[None][feat] Clean up ngram auto mode, add max_concurrency to configs (NVIDIA#6676)

Signed-off-by: Mike Iovine <[email protected]>

Author: mikeiovine

tensorrt_llm/_torch/pyexecutor/py_executor.py

@@ -871,6 +871,10 @@ def _prepare_and_schedule_batch(self):
             self.use_spec_decode = self.drafter.should_use_spec_decode(
                 self.active_requests)
             self.model_engine.enable_spec_decode = self.use_spec_decode
+            # If speculation is off, this function sets py_draft_tokens to None
+            # for all active requests. If it's on, we initialize py_draft_tokens
+            # with dummy draft tokens to make the scheduler aware of the fact
+            # that speculation is about to happen.
             self._prepare_draft_requests()
 
         scheduled_batch, fitting_disagg_gen_init_requests, num_fitting_reqs = self._schedule(

tensorrt_llm/_torch/pyexecutor/py_executor_creator.py

@@ -170,6 +170,14 @@ def _mangle_executor_config(executor_config: ExecutorConfig):
         )
         executor_config.enable_chunked_context = False
 
+    spec_config = executor_config.speculative_config
+    if not executor_config.pytorch_backend_config.disable_overlap_scheduler and spec_config is not None:
+        if not spec_config.spec_dec_mode.support_overlap_scheduler():
+            logger.warning(
+                f"Disable overlap scheduler for speculation mode {spec_config.spec_dec_mode.name}"
+            )
+            executor_config.pytorch_backend_config.disable_overlap_scheduler = True
+
 
 def _get_mapping(executor_config: ExecutorConfig) -> Mapping:
     if executor_config.mapping is None:

tensorrt_llm/_torch/speculative/__init__.py

@@ -1,3 +1,4 @@
+from .auto_heuristic import suggest_spec_config
 from .eagle3 import Eagle3SpecMetadata
 from .interface import SpecMetadata
 from .mtp import MTPEagleWorker, MTPSpecMetadata, MTPWorker
@@ -23,4 +24,5 @@
     "get_spec_resource_manager",
     "get_spec_worker",
     "update_spec_config_from_model_config",
+    "suggest_spec_config",
 ]

tensorrt_llm/_torch/speculative/auto_heuristic.py

@@ -0,0 +1,17 @@
+def suggest_spec_config(max_batch_size: int) -> "DecodingBaseConfig":
+    """
+    Suggests a reasonable draft model free speculation scheme.
+    Used when the user specifies spec_mode == AUTO.
+
+    For now, we always use an ngram scheme that gets disabled at
+    BS>=32.
+    """
+    from tensorrt_llm.llmapi.llm_args import NGramDecodingConfig
+    return NGramDecodingConfig(
+        max_draft_len=5 if max_batch_size <= 4 else 3,
+        max_matching_ngram_size=3 if max_batch_size <= 4 else 5,
+        max_concurrency=32,
+        is_keep_all=True,
+        is_use_oldest=True,
+        is_public_pool=True,
+    )

tensorrt_llm/_torch/speculative/drafter.py

@@ -9,6 +9,9 @@
 class Drafter(ABC):
     """Abstract base class for all drafter implementations."""
 
+    def __init__(self, max_concurrency: Optional[int] = None) -> None:
+        self.max_concurrency = max_concurrency
+
     @abstractmethod
     def prepare_draft_tokens(
         self,
@@ -25,4 +28,6 @@ def prepare_draft_tokens(
 
     def should_use_spec_decode(self, requests: List[LlmRequest]) -> bool:
         """Check if spec decode should be used for the current iteration."""
+        if self.max_concurrency is not None:
+            return len(requests) <= self.max_concurrency
         return True

tensorrt_llm/_torch/speculative/model_drafter.py

@@ -48,6 +48,8 @@ def __init__(
         spec_resource_manager: Optional[BaseResourceManager] = None,
         guided_decoder: Optional[GuidedDecoder] = None,
     ):
+        super().__init__(spec_config.max_concurrency)
+
         # Validate required parameters
         if draft_model_engine is None:
             raise ValueError("draft_model_engine cannot be None")

tensorrt_llm/_torch/speculative/ngram.py

@@ -168,6 +168,7 @@ def __init__(
         spec_config: NGramDecodingConfig,
         ngram_pool_manager: NGramPoolManager = None,
     ):
+        super().__init__(spec_config.max_concurrency)
         assert ngram_pool_manager is not None, "NGram needs a resource manager to maintain the pool."
         self.spec_config = spec_config
         self.max_draft_len = spec_config.max_draft_len
@@ -178,11 +179,6 @@ def prepare_draft_tokens(
         scheduled_requests: ScheduledRequests,
         resource_manager: Optional[ResourceManager] = None,
     ) -> None:
-        # Disable NGram speculative decoding auto heuristic for batch size > 32.
-        if self.spec_config.is_auto_heuristic and len(
-                scheduled_requests.all_requests()) > 32:
-            return
-
         # Sort by request_id when py_batch_idx is None as a fallback.
         # This happens in the disagg case: for a set of new requests, we draft
         # before forward_step, so py_batch_idx is not assigned.

