[Bugfix] Fix hidden state size in EAGLE PD disaggregation buffers

python/sglang/srt/managers/scheduler.py

@@ -338,7 +338,7 @@ def __init__(
 
         # Draft workers are looked up via `SpeculativeAlgorithm` registry; new
         # algorithms should register their factory instead of patching this code.
-        if self.spec_algorithm.name in {"EAGLE", "EAGLE3"}:
+        if self.spec_algorithm.is_eagle():
             draft_worker_kwargs["enable_overlap"] = self.enable_overlap
         self.draft_worker = self.spec_algorithm.create_draft_worker(
             **draft_worker_kwargs
@@ -864,8 +864,16 @@ def init_disaggregation(self):
             )
             self.disagg_metadata_buffers = MetadataBuffers(
                 buffer_size,
-                hidden_size=self.model_config.hf_text_config.hidden_size,
-                hidden_states_dtype=self.model_config.dtype,
+                hidden_size=(
+                    self.draft_worker.model_config.hidden_size
+                    if self.spec_algorithm.is_eagle()
+                    else 64  # For safety reasons, won't be used
+                ),
+                hidden_states_dtype=(
+                    self.draft_worker.model_config.dtype
+                    if self.spec_algorithm.is_eagle()
+                    else torch.float32
+                ),
                 custom_mem_pool=self.token_to_kv_pool_allocator.get_kvcache().maybe_get_custom_mem_pool(),
             )
 
@@ -909,8 +917,16 @@ def init_disaggregation(self):
             )
             self.disagg_metadata_buffers = MetadataBuffers(
                 buffer_size,
-                hidden_size=self.model_config.hf_text_config.hidden_size,
-                hidden_states_dtype=self.model_config.dtype,
+                hidden_size=(
+                    self.draft_worker.model_config.hidden_size
+                    if self.spec_algorithm.is_eagle()
+                    else 64  # For safety reasons, won't be used
+                ),
+                hidden_states_dtype=(
+                    self.draft_worker.model_config.dtype
+                    if self.spec_algorithm.is_eagle()
+                    else torch.float32
+                ),
                 custom_mem_pool=self.token_to_kv_pool_allocator.get_kvcache().maybe_get_custom_mem_pool(),
             )