[SPEC V2][2/N] feat: adaptive spec support spec v2 (#23336)

Co-authored-by: Qiaolin-Yu <liin1211@outlook.com>

python/sglang/srt/managers/scheduler_output_processor_mixin.py

@@ -386,8 +386,16 @@ class SchedulerOutputProcessorMixin:
         result.num_accepted_drafts = sum(accept_lens) - len(batch.reqs)
         result.num_accepted_drafts_per_req_cpu = [x - 1 for x in accept_lens]
 
+        # Feed the adaptive controller now that accept_lens is on CPU,
+        # instead of doing a synchronous GPU→CPU copy in the worker hot path.
+        # BaseSpecWorker provides a no-op default for non-adaptive workers.
+        self.model_worker.on_verify_complete_cpu(result.num_accepted_drafts_per_req_cpu)
+
         predict_tokens = []
-        stride = self.draft_worker.speculative_num_draft_tokens
+        # In adaptive spec-v2, the worker state may already have switched when this
+        # delayed result is processed. Use the draft token count recorded on result.
+        stride = result.speculative_num_draft_tokens
+        assert stride is not None, "spec-v2 result missing speculative_num_draft_tokens"
 
         for i, req in enumerate(batch.reqs):
             # -1 because prepare_for_decode pre-claimed the bonus slot.

python/sglang/srt/managers/utils.py

@@ -39,6 +39,7 @@ class GenerationBatchResult:
     copy_done: Optional[torch.cuda.Event] = None
     delay_sample_func: Optional[callable] = None
     future_indices: Optional[FutureIndices] = None
+    speculative_num_draft_tokens: Optional[int] = None
 
     # FIXME(lsyin): maybe move to a better place?
     # sync path: forward stream -> output processor

python/sglang/srt/speculative/adaptive_spec_params.py

@@ -9,6 +9,8 @@ import json
 import logging
 from typing import TYPE_CHECKING
 
+from sglang.srt.utils import log_info_on_rank0
+
 if TYPE_CHECKING:
     from sglang.srt.server_args import ServerArgs
 
@@ -32,11 +34,6 @@ def adaptive_unsupported_reason(server_args: ServerArgs) -> str | None:
             "enable_dp_attention=True is not supported "
             "(adaptive tier decisions are not synchronized across DP ranks)"
         )
-    if not server_args.disable_overlap_schedule:
-        return (
-            "the overlap scheduler (spec v2) is enabled "
-            "(adaptive is only implemented for EAGLEWorker v1)"
-        )
     if server_args.enable_multi_layer_eagle:
         return (
             "enable_multi_layer_eagle=True is not supported "
@@ -95,7 +92,22 @@ class AdaptiveSpeculativeParams:
     ):
         cfg = config or {}
         # TODO: Wider range of candidate_steps (once lazy init is supported).
-        self.candidate_steps = sorted(set(cfg.get("candidate_steps", [1, 3, 7])))
+        candidates = set(cfg.get("candidate_steps", [1, 3, 7]))
+
+        # Ensure the worker's initial speculative_num_steps is itself a candidate.
+        # Otherwise AdaptiveController.register() would store the worker's pre-built
+        # runtime state under a key that _activate() never queries, leaking that
+        # state's draft attn backend and cuda graph buffers for the process lifetime.
+        if initial_steps not in candidates:
+            log_info_on_rank0(
+                logger,
+                f"Adding initial speculative_num_steps={initial_steps} to "
+                f"candidate_steps={sorted(candidates)} so the pre-built "
+                f"runtime state is reused.",
+            )
+            candidates.add(initial_steps)
+
+        self.candidate_steps = sorted(candidates)
         assert (
             len(self.candidate_steps) >= 2
         ), "candidate_steps must have at least 2 distinct values"
@@ -108,18 +120,16 @@ class AdaptiveSpeculativeParams:
         self.down_hysteresis = cfg.get("down_hysteresis", -0.25)
         self.up_hysteresis = cfg.get("up_hysteresis", 0.0)
 
-        self.current_steps = min(
+        self.current_steps = initial_steps
-            self.candidate_steps,
-            key=lambda step: (abs(step - initial_steps), -step),
-        )
 
