Merge branch 'v0.8.5-zero_overhead' into 'v0.8.5.post1-dev'

tbo增加ds int8支持，修改tb zerooverhead环境变量到envs.py中统一管理

See merge request dcutoolkit/deeplearing/vllm!120

vllm/engine/multiprocessing/engine.py

@@ -6,8 +6,8 @@ from contextlib import contextmanager
 from typing import Iterator, List, Optional, Union
 
 import cloudpickle
+import vllm.envs as envs
 from vllm.zero_overhead.llm_engine import ZeroOverheadEngine
-from vllm.zero_overhead.utils import is_zero_overhead
 import zmq
 
 from vllm import AsyncEngineArgs, SamplingParams
@@ -81,7 +81,7 @@ class MQLLMEngine:
         # the python object to be reused again.
         kwargs['use_cached_outputs'] = True
 
-        if is_zero_overhead():
+        if envs.VLLM_ZERO_OVERHEAD:
             self.engine = ZeroOverheadEngine(*args, **kwargs)
         else:
             self.engine = LLMEngine(*args, **kwargs)

vllm/entrypoints/llm.py

@@ -43,8 +43,8 @@ from vllm.transformers_utils.tokenizer import (AnyTokenizer, MistralTokenizer,
 from vllm.usage.usage_lib import UsageContext
 from vllm.utils import (Counter, Device, deprecate_args, deprecate_kwargs,
                         is_list_of)
+import vllm.envs as envs
 from vllm.zero_overhead.llm_engine import ZeroOverheadEngine
-from vllm.zero_overhead.utils import is_zero_overhead
 
 logger = init_logger(__name__)
 
@@ -248,7 +248,7 @@ class LLM:
         )
 
         # Create the Engine (autoselects V0 vs V1)
-        if is_zero_overhead():
+        if envs.VLLM_ZERO_OVERHEAD:
             self.llm_engine = ZeroOverheadEngine.from_engine_args(
                 engine_args=engine_args, usage_context=UsageContext.LLM_CLASS)
         else:

vllm/envs.py

@@ -125,6 +125,8 @@ if TYPE_CHECKING:
     VLLM_ENFORCE_EAGER_BS_THRESHOLD: Optional[int] = None
     VLLM_HAS_CONTEXT_DEFAULT: bool = False
 
+    VLLM_ENABLE_TBO: bool = False
+    VLLM_ZERO_OVERHEAD: bool = False
 
 def get_default_cache_root():
     return os.getenv(
@@ -796,11 +798,22 @@ environment_variables: dict[str, Callable[[], Any]] = {
     # If set, vLLM will disable the draft model in cudagraph mode.
     "VLLM_ENFORCE_EAGER_BS_THRESHOLD":
     lambda: int(os.environ.get("VLLM_ENFORCE_EAGER_BS_THRESHOLD", "-1")),
-    
-    # If set, vLLM can avoid Device2Host copy during MLA prefill phase
+        
+    # Enable two batch overlap.
+    "VLLM_ENABLE_TBO":
+    lambda: bool(int(os.getenv("VLLM_ENABLE_TBO", "0"))),
+
+    # Enable zero overhead scheduler.
+    "VLLM_ZERO_OVERHEAD":
+    lambda: bool(int(os.getenv("VLLM_ZERO_OVERHEAD", "0"))),
+
+    # 'has_comtext' is a variable in common.py, which is calculated
+    # by metadata by default. However, it may introduce synchronization 
+    # and affect performance, so it is directly assigned as False. 
+    # If there are any problems during use, use environment variables 
+    # to restore the default usage.
     "VLLM_HAS_CONTEXT_DEFAULT":
-    lambda: bool(int(os.environ.get("VLLM_HAS_CONTEXT_DEFAULT", "0"))),
-    
+    lambda: bool(int(os.getenv("VLLM_HAS_CONTEXT_DEFAULT", "0"))),
 }
 
