[Frontend] supports deepseekv32 chat template (#29837)
Signed-off-by: chaunceyjiang <chaunceyjiang@gmail.com>
This commit is contained in:
@@ -105,7 +105,7 @@ from vllm.outputs import CompletionOutput, PoolingRequestOutput, RequestOutput
|
||||
from vllm.pooling_params import PoolingParams
|
||||
from vllm.reasoning import ReasoningParser, ReasoningParserManager
|
||||
from vllm.sampling_params import BeamSearchParams, SamplingParams
|
||||
from vllm.tokenizers import MistralTokenizer, TokenizerLike
|
||||
from vllm.tokenizers import DeepseekV32Tokenizer, MistralTokenizer, TokenizerLike
|
||||
from vllm.tracing import (
|
||||
contains_trace_headers,
|
||||
extract_trace_headers,
|
||||
@@ -1128,6 +1128,13 @@ class OpenAIServing:
|
||||
messages=messages,
|
||||
**_chat_template_kwargs,
|
||||
)
|
||||
elif isinstance(tokenizer, DeepseekV32Tokenizer):
|
||||
request_prompt = tokenizer.apply_chat_template(
|
||||
conversation=conversation,
|
||||
messages=messages,
|
||||
model_config=model_config,
|
||||
**_chat_template_kwargs,
|
||||
)
|
||||
else:
|
||||
request_prompt = apply_hf_chat_template(
|
||||
tokenizer=tokenizer,
|
||||
|
||||
Reference in New Issue
Block a user