         # Initialize EMA at current steps - 1 (neutral starting point)
         self.ema_accept_len = float(self.current_steps - 1)
         self._batch_count = 0
 
-        logger.info(
+        log_info_on_rank0(
+            logger,
             f"AdaptiveSpeculativeParams initialized: "
-            f"steps={self.current_steps}, candidate_steps={self.candidate_steps}"
+            f"steps={self.current_steps}, candidate_steps={self.candidate_steps}",
         )
 
     def update(self, num_accepted_drafts_per_req: list[int]) -> bool:
@@ -171,9 +181,10 @@ class AdaptiveSpeculativeParams:
 
         if target != old_steps:
             self.current_steps = target
-            logger.info(
+            log_info_on_rank0(
+                logger,
                 f"Adaptive spec params updated: steps {old_steps} -> {target} "
-                f"(ema_accept_len={self.ema_accept_len:.2f})"
+                f"(ema_accept_len={self.ema_accept_len:.2f})",
             )
             return True
         return False

python/sglang/srt/speculative/base_spec_worker.py

@@ -32,3 +32,11 @@ class BaseSpecWorker(ABC):
     def clear_cache_pool(self):
         # TODO: move this abstract method to BaseTpWorker and call through self.model_runner
         pass
+
+    def on_verify_complete_cpu(self, num_accepted_drafts_per_req: list[int]) -> None:
+        """Hook called after verify finishes and accept counts are on CPU.
+
+        Default no-op. Adaptive-aware workers override this to feed the
+        controller without forcing a GPU→CPU sync in the worker hot path.
+        """
+        pass

python/sglang/srt/speculative/eagle_info_v2.py

@@ -124,7 +124,12 @@ class EagleDraftInputV2Mixin:
         num_needed_tokens = 0
         for i, r in enumerate(batch.reqs):
             cur = r.kv_allocated_len
-            nxt = r.kv_committed_len + double_alloc
+            # max(cur, ...) clamps so adaptive downswitch (smaller alloc_len_per_decode)
+            # cannot make nxt < cur and corrupt allocator state. kv_committed_len lags
+            # batch.seq_lens by ~1 verify in overlap mode, so we react to adaptive
+            # switches one batch later than a seq_lens-based baseline; the 2*alloc
+            # over-allocation buffer absorbs that lag.
+            nxt = max(cur, r.kv_committed_len + double_alloc)
             cur_kv_lens[i] = cur
             nxt_kv_lens[i] = nxt
             num_needed_tokens += nxt - cur

python/sglang/srt/speculative/eagle_worker.py

@@ -73,6 +73,7 @@ from sglang.srt.utils import (
     is_cuda,
     is_musa,
     is_npu,
+    log_info_on_rank0,
     next_power_of_2,
 )
 from sglang.srt.utils.patch_torch import monkey_patch_torch_reductions
@@ -278,30 +279,34 @@ class EAGLEWorker(TpModelWorker):
         if self.speculative_num_steps > 1:
             tic = time.perf_counter()
             before_mem = get_available_gpu_memory(self.device, self.gpu_id)
-            logger.info(
+            log_info_on_rank0(
-                f"Capture draft cuda graph begin. This can take up to several minutes. avail mem={before_mem:.2f} GB"
+                logger,
+                f"Capture draft cuda graph begin. This can take up to several minutes. avail mem={before_mem:.2f} GB",
             )
             self.cuda_graph_runner = Device2DraftCudaGraphRunner[
                 self.target_worker.device
             ](self)
             after_mem = get_available_gpu_memory(self.device, self.gpu_id)
-            logger.info(
+            log_info_on_rank0(
-                f"Capture draft cuda graph end. Time elapsed: {time.perf_counter() - tic:.2f} s. mem usage={(before_mem - after_mem):.2f} GB. avail mem={after_mem:.2f} GB."
+                logger,
+                f"Capture draft cuda graph end. Time elapsed: {time.perf_counter() - tic:.2f} s. mem usage={(before_mem - after_mem):.2f} GB. avail mem={after_mem:.2f} GB.",
             )
 