 # end-env-vars-definition

vllm/forward_context.py

@@ -30,10 +30,6 @@ forward_start_time: float = 0
 batchsize_logging_interval: float = envs.VLLM_LOG_BATCHSIZE_INTERVAL
 batchsize_forward_time: defaultdict = defaultdict(list)
 
-enable_tbo = os.environ.get('VLLM_ENABLE_TBO') == '1'
-def is_enable_tbo():
-    return enable_tbo
-
 @dataclass
 class DPMetadata:
     cu_tokens_across_dp_cpu: torch.Tensor
@@ -55,7 +51,7 @@ _forward_context: Optional[ForwardContext] = None
 
 
 def get_forward_context() -> ForwardContext:
-    if is_enable_tbo():
+    if envs.VLLM_ENABLE_TBO:
         forward_context = get_tbo_forward_context()
         """Get the current forward context."""
         assert forward_context is not None, (
@@ -125,7 +121,7 @@ def set_forward_context(attn_metadata: Any,
         kv_connector = get_kv_transfer_group()
         assert isinstance(kv_connector, KVConnectorBase_V1)
         kv_connector.start_load_kv(_forward_context)
-    if is_enable_tbo():
+    if envs.VLLM_ENABLE_TBO:
         set_tbo_forward_context(_forward_context)
     try:
         yield
@@ -171,5 +167,5 @@ def set_forward_context(attn_metadata: Any,
             kv_connector.wait_for_save()
 
         _forward_context = prev_context
-        if is_enable_tbo():
+        if envs.VLLM_ENABLE_TBO:
             set_tbo_forward_context(_forward_context)

vllm/model_executor/layers/fused_moe/layer.py

@@ -554,9 +554,8 @@ class FusedMoE(torch.nn.Module):
 
 
         self.quant_method.create_weights(layer=self, **moe_quant_params)
-        from vllm.two_batch_overlap.two_batch_overlap import tbo_all_reduce, is_enable_tbo
+        from vllm.two_batch_overlap.two_batch_overlap import tbo_all_reduce
         self.tbo_all_reduce = tbo_all_reduce
-        self.enable_tbo = is_enable_tbo()
 
     def _load_per_tensor_weight_scale(self, shard_id: str,
                                       param: torch.nn.Parameter,
@@ -940,7 +939,7 @@ class FusedMoE(torch.nn.Module):
 
         if self.reduce_results and (self.tp_size > 1 or self.ep_size > 1):
             # Default set to False. (May have to add shared expert outputs.)
-            if self.enable_tbo:
+            if envs.VLLM_ENABLE_TBO:
                 final_hidden_states = self.tbo_all_reduce(final_hidden_states)
             else:
                 final_hidden_states = tensor_model_parallel_all_reduce(

vllm/model_executor/layers/linear.py

@@ -3,7 +3,7 @@
 import itertools
 from abc import abstractmethod
 from typing import Any, Literal, Optional, Union
-
+import vllm.envs as envs
 import torch
 import torch.nn as nn
 from torch.nn.parameter import Parameter, UninitializedParameter
@@ -1237,9 +1237,8 @@ class RowParallelLinear(LinearBase):
             })
         else:
             self.register_parameter("bias", None)
-        from vllm.two_batch_overlap.two_batch_overlap import tbo_all_reduce, is_enable_tbo
+        from vllm.two_batch_overlap.two_batch_overlap import tbo_all_reduce
         self.tbo_all_reduce = tbo_all_reduce
-        self.enable_tbo = is_enable_tbo()
 
     def weight_loader(self, param: Parameter, loaded_weight: torch.Tensor):
         tp_rank = get_tensor_model_parallel_rank()
@@ -1310,7 +1309,7 @@ class RowParallelLinear(LinearBase):
                                                   input_parallel,
                                                   bias=bias_)
         if self.reduce_results and self.tp_size > 1:
-            if self.enable_tbo:
+            if envs.VLLM_ENABLE_TBO:
                 output = self.tbo_all_reduce(output_parallel)
             else:
                 output = tensor_model_parallel_all_reduce(output_parallel)

