Moroccan Darija Pretraining Corpus for Nanochat

Domain:

natural language processing

Record type:

dataset

Creator:

Lyte

Host:

Parquet shards prepared for nanochat pretraining. 83 train shards 1 validation shard 82,738,910 train rows 50,000 validation rows Built from Lyte/darija-pretraining-corpus using these subsets: arabic_raw bilingual pure Each parquet file contains a single column: text

Lyte/darija-pretraining-corpus

AhmedH005/Al-Atlas-Moroccan-Darija-Pretraining

Pretraining code and dataset analysis for the Al-Atlas family of open-source LLMs trained on authent

atlasia-ma/Al-Atlas-Moroccan-Darija-Pretraining

Training and data analysis code for Al-Atlas dataset and models # Al-Atlas: Data, Models, Evals ##

Lyte/nanochat-darija-73m-base

Lyte/nanochat-darija-73m-instruct

Moroccan Darija Code-Switched Corpus (Moroccan Arabic)

This dataset contains sentence/post-level code-switched Moroccan Darija text with a single label per

Moroccan Darija Pretraining Corpus for Nanochat

Domain:

Record type:

Visit

Tasks

Languages

Tags