         # Capture extend
         if self.draft_extend_attn_backend and not _is_npu:
             tic = time.perf_counter()
             before_mem = get_available_gpu_memory(self.device, self.gpu_id)
-            logger.info(
+            log_info_on_rank0(
-                f"Capture draft extend cuda graph begin. This can take up to several minutes. avail mem={before_mem:.2f} GB"
+                logger,
+                f"Capture draft extend cuda graph begin. This can take up to several minutes. avail mem={before_mem:.2f} GB",
             )
             self.cuda_graph_runner_for_draft_extend = EAGLEDraftExtendCudaGraphRunner(
                 self
             )
             after_mem = get_available_gpu_memory(self.device, self.gpu_id)
-            logger.info(
+            log_info_on_rank0(
-                f"Capture draft extend cuda graph end. Time elapsed: {time.perf_counter() - tic:.2f} s. mem usage={(before_mem - after_mem):.2f} GB. avail mem={after_mem:.2f} GB."
+                logger,
+                f"Capture draft extend cuda graph end. Time elapsed: {time.perf_counter() - tic:.2f} s. mem usage={(before_mem - after_mem):.2f} GB. avail mem={after_mem:.2f} GB.",
             )
 
     def apply_runtime_state(self, state: SpecRuntimeState):
@@ -309,11 +314,12 @@ class EAGLEWorker(TpModelWorker):
         if self.speculative_num_steps == state.speculative_num_steps:
             return
 
-        logger.info(
+        log_info_on_rank0(
+            logger,
             "Switch adaptive runtime state: "
             f"steps {self.speculative_num_steps} -> {state.speculative_num_steps}, "
             f"draft_tokens {self.speculative_num_draft_tokens} -> "
-            f"{state.speculative_num_draft_tokens}"
+            f"{state.speculative_num_draft_tokens}",
         )
 
         self.speculative_num_steps = state.speculative_num_steps
@@ -384,10 +390,11 @@ class EAGLEWorker(TpModelWorker):
             )
 
         after_mem = get_available_gpu_memory(self.device, self.gpu_id)
-        logger.info(
+        log_info_on_rank0(
+            logger,
             f"Built adaptive runtime state steps={speculative_num_steps}: "
             f"elapsed={time.perf_counter() - tic:.2f}s, "
-            f"mem={(before_mem - after_mem):.2f}GB"
+            f"mem={(before_mem - after_mem):.2f}GB",
         )
 
         return state
@@ -403,9 +410,9 @@ class EAGLEWorker(TpModelWorker):
             self.speculative_num_draft_tokens,
             self.draft_attn_backend,
             self.draft_extend_attn_backend,
-            getattr(self.draft_model_runner, "draft_attn_backend", None),
+            self.draft_model_runner.draft_attn_backend,
-            getattr(self, "cuda_graph_runner", None),
+            self.cuda_graph_runner,
-            getattr(self, "cuda_graph_runner_for_draft_extend", None),
+            self.cuda_graph_runner_for_draft_extend,
             sa.speculative_num_steps,
             sa.speculative_num_draft_tokens,
         )
@@ -507,9 +514,8 @@ class EAGLEWorker(TpModelWorker):
                 batch.reqs, "set_spec_draft_extend_end_time", trace_only=True
             )
 
-            controller = getattr(self, "adaptive_controller", None)
+            if self.adaptive_controller is not None:
-            if controller is not None:
+                self.adaptive_controller.on_verify_complete(
-                controller.on_verify_complete(
                     verify_output.num_accepted_drafts_per_req_cpu
                 )
 

