biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[V0 Deprecation] Remove Prompt Adapters (#20588)

Browse Source 
Signed-off-by: mgoin <mgoin64@gmail.com>
This commit is contained in:
Michael Goin
2025-07-23 19:36:48 -04:00
committed by GitHub
parent 78c13e30e1
commit 82ec66f514
60 changed files with 126 additions and 1727 deletions
Download Patch File Download Diff File Expand all files Collapse all files

35
vllm/inputs/preprocess.py
View File

@@ -13,7 +13,6 @@ from vllm.lora.request import LoRARequest
from vllm.multimodal import MULTIMODAL_REGISTRY, MultiModalRegistry
from vllm.multimodal.inputs import (MultiModalDataDict, MultiModalEncDecInputs,
MultiModalInputs)
from vllm.prompt_adapter.request import PromptAdapterRequest
from vllm.transformers_utils.tokenizer import AnyTokenizer
from vllm.transformers_utils.tokenizer_group import TokenizerGroup
@@ -168,18 +167,6 @@ class InputPreprocessor:
return decoder_input_ids
def _apply_prompt_adapter(
self,
prompt_token_ids: list[int],
prompt_adapter_request: Optional[PromptAdapterRequest],
) -> list[int]:
if prompt_adapter_request:
prompt_token_ids = (
[0] * prompt_adapter_request.prompt_adapter_num_virtual_tokens
+ prompt_token_ids)
return prompt_token_ids
def _get_tokenization_kw(
self,
overrides: Optional[dict[str, Any]] = None,
@@ -786,15 +773,10 @@ class InputPreprocessor:
def _build_decoder_only_llm_inputs(
self,
prompt_inputs: DecoderOnlyInputs,
prompt_adapter_request: Optional[PromptAdapterRequest],
) -> DecoderOnlyInputs:
if "prompt_token_ids" in prompt_inputs:
prompt_inputs = cast(Union[TokenInputs, MultiModalInputs],
prompt_inputs) # Needed for mypy
prompt_inputs["prompt_token_ids"] = self._apply_prompt_adapter(
prompt_inputs["prompt_token_ids"],
prompt_adapter_request=prompt_adapter_request,
)
return prompt_inputs
@@ -803,7 +785,6 @@ class InputPreprocessor:
prompt: SingletonPrompt,
tokenization_kwargs: Optional[dict[str, Any]] = None,
lora_request: Optional[LoRARequest] = None,
prompt_adapter_request: Optional[PromptAdapterRequest] = None,
return_mm_hashes: bool = False,
) -> DecoderOnlyInputs:
"""
@@ -815,7 +796,6 @@ class InputPreprocessor:
* prompt: input prompt
* lora_request
* prompt_adapter_request
* return_mm_hashes
Returns:
@@ -830,17 +810,13 @@ class InputPreprocessor:
return_mm_hashes=return_mm_hashes,
)
return self._build_decoder_only_llm_inputs(
prompt_comps,
prompt_adapter_request=prompt_adapter_request,
)
return self._build_decoder_only_llm_inputs(prompt_comps)
async def _process_decoder_only_prompt_async(
self,
prompt: SingletonPrompt,
tokenization_kwargs: Optional[dict[str, Any]] = None,
lora_request: Optional[LoRARequest] = None,
prompt_adapter_request: Optional[PromptAdapterRequest] = None,
return_mm_hashes: bool = False,
) -> DecoderOnlyInputs:
"""
@@ -854,17 +830,13 @@ class InputPreprocessor:
return_mm_hashes=return_mm_hashes,
)
return self._build_decoder_only_llm_inputs(
prompt_comps,
prompt_adapter_request=prompt_adapter_request,
)
return self._build_decoder_only_llm_inputs(prompt_comps)
def preprocess(
self,
prompt: PromptType,
tokenization_kwargs: Optional[dict[str, Any]] = None,
lora_request: Optional[LoRARequest] = None,
prompt_adapter_request: Optional[PromptAdapterRequest] = None,
return_mm_hashes: bool = False,
) -> ProcessorInputs:
"""Preprocess the input prompt."""
@@ -886,7 +858,6 @@ class InputPreprocessor:
prompt,
tokenization_kwargs=tokenization_kwargs,
lora_request=lora_request,
prompt_adapter_request=prompt_adapter_request,
return_mm_hashes=return_mm_hashes,
)
@@ -895,7 +866,6 @@ class InputPreprocessor:
prompt: PromptType,
tokenization_kwargs: Optional[dict[str, Any]] = None,
lora_request: Optional[LoRARequest] = None,
prompt_adapter_request: Optional[PromptAdapterRequest] = None,
return_mm_hashes: bool = False,
) -> ProcessorInputs:
"""
@@ -919,6 +889,5 @@ class InputPreprocessor:
prompt,
tokenization_kwargs=tokenization_kwargs,
lora_request=lora_request,
prompt_adapter_request=prompt_adapter_request,
return_mm_hashes=return_mm_hashes,
)