[GH-ISSUE #49] Model always responds in Chinese when using standard way of vllm + huggingface model #46

Open
opened 2026-10-26 21:07:33 +00:00 by bsakaguchi · 0 comments
Owner

Originally created by @iamhappytoo on GitHub (Sep 17, 2024).
Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-Coder-V2/issues/49

image

Code used:
from transformers import AutoTokenizer
from vllm import LLM, SamplingParams

max_model_len, tp_size = 8192, 1
model_name = "deepseek-ai/DeepSeek-Coder-V2-Instruct"
tokenizer = AutoTokenizer.from_pretrained(model_name)
llm = LLM(model=model_name, tensor_parallel_size=tp_size, max_model_len=max_model_len, trust_remote_code=True, enforce_eager=True)
sampling_params = SamplingParams(temperature=0.3, max_tokens=256, stop_token_ids=[tokenizer.eos_token_id])

messages_list = [
[{"role": "user", "content": "Who are you?"}],
[{"role": "user", "content": "write a quick sort algorithm in python."}],
[{"role": "user", "content": "Write a piece of quicksort code in C++."}],
]

prompt_token_ids = [tokenizer.apply_chat_template(messages, add_generation_prompt=True) for messages in messages_list]

outputs = llm.generate(prompt_token_ids=prompt_token_ids, sampling_params=sampling_params)

generated_text = [output.outputs[0].text for output in outputs]
print(generated_text)

vllm version: 0.5.5 or 0.6.1.post2 all have the Chinese output.

Originally created by @iamhappytoo on GitHub (Sep 17, 2024). Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-Coder-V2/issues/49 ![image](https://github.com/user-attachments/assets/c2c01780-4fb7-4915-8338-9bdd338d3ee1) Code used: from transformers import AutoTokenizer from vllm import LLM, SamplingParams max_model_len, tp_size = 8192, 1 model_name = "deepseek-ai/DeepSeek-Coder-V2-Instruct" tokenizer = AutoTokenizer.from_pretrained(model_name) llm = LLM(model=model_name, tensor_parallel_size=tp_size, max_model_len=max_model_len, trust_remote_code=True, enforce_eager=True) sampling_params = SamplingParams(temperature=0.3, max_tokens=256, stop_token_ids=[tokenizer.eos_token_id]) messages_list = [ [{"role": "user", "content": "Who are you?"}], [{"role": "user", "content": "write a quick sort algorithm in python."}], [{"role": "user", "content": "Write a piece of quicksort code in C++."}], ] prompt_token_ids = [tokenizer.apply_chat_template(messages, add_generation_prompt=True) for messages in messages_list] outputs = llm.generate(prompt_token_ids=prompt_token_ids, sampling_params=sampling_params) generated_text = [output.outputs[0].text for output in outputs] print(generated_text) vllm version: 0.5.5 or 0.6.1.post2 all have the Chinese output.
Sign in to join this conversation.
No labels
pull-request
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
deepseek-ai/DeepSeek-Coder-V2#46
No description provided.