python/sglang/srt/speculative/eagle_worker_v2.py

@@ -30,8 +30,13 @@ from sglang.srt.managers.io_struct import (
 from sglang.srt.managers.schedule_batch import ModelWorkerBatch
 from sglang.srt.managers.scheduler import GenerationBatchResult
 from sglang.srt.managers.tp_worker import TpModelWorker
+from sglang.srt.model_executor.cuda_graph_runner import CudaGraphRunner
 from sglang.srt.model_executor.forward_batch_info import CaptureHiddenMode, ForwardBatch
 from sglang.srt.server_args import ServerArgs
+from sglang.srt.speculative.adaptive_runtime_state import (
+    AdaptiveController,
+    SpecRuntimeState,
+)
 from sglang.srt.speculative.base_spec_worker import BaseDraftWorker, BaseSpecWorker
 from sglang.srt.speculative.draft_utils import DraftBackendFactory
 from sglang.srt.speculative.eagle_draft_cuda_graph_runner import (
@@ -65,6 +70,7 @@ from sglang.srt.utils.common import (
     is_hip,
     is_musa,
     is_npu,
+    log_info_on_rank0,
     next_power_of_2,
 )
 from sglang.srt.utils.patch_torch import monkey_patch_torch_reductions
@@ -273,15 +279,17 @@ class EagleDraftWorker(BaseDraftWorker):
         if self.speculative_num_steps > 1:
             tic = time.perf_counter()
             before_mem = get_available_gpu_memory(self.device, self.gpu_id)
-            logger.info(
+            log_info_on_rank0(
-                f"Capture draft cuda graph begin. This can take up to several minutes. avail mem={before_mem:.2f} GB"
+                logger,
+                f"Capture draft cuda graph begin. This can take up to several minutes. avail mem={before_mem:.2f} GB",
             )
             self.cuda_graph_runner = Device2DraftCudaGraphRunner[
                 self.target_worker.device
             ](self)
             after_mem = get_available_gpu_memory(self.device, self.gpu_id)
-            logger.info(
+            log_info_on_rank0(
-                f"Capture draft cuda graph end. Time elapsed: {time.perf_counter() - tic:.2f} s. mem usage={(before_mem - after_mem):.2f} GB. avail mem={after_mem:.2f} GB."
+                logger,
+                f"Capture draft cuda graph end. Time elapsed: {time.perf_counter() - tic:.2f} s. mem usage={(before_mem - after_mem):.2f} GB. avail mem={after_mem:.2f} GB.",
             )
 
         Device2ExtendCudaGraphRunner = {
@@ -313,15 +321,17 @@ class EagleDraftWorker(BaseDraftWorker):
         ):
             tic = time.perf_counter()
             before_mem = get_available_gpu_memory(self.device, self.gpu_id)
-            logger.info(
+            log_info_on_rank0(
-                f"Capture draft extend cuda graph begin. This can take up to several minutes. avail mem={before_mem:.2f} GB"
+                logger,
+                f"Capture draft extend cuda graph begin. This can take up to several minutes. avail mem={before_mem:.2f} GB",
             )
             self.cuda_graph_runner_for_draft_extend = Device2ExtendCudaGraphRunner[
                 self.target_worker.device
             ](self)
             after_mem = get_available_gpu_memory(self.device, self.gpu_id)
-            logger.info(
+            log_info_on_rank0(
-                f"Capture draft extend cuda graph end. Time elapsed: {time.perf_counter() - tic:.2f} s. mem usage={(before_mem - after_mem):.2f} GB. avail mem={after_mem:.2f} GB."
+                logger,
+                f"Capture draft extend cuda graph end. Time elapsed: {time.perf_counter() - tic:.2f} s. mem usage={(before_mem - after_mem):.2f} GB. avail mem={after_mem:.2f} GB.",
             )
 
     def draft(self, model_worker_batch: ModelWorkerBatch):
@@ -678,6 +688,13 @@ class EAGLEWorkerV2(BaseSpecWorker):
             target_worker,
         )
 
+        # Adaptive speculative
+        self.adaptive_controller: Optional[AdaptiveController] = None
+        if server_args.speculative_adaptive:
+            self.adaptive_controller = AdaptiveController(
+                self, config_path=server_args.speculative_adaptive_config
+            )
+
         # Some dummy tensors
         self.num_new_pages_per_topk = torch.empty(
             (), dtype=torch.int64, device=self.device
@@ -686,6 +703,25 @@ class EAGLEWorkerV2(BaseSpecWorker):
 
         self.plan_stream, self.plan_stream_ctx = _get_plan_stream(self.device)
 
