Remove grok model arch from vllm (#46706)
Signed-off-by: Xianbao QIAN <xianbao.qian@gmail.com>
This commit is contained in:
@@ -1,90 +0,0 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
|
||||
from vllm.config import VllmConfig
|
||||
from vllm.entrypoints.chat_utils import (
|
||||
ChatCompletionMessageParam,
|
||||
ConversationMessage,
|
||||
parse_chat_messages,
|
||||
parse_chat_messages_async,
|
||||
)
|
||||
from vllm.logger import init_logger
|
||||
from vllm.tokenizers.grok2 import Grok2Tokenizer
|
||||
from vllm.utils.async_utils import make_async
|
||||
|
||||
from .base import BaseRenderer
|
||||
from .inputs import DictPrompt
|
||||
from .inputs.preprocess import parse_dec_only_prompt
|
||||
from .params import ChatParams
|
||||
|
||||
logger = init_logger(__name__)
|
||||
|
||||
|
||||
class Grok2Renderer(BaseRenderer[Grok2Tokenizer]):
|
||||
def __init__(
|
||||
self,
|
||||
config: VllmConfig,
|
||||
tokenizer: Grok2Tokenizer | None,
|
||||
) -> None:
|
||||
super().__init__(config, tokenizer)
|
||||
|
||||
self._apply_chat_template_async = make_async(
|
||||
self._apply_chat_template, executor=self._executor
|
||||
)
|
||||
|
||||
def _apply_chat_template(self, *args, **kwargs):
|
||||
return self.get_tokenizer().apply_chat_template(*args, **kwargs)
|
||||
|
||||
def render_messages(
|
||||
self,
|
||||
messages: list[ChatCompletionMessageParam],
|
||||
params: ChatParams,
|
||||
) -> tuple[list[ConversationMessage], DictPrompt]:
|
||||
conversation, mm_data, mm_uuids = parse_chat_messages(
|
||||
messages,
|
||||
self.model_config,
|
||||
content_format="string",
|
||||
media_io_kwargs=params.media_io_kwargs,
|
||||
mm_processor_kwargs=params.mm_processor_kwargs,
|
||||
)
|
||||
|
||||
prompt_raw = self._apply_chat_template(
|
||||
conversation=conversation,
|
||||
messages=messages,
|
||||
**params.get_apply_chat_template_kwargs(),
|
||||
)
|
||||
|
||||
prompt = parse_dec_only_prompt(prompt_raw)
|
||||
if mm_data is not None:
|
||||
prompt["multi_modal_data"] = mm_data
|
||||
if mm_uuids is not None:
|
||||
prompt["multi_modal_uuids"] = mm_uuids
|
||||
|
||||
return conversation, prompt
|
||||
|
||||
async def render_messages_async(
|
||||
self,
|
||||
messages: list[ChatCompletionMessageParam],
|
||||
params: ChatParams,
|
||||
) -> tuple[list[ConversationMessage], DictPrompt]:
|
||||
conversation, mm_data, mm_uuids = await parse_chat_messages_async(
|
||||
messages,
|
||||
self.model_config,
|
||||
content_format="string",
|
||||
media_io_kwargs=params.media_io_kwargs,
|
||||
mm_processor_kwargs=params.mm_processor_kwargs,
|
||||
)
|
||||
|
||||
prompt_raw = await self._apply_chat_template_async(
|
||||
conversation=conversation,
|
||||
messages=messages,
|
||||
**params.get_apply_chat_template_kwargs(),
|
||||
)
|
||||
|
||||
prompt = parse_dec_only_prompt(prompt_raw)
|
||||
if mm_data is not None:
|
||||
prompt["multi_modal_data"] = mm_data
|
||||
if mm_uuids is not None:
|
||||
prompt["multi_modal_uuids"] = mm_uuids
|
||||
|
||||
return conversation, prompt
|
||||
@@ -22,7 +22,6 @@ logger = init_logger(__name__)
|
||||
_VLLM_RENDERERS = {
|
||||
"deepseek_v32": ("deepseek_v32", "DeepseekV32Renderer"),
|
||||
"deepseek_v4": ("deepseek_v4", "DeepseekV4Renderer"),
|
||||
"grok2": ("grok2", "Grok2Renderer"),
|
||||
"hf": ("hf", "HfRenderer"),
|
||||
"kimi_audio": ("hf", "HfRenderer"),
|
||||
"mistral": ("mistral", "MistralRenderer"),
|
||||
|
||||
Reference in New Issue
Block a user