tensorrt_llm/llmapi/llm.py

@@ -32,8 +32,8 @@
 from ..sampling_params import SamplingParams
 from ..scheduling_params import SchedulingParams
 from .llm_args import (TORCH_LLMARGS_EXPLICIT_DOCSTRING,
-                       TRT_LLMARGS_EXPLICIT_DOCSTRING, NGramDecodingConfig,
-                       PeftCacheConfig, PybindMirror, TorchLlmArgs, TrtLlmArgs)
+                       TRT_LLMARGS_EXPLICIT_DOCSTRING, PeftCacheConfig,
+                       PybindMirror, TorchLlmArgs, TrtLlmArgs)
 from .llm_utils import (CachedModelLoader, KvCacheRetentionConfig,
                         LlmBuildStats, ModelLoader, _ModelRuntimeContext)
 from .mpi_session import MpiPoolSession, external_mpi_comm_available
@@ -1015,32 +1015,10 @@ def _build_model(self):
 
         spec_config = self.args.speculative_config
         max_batch_size = self._executor_config.max_batch_size
-        # Apply default heuristic to AutoDecodingConfig based on benchmark results
-        # With concurrency <= 4, max_draft_len = 5, max_matching_ngram_size = 3
-        # With concurrency <= 32, max_draft_len = 3, max_matching_ngram_size = 5
-        # With concurrency > 32, speculative decoding is disabled.
-        if spec_config is not None and spec_config.decoding_type == "AUTO":
-            if not self.args.disable_overlap_scheduler:
-                logger.info(
-                    "Disable overlap scheduler to enable Auto speculative decoding with Ngram."
-                )
-                # From benchmark results, we found that NGram speculative decoding provides better performance than overlap scheduler with low concurrency <= 32.
-                # Therefore, we disable overlap scheduler to enable NGram speculative decoding.
-                self.args.disable_overlap_scheduler = True
-
-            spec_config = NGramDecodingConfig(
-                max_draft_len=5 if max_batch_size <= 4 else 3,
-                max_matching_ngram_size=3 if max_batch_size <= 4 else 5,
-                is_keep_all=True,
-                is_use_oldest=True,
-                is_public_pool=True,
-                # Flag to indicate the NGramDecodingConfig is instantiated by auto heuristic.
-                is_auto_heuristic=True,
-            )
 
-            logger.info(
-                f"Apply heuristic to incomplete NGramDecodingConfig: max_draft_len={spec_config.max_draft_len}, max_matching_ngram_size={spec_config.max_matching_ngram_size}"
-            )
+        if spec_config is not None and spec_config.decoding_type == "AUTO":
+            from tensorrt_llm._torch.speculative import suggest_spec_config
+            spec_config = suggest_spec_config(max_batch_size)
 
         update_executor_config(
             self._executor_config,

tensorrt_llm/llmapi/llm_args.py

@@ -342,6 +342,11 @@ class DecodingBaseConfig(StrictBaseModel):
     max_draft_len: Optional[int] = None
     speculative_model_dir: Optional[Union[str, Path]] = None
 
+    # PyTorch only.
+    # When specified, speculation will be disabled at batch sizes above
+    # this value. Otherwise, speculation will always be on.
+    max_concurrency: Optional[int] = None
+
     @classmethod
     def from_dict(cls, data: dict):
         # dispatch to the correct decoding config
@@ -469,9 +474,6 @@ class NGramDecodingConfig(DecodingBaseConfig):
     is_keep_all: bool = True
     is_use_oldest: bool = True
     is_public_pool: bool = True
-    # Flag to indicate the NGramDecodingConfig is instantiated by auto heuristic.
-    # User should not set this flag. Use AutoDecodingConfig instead.
-    is_auto_heuristic: bool = False
 
     @classmethod
     def from_dict(cls, data: dict):
@@ -535,13 +537,10 @@ class AutoDecodingConfig(DecodingBaseConfig):
     """
     Configuration for auto speculative decoding.
 
-    This config is used to automatically select the best speculative decoding algorithm.
+    This config will automatically select a good, draft-model free
+    speculation algorithm with some heuristic.
 
-    According to benchmark results, the best algorithm in general is NGRAM with low concurrency <= 32.
-    Default heuristic:
-        With concurrency <= 4, max_draft_len = 5, max_matching_ngram_size = 3
-        With concurrency <= 32, max_draft_len = 3, max_matching_ngram_size = 5
-        With concurrency > 32, speculative decoding is disabled.
+    Attributes that are inherited from the base class are ignored.
     """
 
     @classmethod