[V0 Deprecation] Remove Prompt Adapters (#20588)

Signed-off-by: mgoin <mgoin64@gmail.com>

vllm/engine/llm_engine.py

@@ -44,7 +44,6 @@ from vllm.multimodal.processing import EncDecMultiModalProcessor
 from vllm.outputs import (PoolingRequestOutput, RequestOutput,
                           RequestOutputFactory)
 from vllm.pooling_params import PoolingParams
-from vllm.prompt_adapter.request import PromptAdapterRequest
 from vllm.sampling_params import RequestOutputKind, SamplingParams
 from vllm.sequence import (ExecuteModelRequest, ParallelSampleSequenceGroup,
                            PoolingSequenceGroupOutput, Sequence, SequenceGroup,
@@ -223,7 +222,6 @@ class LLMEngine:
         self.load_config = vllm_config.load_config
         self.decoding_config = vllm_config.decoding_config or DecodingConfig(  # noqa
         )
-        self.prompt_adapter_config = vllm_config.prompt_adapter_config  # noqa
         self.observability_config = vllm_config.observability_config or ObservabilityConfig(  # noqa
         )
 
@@ -294,8 +292,6 @@ class LLMEngine:
                     # Feature flags
                     "enable_lora":
                     bool(self.lora_config),
-                    "enable_prompt_adapter":
-                    bool(self.prompt_adapter_config),
                     "enable_prefix_caching":
                     self.cache_config.enable_prefix_caching,
                     "enforce_eager":
@@ -542,9 +538,6 @@ class LLMEngine:
             self.lora_config.verify_with_model_config(self.model_config)
             self.lora_config.verify_with_scheduler_config(
                 self.scheduler_config)
-        if self.prompt_adapter_config:
-            self.prompt_adapter_config.verify_with_model_config(
-                self.model_config)
 
     def _add_processed_request(
         self,
@@ -553,7 +546,6 @@ class LLMEngine:
         params: Union[SamplingParams, PoolingParams],
         arrival_time: float,
         lora_request: Optional[LoRARequest],
-        prompt_adapter_request: Optional[PromptAdapterRequest],
         trace_headers: Optional[Mapping[str, str]] = None,
         priority: int = 0,
     ) -> Optional[SequenceGroup]:
@@ -569,7 +561,6 @@ class LLMEngine:
                 arrival_time=arrival_time,
                 lora_request=lora_request,
                 trace_headers=trace_headers,
-                prompt_adapter_request=prompt_adapter_request,
                 priority=priority,
             )
             return None
@@ -583,11 +574,10 @@ class LLMEngine:
         encoder_inputs, decoder_inputs = split_enc_dec_inputs(processed_inputs)
 
         seq = Sequence(seq_id, decoder_inputs, block_size, eos_token_id,
-                       lora_request, prompt_adapter_request)
+                       lora_request)
 
         encoder_seq = (None if encoder_inputs is None else Sequence(
-            seq_id, encoder_inputs, block_size, eos_token_id, lora_request,
-            prompt_adapter_request))
+            seq_id, encoder_inputs, block_size, eos_token_id, lora_request))
 
         # Create a SequenceGroup based on SamplingParams or PoolingParams
         if isinstance(params, SamplingParams):
@@ -598,7 +588,6 @@ class LLMEngine:
                 arrival_time=arrival_time,
                 lora_request=lora_request,
                 trace_headers=trace_headers,
-                prompt_adapter_request=prompt_adapter_request,
                 encoder_seq=encoder_seq,
                 priority=priority)
         elif isinstance(params, PoolingParams):
@@ -608,7 +597,6 @@ class LLMEngine:
                 params,
                 arrival_time=arrival_time,
                 lora_request=lora_request,
-                prompt_adapter_request=prompt_adapter_request,
                 encoder_seq=encoder_seq,
                 priority=priority)
         else:
@@ -637,7 +625,6 @@ class LLMEngine:
         lora_request: Optional[LoRARequest] = None,
         tokenization_kwargs: Optional[dict[str, Any]] = None,
         trace_headers: Optional[Mapping[str, str]] = None,
-        prompt_adapter_request: Optional[PromptAdapterRequest] = None,
         priority: int = 0,
     ) -> None:
         """Add a request to the engine's request pool.
@@ -658,7 +645,6 @@ class LLMEngine:
                 the current monotonic time.
             lora_request: The LoRA request to add.
             trace_headers: OpenTelemetry trace headers.
-            prompt_adapter_request: The prompt adapter request to add.
             priority: The priority of the request.
                 Only applicable with priority scheduling.
 