+        # Build adaptive runtime states (must be after draft worker is fully initialized)
+        if self.adaptive_controller is not None:
+            with self._draft_worker.draft_tp_context(
+                self._draft_worker.draft_runner.tp_group
+            ), speculative_moe_backend_context(), speculative_moe_a2a_backend_context():
+                self.adaptive_controller.register(
+                    SpecRuntimeState(
+                        speculative_num_steps=self.speculative_num_steps,
+                        speculative_num_draft_tokens=self.speculative_num_draft_tokens,
+                        draft_attn_backend=self._draft_worker.draft_attn_backend,
+                        cuda_graph_runner=self._draft_worker.cuda_graph_runner,
+                        target_attn_backend=self._target_worker.model_runner.attn_backend,
+                        target_graph_runner=self._target_worker.model_runner.graph_runner,
+                        draft_extend_attn_backend=self._draft_worker.draft_extend_attn_backend,
+                        cuda_graph_runner_for_draft_extend=self._draft_worker.cuda_graph_runner_for_draft_extend,
+                    )
+                )
+                self.adaptive_controller.init_states()
+
     @property
     def target_worker(self):
         return self._target_worker
@@ -754,8 +790,150 @@ class EAGLEWorkerV2(BaseSpecWorker):
                 self.draft_worker._draft_extend_for_decode(
                     model_worker_batch, batch_output
                 )
+
             return batch_output
 
