dataset_info:
features:
- name: messages
list:
- name: role
dtype: string
- name: content
dtype: string
- name: metadata
struct:
- name: category
dtype: string
- name: seed_prompt
dtype: string
- name: num_turns
dtype: int64
- name: timestamp
dtype: string
configs:

• config_name: default
  data_files:
  • split: train
    path: "data/train-*.parquet"

language:

• en
  license: apache-2.0
  task_categories:
• text-generation
  tags:
• chat
• synthetic
• roleplay
• waifu
• multi-turn
• chatml
  pretty_name: Wifu Chat Dataset
  size_categories:
• n<1K

Wifu Chat Dataset

Synthetic multi-turn chat dataset for fine-tuning companion/waifu chatbots. Contains 403 uncensored conversations across categories like greetings, emotional support, flirty, roleplay, and NSFW scenarios.

Format: ChatML JSONL -- messages array with user/assistant roles. No system prompt included.

Disclaimer: This dataset contains explicit adult content (NSFW). Intended for research and fine-tuning uncensored models.

hiii~ wanna support me? 💕

sooo generating these datasets isn't free lol 😭 like it literally costs me real money to rent GPUs and run these models for hours and hours~ every conversation you see here was cooked on expensive H100s and honestly my wallet is crying rn (,,>_<,,)

if this dataset helped you out or saved you a bunch of time, maybe consider buying me a coffee?? it would mean the world to me and help me keep making more cute datasets for everyone~ ✨

every little bit helps me pay for compute, try out new models, and drop more open-source goodies for the community~ think of it as fueling the next batch of waifu training data hehe 🔥

tysm for even reading this far, you're already amazing 💗