[Frontend] Use new Renderer for Completions and Tokenize API (#32863)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2026-01-31 04:51:15 -08:00
committed by GitHub
parent 8980001c93
commit f0a1c8453a
64 changed files with 2116 additions and 2003 deletions
+9 -47
View File
@@ -1,11 +1,10 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
from collections.abc import Sequence
from typing import TYPE_CHECKING, Literal, NamedTuple, TypeAlias, TypedDict, cast
from typing import TYPE_CHECKING, Literal, NamedTuple, TypeAlias, TypedDict
from typing_extensions import TypeIs
from vllm.utils.collection_utils import is_list_of
from vllm.utils import length_from_prompt_token_ids_or_embeds
from .data import (
EmbedsPrompt,
@@ -22,50 +21,6 @@ if TYPE_CHECKING:
import torch
def parse_raw_prompts(
prompt: str | list[str] | list[int] | list[list[int]],
) -> Sequence[TextPrompt] | Sequence[TokensPrompt]:
if isinstance(prompt, str):
# case 1: a string
return [TextPrompt(prompt=prompt)]
if isinstance(prompt, list):
if len(prompt) == 0:
raise ValueError("please provide at least one prompt")
# case 2: array of strings
if is_list_of(prompt, str):
prompt = cast(list[str], prompt)
return [TextPrompt(prompt=elem) for elem in prompt]
# case 3: array of tokens
if is_list_of(prompt, int):
prompt = cast(list[int], prompt)
return [TokensPrompt(prompt_token_ids=prompt)]
# case 4: array of token arrays
if is_list_of(prompt, list):
if len(prompt) == 1 and isinstance(prompt[0], list) and len(prompt[0]) == 0:
raise ValueError("please provide at least one prompt")
for elem in prompt:
if not isinstance(elem, list):
raise TypeError(
"prompt must be a list of lists, but found a non-list element."
)
if not is_list_of(elem, int):
raise TypeError(
"Nested lists of tokens must contain only integers."
)
prompt = cast(list[list[int]], prompt)
return [TokensPrompt(prompt_token_ids=elem) for elem in prompt]
raise TypeError(
"prompt must be a string, array of strings, "
"array of tokens, or array of token arrays"
)
class ParsedStrPrompt(TypedDict):
type: Literal["str"]
content: str
@@ -145,3 +100,10 @@ def get_prompt_components(prompt: PromptType) -> PromptComponents:
token_ids=prompt.get("prompt_token_ids"), # type: ignore[arg-type]
embeds=prompt.get("prompt_embeds"),
)
def get_prompt_len(prompt: TokensPrompt | EmbedsPrompt):
return length_from_prompt_token_ids_or_embeds(
prompt.get("prompt_token_ids"), # type: ignore[arg-type]
prompt.get("prompt_embeds"), # type: ignore[arg-type]
)