# glm53-dflash-01-aux-hidden-states.patch — give the GLM trunk the EAGLE3 / DFlash2
# auxiliary-hidden-state interface.
#
# STATUS: WIP, and INDEPENDENT of the compile changelist (01-05). It patches the same
# file as glm53-compile-01-model-decorator.patch, so if both are ever applied, apply 01
# first. Not part of the deployed shape.
#
# WHY.  `--speculative-config '{"method":"dflash","model":"/path/to/drafter", ...}'` with
# incoai/GLM-5.3-Flash-DFlash2 is accepted -- the config parses, the drafter loads -- and
# then the worker dies at init with:
#
#     RuntimeError: Model does not support EAGLE3 interface
#
# raised by v1/worker/gpu/spec_decode/eagle/eagle3_utils.py:20 through
# `supports_eagle3(model)` == `isinstance(model, SupportsEagle3)` (a runtime-checkable
# Protocol).  Nothing about the drafter, the method name or the payload is wrong:
# `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`.
#   * 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().
#   * 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.
#
# WHAT IT ADDS for GLM: the mixin on Glm5NextModel; aux capture in the layer loop, where
# the aux state is the single-stream `hidden_states` ENTERING that layer (GLM's loop
# 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).
#
# 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
# decode-speed fix (a verification pass pays pointwise cost proportional to its token
# count, which is why MTP4 measured flat).  It is the prerequisite for driving DFlash2
# on this endpoint at all.
#
# TARGET: vllm/models/glm5next/common/model.py   (inside the pinned image)
#   base    sha256 008eeaa2a07e091465319f86900b5a321d921238723ccbe22e6b9e971e1d2b28
#   patched sha256 4ff6ff8ddf8af3d74bef699f98fc95a99ee47586f6e785248e7dc4b683b72d7a
#
# 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 @@
 logger = init_logger(__name__)
 
 
+from vllm.model_executor.models.interfaces import EagleModelMixin
+
+
 class Glm5NextMLP(nn.Module):
     def __init__(
         self,
@@ -623,7 +626,7 @@
         )
 
 
-class Glm5NextModel(nn.Module):
+class Glm5NextModel(nn.Module, EagleModelMixin):
     def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
         super().__init__()
 
@@ -730,7 +733,14 @@
         if self.is_sequence_parallel:
             hidden_states = sp_shard(hidden_states)
 
-        for layer in self._active_layers:
+        aux_hidden_states: list[torch.Tensor] = []
+        for local_idx, layer in enumerate(self._active_layers):
+            if (getattr(self, "start_layer", 0) + local_idx) in self.aux_hidden_state_layers:
+                # Auxiliary hidden state for EAGLE-3 / DFlash2 drafts: the
+                # single-stream state ENTERING this layer, before its input norm --
+                # the same semantic deepseek_v2 implements by appending
+                # hidden_states + residual at this point in its loop.
+                aux_hidden_states.append(hidden_states)
             hidden_states, residual, post, comb = layer(
                 positions, hidden_states, residual, post, comb
             )
@@ -749,6 +759,8 @@
             hidden_states = sp_all_gather(hidden_states)[:full_num_tokens]
 
         hidden_states = self.norm(hidden_states)
+        if len(aux_hidden_states) > 0:
+            return hidden_states, aux_hidden_states
         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)
 
+    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,