ref)` then
-#     `assert isinstance(holder, EagleModelMixin)` then
-#     `holder._set_aux_hidden_state_layers(layers)`.  The INNER trunk must inherit
-#     EagleModelMixin, which supplies `aux_hidden_state_layers` and the slot helpers.
-#   * interfaces.py:1712-1719 -- the default layer set is (2, L // 2, L - 3) and needs
-#     the holder to expose `layers`.
+#   * interfaces.py:1663 -- `supports_eagle3: ClassVar[Literal[True]] = True`.  THE FLAG
+#     IS A REQUIRED MEMBER FOR `isinstance`: satisfying the Protocol with methods alone
+#     still raises the same RuntimeError -- v1 of this patch did exactly that and the boot
+#     failed identically.  INHERIT the Protocol instead.  That is the idiom in this image
+#     (LagunaForCausalLM, MiMoV2FlashForCausalLM and DFlashLagunaForCausalLM all list
+#     SupportsEagle3 among their bases) and it brings the flag, SupportsEagleBase's
+#     has_own_lm_head/has_own_embed_tokens, AND working non-stub defaults for both methods.
+#   * interfaces.py:1673-1719 -- the inherited defaults unwrap
+#     `holder = getattr(parent_ref, "model", parent_ref)`, assert
+#     `isinstance(holder, EagleModelMixin)`, set `holder._set_aux_hidden_state_layers`, and
+#     default to (2, L // 2, L - 3).  So the INNER trunk must inherit EagleModelMixin,
+#     which supplies `aux_hidden_state_layers` and the slot helpers.
 #   * eagle3_utils.py:26-31 -- layers come from the spec config when it carries them
-#     (this drafter ships target_layer_ids [5, 14, 24, 33, 42], mapped to trunk
-#     indices), else from get_eagle3_default_aux_hidden_state_layers().
+#     (this drafter ships target_layer_ids [5, 14, 24, 33, 42]), else from the default.
 #   * deepseek_v2.py:1551-1585 and v1/worker/gpu/cudagraph_utils.py:621 -- the trunk
 #     appends one tensor per aux layer while looping and returns the TUPLE
 #     (hidden_states, aux_hidden_states) only when that list is non-empty.
@@ -35,9 +40,7 @@
 # carries the already-summed single-stream state -- its non-mHC branch re-splits
 # `residual = hidden_states` inside the layer -- so this is the equivalent of
 # deepseek_v2's `hidden_states + residual`); the conditional tuple return; and
-# set_aux_hidden_state_layers / get_eagle3_default_aux_hidden_state_layers on
-# Glm5NextForCausalLM, delegating to the inner trunk (the same two-liner shape
-# kimi_k25.py:466 and pixtral.py:503 use).
+# SupportsEagle3 among Glm5NextForCausalLM's bases.
 #
 # HONEST SCOPE: this makes the interface PRESENT, which is exactly what the engine
 # checks -- it does not make drafts GOOD, and per the measured profile it is not a
@@ -47,23 +50,26 @@
 #
 # TARGET: vllm/models/glm5next/common/model.py   (inside the pinned image)
 #   base    sha256 008eeaa2a07e091465319f86900b5a321d921238723ccbe22e6b9e971e1d2b28
-#   patched sha256 4ff6ff8ddf8af3d74bef699f98fc95a99ee47586f6e785248e7dc4b683b72d7a
+#   patched sha256 5eea89c51d4ab37ef823526286b5103e39996802ca08aa1934f6373fc4728062
 #
 # APPLY: `patch -p1` from the vllm package root, or bind-mount the patched file over
 # site-packages (which is what a scratch-launcher DFLASH_PATCH=1 does).
 --- a/vllm/models/glm5next/common/model.py
 +++ b/vllm/models/glm5next/common/model.py
-@@ -95,6 +95,9 @@
+@@ -95,6 +95,12 @@
  logger = init_logger(__name__)
  
  
-+from vllm.model_executor.models.interfaces import EagleModelMixin
++from vllm.model_executor.models.interfaces import (
++    EagleModelMixin,
++    SupportsEagle3,
++)
 +
 +
  class Glm5NextMLP(nn.Module):
      def __init__(
          self,
-@@ -623,7 +626,7 @@
+@@ -623,7 +629,7 @@
          )
  
  
@@ -72,7 +78,7 @@
      def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
          super().__init__()
  
-@@ -730,7 +733,14 @@
+@@ -730,7 +736,14 @@
          if self.is_sequence_parallel:
              hidden_states = sp_shard(hidden_states)
  
@@ -88,7 +94,7 @@
              hidden_states, residual, post, comb = layer(
                  positions, hidden_states, residual, post, comb
              )
-@@ -749,6 +759,8 @@
+@@ -749,6 +762,8 @@
              hidden_states = sp_all_gather(hidden_states)[:full_num_tokens]
  
          hidden_states = self.norm(hidden_states)
@@ -97,21 +103,17 @@
          return hidden_states
  
      def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]) -> set[str]:
-@@ -963,6 +975,17 @@
-     def embed_input_ids(self, input_ids: torch.Tensor) -> torch.Tensor:
-         return self.model.embed_input_ids(input_ids)
+@@ -935,7 +950,12 @@
  
-+    def set_aux_hidden_state_layers(self, layers: tuple[int, ...]) -> None:
-+        # EAGLE-3 / DFlash2 interface (interfaces.py asserts the holder inherits
-+        # EagleModelMixin, which supplies the slot plumbing behind this attribute).
-+        self.model.aux_hidden_state_layers = layers
-+
-+    def get_eagle3_default_aux_hidden_state_layers(self) -> tuple[int, ...]:
-+        # Only used when the spec config carries no explicit aux layers; the
-+        # DFlash2 drafters this fleet runs ship target_layer_ids instead.
-+        num_layers = len(self.model.layers)
-+        return (2, num_layers // 2, num_layers - 3)
-+
-     def forward(
-         self,
-         input_ids: torch.Tensor | None,
+ 
+ class Glm5NextForCausalLM(
+-    nn.Module, HasInnerState, SupportsPP, MixtureOfExperts, IsHybrid
++    nn.Module,
++    HasInnerState,
++    SupportsPP,
++    MixtureOfExperts,
++    IsHybrid,
++    SupportsEagle3,
+ ):
+     def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
+         super().__init__()