+    def on_verify_complete_cpu(self, accepted_draft_tokens: list[int]) -> None:
+        if self.adaptive_controller is not None:
+            self.adaptive_controller.on_verify_complete(accepted_draft_tokens)
+
+    # -- Adaptive speculative decoding protocol --
+
+    def build_adaptive_runtime_state(
+        self, speculative_num_steps: int, speculative_num_draft_tokens: int
+    ) -> SpecRuntimeState:
+        """Build a SpecRuntimeState for the given step configuration."""
+        tic = time.perf_counter()
+        before_mem = get_available_gpu_memory(self.device, self.gpu_id)
+
+        with self._override_worker_state(
+            speculative_num_steps, speculative_num_draft_tokens
+        ):
+            self._draft_worker.init_attention_backend()
+            self._draft_worker.init_cuda_graphs()
+
+            # Build target attention backend and CUDA graph runner
+            target_model_runner = self._target_worker.model_runner
+            backup_init = target_model_runner.init_new_workspace
+            try:
+                target_attn_backend = target_model_runner._get_attention_backend(
+                    init_new_workspace=True
+                )
+            finally:
+                target_model_runner.init_new_workspace = backup_init
+
+            target_graph_runner = None
+            if not self.server_args.disable_cuda_graph:
+                target_graph_runner = CudaGraphRunner(
+                    target_model_runner,
+                    attn_backend=target_attn_backend,
+                    speculative_num_steps=speculative_num_steps,
+                    speculative_num_draft_tokens=speculative_num_draft_tokens,
+                )
+
+            state = SpecRuntimeState(
+                speculative_num_steps=speculative_num_steps,
+                speculative_num_draft_tokens=speculative_num_draft_tokens,
+                draft_attn_backend=self._draft_worker.draft_attn_backend,
+                cuda_graph_runner=self._draft_worker.cuda_graph_runner,
+                target_attn_backend=target_attn_backend,
+                target_graph_runner=target_graph_runner,
+                draft_extend_attn_backend=self._draft_worker.draft_extend_attn_backend,
+                cuda_graph_runner_for_draft_extend=self._draft_worker.cuda_graph_runner_for_draft_extend,
+            )
+
+        after_mem = get_available_gpu_memory(self.device, self.gpu_id)
+        log_info_on_rank0(
+            logger,
+            f"Built adaptive runtime state steps={speculative_num_steps}: "
+            f"elapsed={time.perf_counter() - tic:.2f}s, "
+            f"mem={(before_mem - after_mem):.2f}GB",
+        )
+
+        return state
+
+    def apply_runtime_state(self, state: SpecRuntimeState) -> None:
+        """Apply a pre-built runtime state to this worker."""
+        if self.speculative_num_steps == state.speculative_num_steps:
+            return
+
+        log_info_on_rank0(
+            logger,
+            "Switch adaptive runtime state: "
+            f"steps {self.speculative_num_steps} -> {state.speculative_num_steps}, "
+            f"draft_tokens {self.speculative_num_draft_tokens} -> "
+            f"{state.speculative_num_draft_tokens}",
+        )
+
+        # Top-level
+        self.speculative_num_steps = state.speculative_num_steps
+        self.speculative_num_draft_tokens = state.speculative_num_draft_tokens
+
+        # Draft side
+        dw = self._draft_worker
+        dw.speculative_num_steps = state.speculative_num_steps
+        dw.speculative_num_draft_tokens = state.speculative_num_draft_tokens
+        dw.draft_attn_backend = state.draft_attn_backend
+        dw.draft_runner.draft_attn_backend = state.draft_attn_backend
+        dw.cuda_graph_runner = state.cuda_graph_runner
+        dw.draft_extend_attn_backend = state.draft_extend_attn_backend
+        dw.cuda_graph_runner_for_draft_extend = state.cuda_graph_runner_for_draft_extend
+
+        # Target side
+        self._target_worker.model_runner.attn_backend = state.target_attn_backend
+        self._target_worker.model_runner.graph_runner = state.target_graph_runner
+
+        # Sync server_args
+        self.server_args.speculative_num_steps = state.speculative_num_steps
+        self.server_args.speculative_num_draft_tokens = (
+            state.speculative_num_draft_tokens
+        )
+
+    @contextlib.contextmanager
+    def _override_worker_state(
+        self, speculative_num_steps: int, speculative_num_draft_tokens: int
+    ):
+        """Temporarily override server_args and worker attributes for graph capture."""
+        sa = self.server_args
+        dw = self._draft_worker
+        backup = (
+            self.speculative_num_steps,
+            self.speculative_num_draft_tokens,
+            dw.speculative_num_steps,
+            dw.speculative_num_draft_tokens,
+            dw.draft_attn_backend,
+            dw.draft_extend_attn_backend,
+            dw.draft_runner.draft_attn_backend,
+            dw.cuda_graph_runner,
+            dw.cuda_graph_runner_for_draft_extend,
+            sa.speculative_num_steps,
+            sa.speculative_num_draft_tokens,
+        )
+
+        self.speculative_num_steps = speculative_num_steps
+        self.speculative_num_draft_tokens = speculative_num_draft_tokens
+        dw.speculative_num_steps = speculative_num_steps
+        dw.speculative_num_draft_tokens = speculative_num_draft_tokens
+        sa.speculative_num_steps = speculative_num_steps
+        sa.speculative_num_draft_tokens = speculative_num_draft_tokens
+
+        try:
+            yield
+        finally:
+            (
+                self.speculative_num_steps,
+                self.speculative_num_draft_tokens,
+                dw.speculative_num_steps,
+                dw.speculative_num_draft_tokens,
+                dw.draft_attn_backend,
+                dw.draft_extend_attn_backend,
+                dw.draft_runner.draft_attn_backend,
+                dw.cuda_graph_runner,
+                dw.cuda_graph_runner_for_draft_extend,
+                sa.speculative_num_steps,
+                sa.speculative_num_draft_tokens,
+            ) = backup
+
     def verify(self, batch: ModelWorkerBatch):
         # Since batch.seq_lens is allocated in another stream, we need
         # record_stream() to prevent pytorch gc and reuse the gpu memory
@@ -890,6 +1068,7 @@ class EAGLEWorkerV2(BaseSpecWorker):
             logits_output=logits_output,
             next_token_ids=predict,
             can_run_cuda_graph=can_run_cuda_graph,
+            speculative_num_draft_tokens=self.speculative_num_draft_tokens,
             next_draft_input=next_draft_input,
             accept_lens=accept_lens,
             routed_experts_output=forward_batch_output.routed_experts_output,

