[Frontend] Introduce Renderer for processing chat messages (using ModelConfig) (#30200)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
@@ -0,0 +1,48 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
from typing import TYPE_CHECKING, Any, Protocol
|
||||
|
||||
from vllm.inputs import TextPrompt, TokensPrompt
|
||||
from vllm.tokenizers import TokenizerLike
|
||||
|
||||
if TYPE_CHECKING:
|
||||
from vllm.config import ModelConfig
|
||||
from vllm.entrypoints.chat_utils import (
|
||||
ChatCompletionMessageParam,
|
||||
ConversationMessage,
|
||||
)
|
||||
|
||||
|
||||
class RendererLike(Protocol):
|
||||
@classmethod
|
||||
def from_config(
|
||||
cls,
|
||||
config: "ModelConfig",
|
||||
tokenizer_kwargs: dict[str, Any],
|
||||
) -> "RendererLike":
|
||||
raise NotImplementedError
|
||||
|
||||
@property
|
||||
def tokenizer(self) -> TokenizerLike | None:
|
||||
raise NotImplementedError
|
||||
|
||||
def get_tokenizer(self) -> TokenizerLike:
|
||||
tokenizer = self.tokenizer
|
||||
if tokenizer is None:
|
||||
raise ValueError("Tokenizer not available when `skip_tokenizer_init=True`")
|
||||
|
||||
return tokenizer
|
||||
|
||||
def render_messages(
|
||||
self,
|
||||
messages: list["ChatCompletionMessageParam"],
|
||||
**kwargs,
|
||||
) -> tuple[list["ConversationMessage"], TextPrompt | TokensPrompt]:
|
||||
raise NotImplementedError
|
||||
|
||||
async def render_messages_async(
|
||||
self,
|
||||
messages: list["ChatCompletionMessageParam"],
|
||||
**kwargs,
|
||||
) -> tuple[list["ConversationMessage"], TextPrompt | TokensPrompt]:
|
||||
return self.render_messages(messages, **kwargs)
|
||||
Reference in New Issue
Block a user