Datasets:
metadata
license: apache-2.0
language:
- en
- hi
tags:
- chat
- conversational
- hinglish
- bilingual
- 50-50
- multi-turn
- viuai
size_categories:
- 1M<n<10M
task_categories:
- text-generation
🌐 ViuAI-10M-Chat-Data (50% English + 50% Hinglish)
The complete, uncapped 50:50 bilingual chat dataset curated for training the ViuAI-10M 32-layer conversational model.
📊 Dataset Composition (Strict 50:50 Interleaved)
| Language | Proportion | Primary Sources | Nature of Conversations |
|---|---|---|---|
| English | 50% | WildChat-1M, UltraChat, OpenAssistant, Dolly-15k | Reasoning, coding, multi-turn assistance, factual QA, complex discussion |
| Hinglish | 50% | Tensoic Hinglish, GritAI Conversations, WhatsApp Banter, Alpaca-Hindi | Casual Roman-Hindi chat, Indian slang, daily life, friendly chit-chat |
💬 Format
Every conversation follows the standard OpenAI / Hugging Face chat message format:
{
"messages": [
{"role": "user", "content": "Bhai ek python script likh ke do jo web scrape kare"},
{"role": "assistant", "content": "Haan bhai, bilkul! BeautifulSoup aur requests use karke ek clean script ye rahi..."}
]
}
🚀 Purpose
Built by ViuAI to train ViuAI-10M into a bilingual chat model.
Maintained with ❤️ by ViuAI Team