Commit History
-
Upload MultiDomain-100M.txt
Verified signatureA multi-domain text dataset created for training lightweight GPT-style language models on low-VRAM consumer hardware. Dataset Composition: * Wikipedia * Books * News * Question Answering * Web Text Approximate Token Distribution: * Wiki: 40M * Books: 25M * News: 6M * QA: 10M * Web: 20M The dataset was cleaned, deduplicated, balanced, and formatted for transformer training.f3a1cdb -
initial commit
Verified signaturea14680e