vllm/model_executor/layers/sampler.py

@@ -21,7 +21,6 @@ from vllm.sequence import (VLLM_INVALID_TOKEN_ID,
                            CompletionSequenceGroupOutput, Logprob,
                            PromptLogprobs, SampleLogprobs, SequenceOutput)
 from vllm.spec_decode.metrics import SpecDecodeWorkerMetrics
-from vllm.zero_overhead.utils import is_zero_overhead
 
 if envs.VLLM_USE_FLASHINFER_SAMPLER and find_spec("flashinfer"):
     import flashinfer.sampling
@@ -39,7 +38,7 @@ def get_sampler() -> torch.nn.Module:
         # Lazy import: the v1 package isn't distributed
         from vllm.v1.sample.sampler import Sampler as V1Sampler
         return V1Sampler()
-    if is_zero_overhead():
+    if envs.VLLM_ZERO_OVERHEAD:
         from vllm.zero_overhead.sampler import ZeroOverheadSampler
         return ZeroOverheadSampler()
     return Sampler()

vllm/model_executor/layers/vocab_parallel_embedding.py

@@ -2,7 +2,7 @@
 
 from dataclasses import dataclass
 from typing import List, Optional, Sequence, Tuple
-
+import vllm.envs as envs
 import os
 import torch
 import torch.nn.functional as F
@@ -283,9 +283,8 @@ class VocabParallelEmbedding(torch.nn.Module):
                                          self.num_embeddings_padded,
                                          params_dtype=params_dtype,
                                          weight_loader=self.weight_loader)
-        from vllm.two_batch_overlap.two_batch_overlap import tbo_all_reduce, is_enable_tbo
+        from vllm.two_batch_overlap.two_batch_overlap import tbo_all_reduce
         self.tbo_all_reduce = tbo_all_reduce
-        self.enable_tbo = is_enable_tbo()
 
     @classmethod
     def _get_indices(cls, vocab_size_padded: int, org_vocab_size_padded: int,
@@ -437,7 +436,7 @@ class VocabParallelEmbedding(torch.nn.Module):
         if self.tp_size > 1:
             output_parallel.masked_fill_(input_mask.unsqueeze(-1), 0)
         # Reduce across all the model parallel GPUs.
-        if self.enable_tbo:
+        if envs.VLLM_ENABLE_TBO:
             output = self.tbo_all_reduce(output_parallel)
         else:
             output = tensor_model_parallel_all_reduce(output_parallel)

vllm/model_executor/models/deepseek_v2.py

@@ -155,9 +155,8 @@ class DeepseekV2MoE(nn.Module):
                 reduce_results=False,
                 prefix=f"{prefix}.shared_experts",
             )
-        from vllm.two_batch_overlap.two_batch_overlap import tbo_all_reduce, is_enable_tbo
+        from vllm.two_batch_overlap.two_batch_overlap import tbo_all_reduce
         self.tbo_all_reduce = tbo_all_reduce
-        self.enable_tbo = is_enable_tbo()
 
     def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
         num_tokens, hidden_dim = hidden_states.shape
@@ -191,7 +190,7 @@ class DeepseekV2MoE(nn.Module):
         #         final_hidden_states = final_hidden_states + shared_output \
         #             * (1. / self.routed_scaling_factor)
         if self.tp_size > 1:
-            if self.enable_tbo:
+            if envs.VLLM_ENABLE_TBO:
                 final_hidden_states = self.tbo_all_reduce(final_hidden_states)
             else:
                 final_hidden_states = tensor_model_parallel_all_reduce(

