Cleaned Thai text corpus assembled for pretraining a small language model (SLM) from scratch. Built by the data_pipeline project: each upstream source is normalized to a single text field, unicode-normalized, whitespace-cleaned, length-filtered, and exact-deduplicated.

	Schema

column type description

text string one cleaned document

source string which curated source it came from

Build summary

Total… See the full description on the dataset page: https://huggingface.co/datasets/SPAISS6F1/slm-pretrain-corpus.