Qwen

/

Qwen3-1.7B

like 376

Follow

Qwen 63.6k

Code:

from transformers import AutoModelForCausalLM, AutoTokenizer

tokenizer1 = AutoTokenizer.from_pretrained("Qwen/Qwen3-1.7B")
prompt = "Give me a short introduction to large language model."
messages = [
    {"role": "user", "content": prompt}
]
text1 = tokenizer1.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
    enable_thinking=True,
)

print(text1)

Output:

<|im_start|>user
Give me a short introduction to large language model.<|im_end|>
<|im_start|>assistant

And I used EvalScope to test AIME25, and although I obtained similar results to the think-mode results in the report, the output also didn't contain <think>. So I'm curious: will Qwen3-1.7B should output something like "<think>{thinking context}</think>{answer}"?