vllm/model_executor/models/deepseek_v3.py

@@ -23,7 +23,7 @@
 import os
 import re
 from typing import Any, Dict, Iterable, List, Optional, Set, Tuple, Union
-
+import vllm.envs as envs
 import torch
 from torch import nn
 from transformers import PretrainedConfig
@@ -150,9 +150,8 @@ class DeepseekV3MoE(nn.Module):
                 quant_config=quant_config,
                 reduce_results=False,
             )
-        from vllm.two_batch_overlap.two_batch_overlap import tbo_all_reduce, is_enable_tbo
+        from vllm.two_batch_overlap.two_batch_overlap import tbo_all_reduce
         self.tbo_all_reduce = tbo_all_reduce
-        self.enable_tbo = is_enable_tbo()
 
     def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
         num_tokens, hidden_dim = hidden_states.shape
@@ -167,7 +166,7 @@ class DeepseekV3MoE(nn.Module):
         if shared_output is not None:
             final_hidden_states = final_hidden_states + shared_output
         if self.tp_size > 1:
-            if self.enable_tbo:
+            if envs.VLLM_ENABLE_TBO:
                 final_hidden_states = self.tbo_all_reduce(final_hidden_states)
             else:
                 final_hidden_states = tensor_model_parallel_all_reduce(

vllm/spec_decode/spec_decode_worker.py

@@ -54,7 +54,7 @@ from vllm.worker.worker_base import LoRANotSupportedWorkerBase, WorkerBase
 from vllm.worker.cache_engine import CacheEngine
 from vllm.attention.ops.paged_attn import PagedAttention
 from vllm.spec_decode.proposer_worker_base import NonLLMProposerWorkerBase
-from vllm.zero_overhead.utils import is_zero_overhead
+import vllm.envs as envs
 
 logger = init_logger(__name__)
 
@@ -207,7 +207,7 @@ class SpecDecodeWorker(LoRANotSupportedWorkerBase):
                 # Load lm_head weight for eagle in init_device
                 if draft_model_config.hf_config.model_type == "eagle":
                     enable_lm_head_weight_load = True
-                if is_zero_overhead():
+                if envs.VLLM_ZERO_OVERHEAD:
                     assert False, (
                                 "speculative decoding not support zero overhead scheduler yet"
                             )
@@ -261,7 +261,7 @@ class SpecDecodeWorker(LoRANotSupportedWorkerBase):
                     "[Speculative Decoding] Disabling MQA scorer as the "
                     "target model is not running in eager mode.")
 
-        if is_zero_overhead():
+        if envs.VLLM_ZERO_OVERHEAD:
             from vllm.zero_overhead.spec_decode.spec_decode_worker import ZeroOverheadSpecDecodeWorker
             return ZeroOverheadSpecDecodeWorker(
                 proposer_worker,

vllm/two_batch_overlap/model_input_split.py

 import torch
 from vllm.attention.backends.flashmla import FlashMLAMetadata
+from vllm.attention.backends.mla.common import MLACommonMetadata
 from vllm.attention.backends.rocm_flash_attn import ROCmFlashAttentionMetadata
 from vllm.model_executor.sampling_metadata import SamplingMetadata
 from vllm.utils import async_tensor_h2d
@@ -12,6 +13,11 @@ def cumsum(lst):
         cum_lst.append(sum)
     return cum_lst
 
+def is_supported_attention_metadata(atten_metadata):
+    return isinstance(atten_metadata, ROCmFlashAttentionMetadata) or \
+            isinstance(atten_metadata, FlashMLAMetadata) or \
+            isinstance(atten_metadata, MLACommonMetadata)
+
 def split_model_input(model_input, self_device, batch_size_left, batch_size_right):
     from vllm.worker.model_runner import ModelInputForGPUWithSamplingMetadata
     query_tokens_split = [sum(model_input.query_lens[0:batch_size_left]), sum(model_input.query_lens[batch_size_left:])]
@@ -93,6 +99,66 @@ def split_model_input(model_input, self_device, batch_size_left, batch_size_righ
     selected_token_indices_left = split_seq_lens_tensor[0].cumsum(dim=0) - 1
     selected_token_indices_right = split_seq_lens_tensor[1].cumsum(dim=0) - 1
     
