[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
+25 -4
View File
@@ -3,12 +3,10 @@
import time
from typing import Any, TypeAlias
from pydantic import (
BaseModel,
Field,
)
from pydantic import BaseModel, Field
from vllm import PoolingParams
from vllm.config import ModelConfig
from vllm.config.pooler import get_use_activation
from vllm.entrypoints.openai.engine.protocol import OpenAIBaseModel, UsageInfo
from vllm.entrypoints.pooling.base.protocol import (
@@ -19,6 +17,7 @@ from vllm.entrypoints.pooling.score.utils import (
ScoreContentPartParam,
ScoreMultiModalParam,
)
from vllm.renderers import TokenizeParams
from vllm.utils import random_uuid
@@ -30,6 +29,17 @@ class ScoreRequestMixin(PoolingBasicRequestMixin, ClassifyRequestMixin):
)
# --8<-- [end:score-extra-params]
def build_tok_params(self, model_config: ModelConfig) -> TokenizeParams:
encoder_config = model_config.encoder_config or {}
return TokenizeParams(
max_total_tokens=model_config.max_model_len,
max_output_tokens=0,
truncate_prompt_tokens=self.truncate_prompt_tokens,
do_lower_case=encoder_config.get("do_lower_case", False),
max_total_tokens_param="max_model_len",
)
def to_pooling_params(self):
return PoolingParams(
truncate_prompt_tokens=self.truncate_prompt_tokens,
@@ -85,6 +95,17 @@ class RerankRequest(PoolingBasicRequestMixin, ClassifyRequestMixin):
)
# --8<-- [end:rerank-extra-params]
def build_tok_params(self, model_config: ModelConfig) -> TokenizeParams:
encoder_config = model_config.encoder_config or {}
return TokenizeParams(
max_total_tokens=model_config.max_model_len,
max_output_tokens=0,
truncate_prompt_tokens=self.truncate_prompt_tokens,
do_lower_case=encoder_config.get("do_lower_case", False),
max_total_tokens_param="max_model_len",
)
class RerankDocument(BaseModel):
text: str | None = None