python/sglang/srt/speculative/multi_layer_eagle_worker_v2.py

@@ -788,6 +788,7 @@ class MultiLayerEagleWorkerV2(BaseSpecWorker):
             logits_output=logits_output,
             next_token_ids=predict,
             can_run_cuda_graph=can_run_cuda_graph,
+            speculative_num_draft_tokens=self.speculative_num_draft_tokens,
             next_draft_input=next_draft_input,
             accept_lens=accept_lens,
             routed_experts_output=forward_batch_output.routed_experts_output,

python/sglang/srt/speculative/standalone_worker.py

@@ -9,6 +9,9 @@ from sglang.srt.layers.moe.utils import (
 )
 from sglang.srt.managers.tp_worker import TpModelWorker
 from sglang.srt.server_args import ServerArgs
+from sglang.srt.speculative.adaptive_runtime_state import (
+    AdaptiveController,
+)
 from sglang.srt.speculative.eagle_worker import EAGLEWorker
 from sglang.srt.speculative.spec_info import SpeculativeAlgorithm
 from sglang.srt.speculative.spec_utils import draft_tp_context, load_token_map
@@ -48,6 +51,9 @@ class StandaloneWorker(EAGLEWorker):
             server_args.speculative_algorithm
         )
 
+        # TODO: Adaptive speculative
+        self.adaptive_controller: Optional[AdaptiveController] = None
+
         # Override the context length of the draft model to be the same as the target model.
         server_args.context_length = target_worker.model_runner.model_config.context_len
 

python/sglang/srt/speculative/standalone_worker_v2.py

@@ -8,6 +8,9 @@ from sglang.srt.environ import envs
 from sglang.srt.layers.moe.utils import speculative_moe_backend_context
 from sglang.srt.managers.tp_worker import TpModelWorker
 from sglang.srt.server_args import ServerArgs
+from sglang.srt.speculative.adaptive_runtime_state import (
+    AdaptiveController,
+)
 from sglang.srt.speculative.eagle_utils import TreeMaskMode
 from sglang.srt.speculative.eagle_worker_v2 import EagleDraftWorker, EAGLEWorkerV2
 from sglang.srt.speculative.spec_info import SpeculativeAlgorithm
@@ -183,3 +186,6 @@ class StandaloneWorkerV2(EAGLEWorkerV2):
         self.extend_lens = torch.empty((), dtype=torch.int64, device=self.device)
 
         self.plan_stream, self.plan_stream_ctx = _get_plan_stream(self.device)
+
+        # TODO: Adaptive speculative
+        self.adaptive_controller: Optional[AdaptiveController] = None

test/registered/spec/eagle/test_adaptive_speculative.py

@@ -6,7 +6,6 @@ from types import SimpleNamespace
 
 import requests
 
-from sglang.srt.environ import envs
 from sglang.srt.utils import kill_process_tree
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.run_eval import run_eval
@@ -59,7 +58,6 @@ class TestAdaptiveSpeculativeServer(CustomTestCase):
             cls.adaptive_config_path = f.name
 
         try:
-            with envs.SGLANG_ENABLE_SPEC_V2.override(False):
             cls.process = popen_launch_server(
                 cls.model,
                 cls.base_url,

test/registered/unit/spec/test_adaptive_spec_params.py

@@ -7,14 +7,15 @@ register_cpu_ci(est_time=6, suite="stage-a-test-cpu")
 
 
 class TestAdaptiveSpeculativeParams(unittest.TestCase):
-    def test_initial_steps_snap_to_nearest_candidate_preferring_larger_step(self):
+    def test_initial_steps_added_to_candidates_when_missing(self):
         params = AdaptiveSpeculativeParams(
             initial_steps=2,
             config={"candidate_steps": [1, 3, 7]},
         )
 
-        self.assertEqual(params.current_steps, 3)
+        self.assertEqual(params.candidate_steps, [1, 2, 3, 7])
-        self.assertEqual(params.ema_accept_len, 2.0)
+        self.assertEqual(params.current_steps, 2)
+        self.assertEqual(params.ema_accept_len, 1.0)
 
     def test_update_respects_warmup_and_interval(self):
         params = AdaptiveSpeculativeParams(