allenai/dolma3_pool

⚠️ IMPORTANT NOTICE ⚠️ This is the Dolma 3 pool, pre–quality upsampling and mixing. If you are interested in the data used to train Olmo 3 7B and Olmo 3 32B, visit allenai/dolma3_mix-6T-1025. Dolma 3 Pool The Dolma 3 pool is a dataset of over 9 trillion tokens from a diverse mix of web content, academic publications, code, and more. For detailed documenation on Dolma 3 processing and data, please see our Dolma 3 Github repository. For more information on Dolma in general… See the full description on the dataset page: https://huggingface.co/datasets/allenai/dolma3_pool.