0x-genesys/

mix_wiki_code_chat_data_300M_tokens

This guide explains how to use the new automatic dataset download system that fetches pre-processed datasets from HuggingFace Hub instead of building them locally.

	📊 Dataset Composition




	

Source Tokens Purpose

WikiText-103 102M (34%) Knowledge

UltraChat 150M (50%) Conversation

The Stack 48M (16%) Code

DailyDialog 10M (3%) Dialogue

TOTAL 300M Balanced

Epochs: 8 Total tokens: 2.4B (300M… See the full description on the dataset page: https://huggingface.co/datasets/0x-genesys/mix_wiki_code_chat_data_300M_tokens.