+    if isinstance(model_input.attn_metadata, MLACommonMetadata):
+        attn_metadata_left = MLACommonMetadata(
+            num_prefills = num_prefills_left,
+            num_prefill_tokens = num_prefill_tokens_left,
+            num_decode_tokens = num_decode_tokens_left,
+            slot_mapping = split_slot_mapping[0],
+            multi_modal_placeholder_index_maps = model_input.attn_metadata.multi_modal_placeholder_index_maps,
+            enable_kv_scales_calculation = model_input.attn_metadata.enable_kv_scales_calculation,
+            use_cuda_graph = model_input.attn_metadata.use_cuda_graph,
+            input_positions = split_input_positions[0],
+            seq_lens = seq_lens_left,
+            seq_lens_tensor = split_seq_lens_tensor[0],
+            max_prefill_seq_len = max_prefill_seq_len_left,
+            max_decode_seq_len = max_decode_seq_len_left,
+            context_lens_tensor = split_context_lens_tensor[0],
+            block_tables = split_block_tables[0],
+            max_query_len = max_query_len_left,
+            max_decode_query_len = max_decode_query_len_left,
+            query_start_loc = query_start_loc_left,
+            seq_start_loc = seq_start_loc_left,
+            _cached_prefill_metadata = None,
+            _cached_decode_metadata = None,
+            head_dim = model_input.attn_metadata.head_dim,
+            is_profile_run = model_input.attn_metadata.is_profile_run,
+            context_chunk_cu_seq_lens=model_input.attn_metadata.context_chunk_cu_seq_lens, 
+            context_chunk_starts=model_input.attn_metadata.context_chunk_starts, 
+            context_chunk_seq_tot=model_input.attn_metadata.context_chunk_seq_tot, 
+            context_chunk_max_seq_lens=model_input.attn_metadata.context_chunk_max_seq_lens, 
+            context_chunk_workspace=model_input.attn_metadata.context_chunk_workspace, 
+        )
+        attn_metadata_right = MLACommonMetadata(
+            num_prefills = num_prefills_right,
+            num_prefill_tokens = num_prefill_tokens_right,
+            num_decode_tokens = num_decode_tokens_right,
+            slot_mapping = split_slot_mapping[1],
+            multi_modal_placeholder_index_maps = model_input.attn_metadata.multi_modal_placeholder_index_maps,
+            enable_kv_scales_calculation = model_input.attn_metadata.enable_kv_scales_calculation,
+            use_cuda_graph = model_input.attn_metadata.use_cuda_graph,
+            input_positions = split_input_positions[1],
+            seq_lens = seq_lens_right,
+            seq_lens_tensor = split_seq_lens_tensor[1],
+            max_prefill_seq_len = max_prefill_seq_len_right,
+            max_decode_seq_len = max_decode_seq_len_right,
+            context_lens_tensor = split_context_lens_tensor[1],
+            block_tables = split_block_tables[1],
+            max_query_len = max_query_len_right,
+            max_decode_query_len = max_decode_query_len_right,
+            query_start_loc = query_start_loc_right,
+            seq_start_loc = seq_start_loc_right,
+            _cached_prefill_metadata = None,
+            _cached_decode_metadata = None,
+            head_dim = model_input.attn_metadata.head_dim,
+            is_profile_run = model_input.attn_metadata.is_profile_run,
+            context_chunk_cu_seq_lens=model_input.attn_metadata.context_chunk_cu_seq_lens, 
+            context_chunk_starts=model_input.attn_metadata.context_chunk_starts, 
+            context_chunk_seq_tot=model_input.attn_metadata.context_chunk_seq_tot, 
+            context_chunk_max_seq_lens=model_input.attn_metadata.context_chunk_max_seq_lens, 
+            context_chunk_workspace=model_input.attn_metadata.context_chunk_workspace, 
+        )
+    
     if isinstance(model_input.attn_metadata, ROCmFlashAttentionMetadata):
         block_tables_list_left = model_input.attn_metadata.block_tables_list[0:batch_size_left]
         block_tables_list_right = model_input.attn_metadata.block_tables_list[batch_size_left:]

