Dataset Card for uplimit-synthetic-data-week-1-with-multi-turn

This dataset has been created with distilabel.

The pipeline script was uploaded to easily reproduce the dataset: colab_kernel_launcher.py.

It can be run directly using the CLI:

distilabel pipeline run --script "https://huggingface.co/datasets/jsimon-pfpt/uplimit-synthetic-data-week-1-with-multi-turn/raw/main/colab_kernel_launcher.py"

Dataset Summary

This dataset contains a pipeline.yaml which can be used to reproduce the pipeline that generated it in distilabel using the distilabel CLI:

distilabel pipeline run --config "https://huggingface.co/datasets/jsimon-pfpt/uplimit-synthetic-data-week-1-with-multi-turn/raw/main/pipeline.yaml"

or explore the configuration:

distilabel pipeline info --config "https://huggingface.co/datasets/jsimon-pfpt/uplimit-synthetic-data-week-1-with-multi-turn/raw/main/pipeline.yaml"

Dataset structure

The examples have the following structure per configuration:

Configuration: default

---

{
    "conversation": [
        {
            "content": "I\u0027m planning a trip to Japan and I\u0027m interested in visiting Tokyo. I\u0027ve heard that Tokyo is a city that\u0027s full of contrasts, and I\u0027m excited to experience the traditional and modern aspects of the city. I\u0027d like to know more about the different neighborhoods to visit in Tokyo. What are some must-visit neighborhoods to consider?\n\nI\u0027m particularly interested in exploring the traditional side of Tokyo, and I\u0027d like to know about the Asakusa district, which I\u0027ve heard is a great place to experience the old Japan.",
            "role": "user"
        },
        {
            "content": "You\u0027re in for a treat! Tokyo is indeed a city of contrasts, and exploring its different neighborhoods is a big part of the fun. Asakusa is an excellent choice for experiencing traditional Japan, and I\u0027d be happy to tell you more about it.\n\nAsakusa is one of the oldest districts in Tokyo, with a rich history dating back to the 15th century. It\u0027s a great place to immerse yourself in traditional Japanese culture, and it\u0027s still relatively untouched by modernization. Here are some must-visit spots in Asakusa:\n\n1. Senso-ji Temple: This iconic Buddhist temple is the main",
            "role": "assistant"
        },
        {
            "content": "...main attraction in Asakusa, and it\u0027s a must-visit for anyone interested in Japanese culture. The temple is famous for its giant paper lantern, Nakamise Shopping Street, and the nearby Nakamise Shopping Street, which is a historic shopping street filled with traditional snacks, souvenirs, and local goods.\n\n2. Asakusa Shrine: This Shinto shrine is dedicated to the god of learning and is located near Senso-ji Temple. It\u0027s a peaceful oasis in the midst of the bustling city, and it\u0027s a great place to experience traditional Japanese spirituality.\n\n3. Sumida River: Asakusa",
            "role": "user"
        },
        {
            "content": "...is situated on the banks of the Sumida River, which offers beautiful views of the Tokyo Skytree and the city skyline. You can take a stroll along the riverbank, rent a boat, or simply sit and enjoy the scenery.\n\nIn addition to these attractions, Asakusa is also known for its traditional Japanese cuisine, such as tempura, soba noodles, and traditional sweets. Be sure to try some of the local specialties, like takoyaki and okonomiyaki, while you\u0027re in the area.\n\nIf you\u0027re interested in exploring more of Tokyo\u0027s traditional side, you might also want to consider visiting the nearby",
            "role": "assistant"
        }
    ],
    "distilabel_metadata": {
        "statistics_magpie_generator_0": {
            "input_tokens": [
                85,
                193,
                331,
                459
            ],
            "output_tokens": [
                109,
                128,
                128,
                128
            ]
        }
    },
    "model_name": "https://api-inference.huggingface.co/models/meta-llama/Llama-3.2-3B-Instruct"
}

This subset can be loaded as:

from datasets import load_dataset

ds = load_dataset("jsimon-pfpt/uplimit-synthetic-data-week-1-with-multi-turn", "default")

Or simply as it follows, since there's only one configuration and is named default:

from datasets import load_dataset

ds = load_dataset("jsimon-pfpt/uplimit-synthetic-data-week-1-with-multi-turn")

References

@misc{xu2024magpiealignmentdatasynthesis,
    title={Magpie: Alignment Data Synthesis from Scratch by Prompting Aligned LLMs with Nothing},
    author={Zhangchen Xu and Fengqing Jiang and Luyao Niu and Yuntian Deng and Radha Poovendran and Yejin Choi and Bill Yuchen Lin},
    year={2024},
    eprint={2406.08464},
    archivePrefix={arXiv},
    primaryClass={cs.CL},
    url={https://arxiv.org/abs/2406.08464},
}