Revert "."

This reverts commit a5e0ea09ea.

vllm/envs.py

@@ -245,7 +245,6 @@ if TYPE_CHECKING:
     VLLM_DEBUG_WORKSPACE: bool = False
     VLLM_DISABLE_SHARED_EXPERTS_STREAM: bool = False
     VLLM_SHARED_EXPERTS_STREAM_TOKEN_THRESHOLD: int = 256
-    VLLM_ENABLE_MULTI_STREAM_GEMM: bool = False
     VLLM_COMPILE_CACHE_SAVE_FORMAT: Literal["binary", "unpacked"] = "binary"
     VLLM_USE_V2_MODEL_RUNNER: bool = False
     VLLM_LOG_MODEL_INSPECTION: bool = False
@@ -1663,13 +1662,6 @@ environment_variables: dict[str, Callable[[], Any]] = {
     "VLLM_SHARED_EXPERTS_STREAM_TOKEN_THRESHOLD": lambda: int(
         int(os.getenv("VLLM_SHARED_EXPERTS_STREAM_TOKEN_THRESHOLD", 256))
     ),
-    # Enables multi-stream overlap of the attention input GEMM with auxiliary
-    # GEMMs (e.g. fused_wqa_wkv overlapped with indexer weights / kv-score
-    # projections in DeepSeek-V4). When unset, those callables run
-    # sequentially on the current stream even if aux streams are provided.
-    "VLLM_ENABLE_MULTI_STREAM_GEMM": lambda: bool(
-        int(os.getenv("VLLM_ENABLE_MULTI_STREAM_GEMM", "0"))
-    ),
     # Format for saving torch.compile cache artifacts
     # - "binary": saves as binary file
     #     Safe for multiple vllm serve processes accessing the same torch compile cache.

vllm/model_executor/layers/deepseek_v4_attention.py

@@ -13,7 +13,6 @@ import torch.nn as nn
 import torch.nn.functional as F
 from transformers import DeepseekV2Config, DeepseekV3Config
 
-import vllm.envs as envs
 from vllm.model_executor.layers.linear import (
     ReplicatedLinear,
 )
@@ -383,7 +382,6 @@ class DeepseekV4MultiHeadLatentAttentionWrapper(PluggableLayer):
             self.ln_events[0],
             self.ln_events[1:4],
             self.aux_stream_list[:3],
-            enable=envs.VLLM_ENABLE_MULTI_STREAM_GEMM,
         )
 
         return qr_kv, kv_score, indexer_kv_score, indexer_weights

vllm/utils/multi_stream_utils.py

@@ -64,7 +64,6 @@ def execute_in_parallel(
     start_event: torch.cuda.Event,
     done_events: list[torch.cuda.Event],
     aux_streams: list[torch.cuda.Stream] | None = None,
-    enable: bool = False,
 ) -> tuple[Any, list[Any]]:
     """Run default_fn on the current stream and aux_fns concurrently on
     aux_streams.
@@ -75,9 +74,8 @@ def execute_in_parallel(
 
     start_event fans out from the current stream to every launched aux stream;
     done_events[i] is recorded after aux_fns[i] so the current stream joins
-    before returning. Falls back to sequential execution on the current stream
-    when aux_streams is None or enable is False; in that case default_fn runs
-    first, then aux_fns in order.
+    before returning. When aux_streams is None, all aux_fns run sequentially
+    on the current stream.
 
     Args:
         default_fn: Callable for the default (current) stream.
@@ -88,17 +86,13 @@ def execute_in_parallel(
             corresponding aux_fn. Length must match aux_fns.
         aux_streams: Per-aux CUDA streams. Length must match aux_fns.
             Multi-stream is disabled when None.
-        enable: Opt-in switch for the multi-stream path. Defaults to False,
-            so callers that pass aux_streams must also pass enable=True
-            (typically gated by an env var) to actually overlap. When False,
-            execution falls back to sequential on the current stream.
 
     Returns:
         Tuple of (default_result, aux_results) where aux_results[i] is the
         result of aux_fns[i] (or None when skipped).
     """
     aux_results: list[Any]
-    if aux_streams is None or not enable:
+    if aux_streams is None:
         default_result = default_fn()
         aux_results = [fn() if fn is not None else None for fn in aux_fns]
         return default_result, aux_results