vllm/two_batch_overlap/two_batch_overlap.py

@@ -3,18 +3,14 @@ import os
 import queue
 import threading
 import torch
-from vllm.attention.backends.flashmla import FlashMLAMetadata
-from vllm.attention.backends.rocm_flash_attn import ROCmFlashAttentionMetadata
 from vllm.distributed.communication_op import tensor_model_parallel_all_reduce
 from vllm.forward_context import set_forward_context
 from vllm.multimodal.inputs import MultiModalKwargs
 from vllm.two_batch_overlap.forward_context import init_tbo_forward_context
-from vllm.two_batch_overlap.model_input_split import split_model_input
-from vllm.utils import async_tensor_h2d
+from vllm.two_batch_overlap.model_input_split import is_supported_attention_metadata, split_model_input
 from vllm.logger import init_logger
 from vllm.profiler.prof import profile
-
-enable_tbo = os.environ.get('VLLM_ENABLE_TBO') == '1'
+from vllm import envs
 
 enable_tbo_decode = os.environ.get('VLLM_TBO_DECODE') == '1'
 
@@ -22,9 +18,6 @@ tbo_one_stream = os.environ.get('VLLM_TBO_ONE_STREAM') == '1'
 
 logger = init_logger(__name__)
 
-def is_enable_tbo():
-    return enable_tbo
-
 tbo_step_stream = None
 all_reduce_stream = None
 
@@ -62,6 +55,7 @@ class TwoBatchOverlap():
         self.right_thread = threading.Thread(target=self.thread_two_batch_overlap, args=(self.model_input_right_queue,))
         self.left_thread.start()
         self.right_thread.start()
+        logger.info('tbo:two batch overlap threads start')
 
     def finish_thread(self):
         if self.left_thread != None:
@@ -81,11 +75,9 @@ class TwoBatchOverlap():
         if queue == self.model_input_left_queue:
             self.left_tid = tid
             is_left_thread = True
-            logger.info('tbo:new thread %d', self.left_tid)
             init_tbo_forward_context(True, self.left_tid)
         else:
             self.right_tid = tid
-            logger.info('tbo:new thread %d', self.right_tid)
             init_tbo_forward_context(False, self.right_tid)
         with torch.cuda.stream(tbo_step_stream):
             while True:
@@ -177,7 +169,7 @@ class TwoBatchOverlap():
 tbo_obj = None
 
 def init_two_batch_overlap():
-    if enable_tbo:
+    if envs.VLLM_ENABLE_TBO:
         global tbo_obj
         if tbo_obj == None:
             tbo_obj = TwoBatchOverlap()
@@ -189,7 +181,7 @@ def finish_two_batch_overlap():
         tbo_obj = None
 