@@ -719,7 +705,6 @@ class LLMEngine:
             prompt,
             tokenization_kwargs=tokenization_kwargs,
             lora_request=lora_request,
-            prompt_adapter_request=prompt_adapter_request,
         )
 
         self._add_processed_request(
@@ -728,7 +713,6 @@ class LLMEngine:
             params=params,
             arrival_time=arrival_time,
             lora_request=lora_request,
-            prompt_adapter_request=prompt_adapter_request,
             trace_headers=trace_headers,
             priority=priority,
         )
@@ -741,7 +725,6 @@ class LLMEngine:
         arrival_time: float,
         lora_request: Optional[LoRARequest],
         trace_headers: Optional[Mapping[str, str]] = None,
-        prompt_adapter_request: Optional[PromptAdapterRequest] = None,
         encoder_seq: Optional[Sequence] = None,
         priority: int = 0,
     ) -> SequenceGroup:
@@ -769,17 +752,15 @@ class LLMEngine:
         if self.vllm_config.speculative_config is not None:
             draft_size = \
                 self.vllm_config.speculative_config.num_speculative_tokens + 1
-        seq_group = SequenceGroup(
-            request_id=request_id,
-            seqs=[seq],
-            arrival_time=arrival_time,
-            sampling_params=sampling_params,
-            lora_request=lora_request,
-            trace_headers=trace_headers,
-            prompt_adapter_request=prompt_adapter_request,
-            encoder_seq=encoder_seq,
-            priority=priority,
-            draft_size=draft_size)
+        seq_group = SequenceGroup(request_id=request_id,
+                                  seqs=[seq],
+                                  arrival_time=arrival_time,
+                                  sampling_params=sampling_params,
+                                  lora_request=lora_request,
+                                  trace_headers=trace_headers,
+                                  encoder_seq=encoder_seq,
+                                  priority=priority,
+                                  draft_size=draft_size)
 
         return seq_group
 
@@ -790,7 +771,6 @@ class LLMEngine:
         pooling_params: PoolingParams,
         arrival_time: float,
         lora_request: Optional[LoRARequest],
-        prompt_adapter_request: Optional[PromptAdapterRequest],
         encoder_seq: Optional[Sequence] = None,
         priority: int = 0,
     ) -> SequenceGroup:
@@ -798,15 +778,13 @@ class LLMEngine:
         # Defensive copy of PoolingParams, which are used by the pooler
         pooling_params = pooling_params.clone()
         # Create the sequence group.
-        seq_group = SequenceGroup(
-            request_id=request_id,
-            seqs=[seq],
-            arrival_time=arrival_time,
-            lora_request=lora_request,
-            pooling_params=pooling_params,
-            prompt_adapter_request=prompt_adapter_request,
-            encoder_seq=encoder_seq,
-            priority=priority)
+        seq_group = SequenceGroup(request_id=request_id,
+                                  seqs=[seq],
+                                  arrival_time=arrival_time,
+                                  lora_request=lora_request,
+                                  pooling_params=pooling_params,
+                                  encoder_seq=encoder_seq,
+                                  priority=priority)
         return seq_group
 
     def abort_request(self, request_id: Union[str, Iterable[str]]) -> None:
@@ -1834,16 +1812,6 @@ class LLMEngine:
     def pin_lora(self, lora_id: int) -> bool:
         return self.model_executor.pin_lora(lora_id)
 
-    def add_prompt_adapter(
-            self, prompt_adapter_request: PromptAdapterRequest) -> bool:
-        return self.model_executor.add_prompt_adapter(prompt_adapter_request)
-
-    def remove_prompt_adapter(self, prompt_adapter_id: int) -> bool:
-        return self.model_executor.remove_prompt_adapter(prompt_adapter_id)
-
-    def list_prompt_adapters(self) -> List[int]:
-        return self.model_executor.list_prompt_adapters()
-
     def start_profile(self) -> None:
         self.model_executor.start_profile()