lukasm-uel0417/

bart-tokenizer-tryout

	build_dataset.py




	


	Dataset Summary

A nlp sentiment dataset with pointcloud text modality, stored in npy sharded format.

Preprocessing: curriculum Augmentation: randaugment

Split strategy: leave one out Sampling: curriculum

Quality filtering: adaptive Labeling: weak supervision

build_dataset.py — main artifact of this repository… See the full description on the dataset page: https://huggingface.co/datasets/lukasm-uel0417/bart-tokenizer-tryout.