 def tbo_all_reduce(obj):
-    if enable_tbo and tbo_obj != None and tbo_obj.tbo_running:
+    if envs.VLLM_ENABLE_TBO and tbo_obj != None and tbo_obj.tbo_running:
         tid = threading.get_ident()
         if not tbo_one_stream:
             if tid == tbo_obj.left_tid:
@@ -219,14 +211,14 @@ def tbo_model_executable(
         seqlen_agnostic_kwargs,
         model_kwargs,
     ):
-    init_two_batch_overlap()
-    is_rocm_fa = isinstance(model_input.attn_metadata, ROCmFlashAttentionMetadata)
-    is_mla_fa = isinstance(model_input.attn_metadata, FlashMLAMetadata)
+    is_support = is_supported_attention_metadata(model_input.attn_metadata)
+    if not is_support:
+        logger.info("tbo:not surpport yet ", type(model_input.attn_metadata))
     is_cuda_graph_decode = model_input.attn_metadata.use_cuda_graph and not model_input.is_prompt    
     batch_size = len(model_input.attn_metadata.seq_lens)
     if batch_size == 1 or \
         (not model_input.is_prompt and not enable_tbo_decode) or \
-        not (is_rocm_fa or is_mla_fa) or \
+        not is_support or \
         is_cuda_graph_decode:
         with set_forward_context(model_input.attn_metadata,
                                     vllm_config, virtual_engine):
@@ -241,6 +233,7 @@ def tbo_model_executable(
             )
         return hidden_or_intermediate_states
     profile.ProfRangePush('tbo_model_executable')
+    init_two_batch_overlap()
     tbo_obj.tbo_running = True
     tbo_obj.left_first = True
     batch_size_left = int(batch_size / 2)

vllm/worker/model_runner.py

@@ -50,7 +50,7 @@ from vllm.prompt_adapter.worker_manager import (
     LRUCacheWorkerPromptAdapterManager)
 from vllm.sampling_params import SamplingParams
 from vllm.sequence import IntermediateTensors, SequenceGroupMetadata
-from vllm.two_batch_overlap.two_batch_overlap import is_enable_tbo, tbo_model_executable
+from vllm.two_batch_overlap.two_batch_overlap import tbo_model_executable
 from vllm.utils import (DeviceMemoryProfiler, GiB_bytes, PyObjectCache,
                         async_tensor_h2d, flatten_2d_lists,
                         is_pin_memory_available, supports_dynamo,
@@ -61,7 +61,6 @@ from vllm.worker.model_runner_base import (
     _add_sampling_metadata_broadcastable_dict,
     _init_attn_metadata_from_tensor_dict,
     _init_sampling_metadata_from_tensor_dict)
-from vllm.zero_overhead.utils import is_zero_overhead
 
 if TYPE_CHECKING:
     from vllm.attention.backends.abstract import AttentionBackend
@@ -1640,7 +1639,7 @@ class ModelRunner(GPUModelRunnerBase[ModelInputForGPUWithSamplingMetadata]):
     _model_input_cls: Type[ModelInputForGPUWithSamplingMetadata] = (
         ModelInputForGPUWithSamplingMetadata)
     _builder_cls: Type[ModelInputForGPUBuilder] = ModelInputForGPUBuilder
-    if is_zero_overhead():
+    if envs.VLLM_ZERO_OVERHEAD:
         from vllm.zero_overhead.model_runner import ZeroOverheadModelInputForGpuBuilder
         _builder_cls = ZeroOverheadModelInputForGpuBuilder
 
@@ -1779,7 +1778,7 @@ class ModelRunner(GPUModelRunnerBase[ModelInputForGPUWithSamplingMetadata]):
             model_forward_start.record()
 
         if not bypass_model_exec:
-            if is_enable_tbo():
+            if envs.VLLM_ENABLE_TBO:
                 hidden_or_intermediate_states = tbo_model_executable(
                     model_input, 
                     self.vllm_config,

vllm/zero_overhead/utils.py

@@ -3,15 +3,12 @@
 from enum import Enum
 import os
 import torch
+import vllm.envs as envs
 
-zero_overhead = os.environ.get('VLLM_ZERO_OVERHEAD') == '1'
 zero_no_thread = os.environ.get('VLLM_ZERO_NO_THREAD') == '1'
 
-def is_zero_overhead():
-    return zero_overhead
-
 def is_zero_no_thread():
-    return zero_no_thread and zero_overhead
+    return zero_no_thread and envs.VLLM_ZERO_OVERHEAD
 
 class SpecStepKind(Enum):
     KIND_DEFAULT = 0