Hi, the format of the training samples is the same as the case we provided in Figure 3. And the format of the system prompt and user prompt is given in the example codes (in README)

That is:

[SYSTEM]: You are a philosopher skilled in deep thinking, accustomed to exploring complex problems with profound insight.
[USER]: Please translate the following text from English to Chinese:\n{An English Sentence/Paragraph}
[ASSISTANT]: <thought>\n{the corresponding long thought}\n</thought>\n<output>\n{the final translation result}\n</output>

@Krystalan Thank you for the response. How can someone create more data in order to train on an additional layer of fine-tuning? Are there any resources from experiments on how to create such a dataset? (Prompt for synthetic data? )

Thanks for your interest! We plan to update our preprint paper to give more details, maybe in the next two weeks.

Calling LLMs:

from openai import OpenAI

Set OpenAI's API key and API base to use vLLM's API server.

openai_api_key = "EMPTY"
openai_api_base = "http://localhost:8000/v1"

client = OpenAI(
api_key=openai_api_key,
base_url=openai_api_base,
)
去哪里申请 openai_api_key 和openai_api_base 来调用DRT-o1 模型呢

文本只能支持中英互译，还是可以支持，多种语言，如泰语、日语、马来语、韩语等