CoolFace
Datasetpublic

ncylich/temporal-moe-corpus

temporal-moe-corpus Tokenized training corpus for the Temporal-MoE experiments, 31.3 GiB. This repository holds the tokenized corpus in Megatron indexed-dataset format, plus the 16k tokenizer. It is what the training runs actually read. The raw web-crawl text is not hosted here, because it is a byte-reproducible subset of a public dataset. The exact recipe and the checksums needed to verify a reproduction are below. Contents dclm_tokenized/ 22 x… See the full description on the dataset page: https://huggingface.co/datasets/ncylich/temporal-moe-corpus.

sourceHugging Facecc-by-4.0updated 2mo agoView on Hugging Face
0likes200downloads

ncylich/temporal-moe-corpus · main · files are served by the source, never re-hosted here