[Cleanup] Remove no-longer-used SpeculativeConfig.enable_chunked_prefill (#27826)

Signed-off-by: Nick Hill <nhill@redhat.com>

vllm/config/speculative.py

@@ -78,10 +78,6 @@ class SpeculativeConfig:
     draft_tensor_parallel_size: int | None = Field(default=None, ge=1)
     """The degree of the tensor parallelism for the draft model. Can only be 1
     or the same as the target model's tensor parallel size."""
-    disable_logprobs: bool = True
-    """If set to True, token log probabilities are not returned during
-    speculative decoding. If set to False, token log probabilities are returned
-    according to the log probability settings in SamplingParams."""
 
     # Draft model configuration
     quantization: me_quant.QuantizationMethods | None = None
@@ -126,12 +122,6 @@ class SpeculativeConfig:
     """The configuration of the target model."""
     target_parallel_config: SkipValidation[ParallelConfig] = None  # type: ignore
     """The parallel configuration for the target model."""
-    enable_chunked_prefill: SkipValidation[bool] = None  # type: ignore
-    """Whether vLLM is configured to use chunked prefill or not. Used for
-    raising an error since it's not yet compatible with speculative decode."""
-    disable_log_stats: SkipValidation[bool] = None  # type: ignore
-    """Whether to disable the periodic printing of stage times in speculative
-    decoding."""
 
     # params generated in the post-init stage
     draft_model_config: SkipValidation[ModelConfig] = None  # type: ignore

vllm/engine/arg_utils.py

@@ -1246,8 +1246,6 @@ class EngineArgs:
         self,
         target_model_config: ModelConfig,
         target_parallel_config: ParallelConfig,
-        enable_chunked_prefill: bool,
-        disable_log_stats: bool,
     ) -> SpeculativeConfig | None:
         """Initializes and returns a SpeculativeConfig object based on
         `speculative_config`.
@@ -1267,8 +1265,6 @@ class EngineArgs:
             {
                 "target_model_config": target_model_config,
                 "target_parallel_config": target_parallel_config,
-                "enable_chunked_prefill": enable_chunked_prefill,
-                "disable_log_stats": disable_log_stats,
             }
         )
         return SpeculativeConfig(**self.speculative_config)
@@ -1561,8 +1557,6 @@ class EngineArgs:
         speculative_config = self.create_speculative_config(
             target_model_config=model_config,
             target_parallel_config=parallel_config,
-            enable_chunked_prefill=self.enable_chunked_prefill,
-            disable_log_stats=self.disable_log_stats,
         )
 
         # make sure num_lookahead_slots is set appropriately depending on

vllm/entrypoints/openai/api_server.py

@@ -241,6 +241,7 @@ async def build_async_engine_client_from_engine_args(
         )
 
         # Don't keep the dummy data in memory
+        assert async_llm is not None
         await async_llm.reset_mm_cache()
 
         yield async_llm