4 ms·
The dataset is 7 times bigger than the dataset used for Llama 2 as reported by Meta.
by GaggiX 3y ago
The dataset is 7 times bigger than the dataset used for Llama 2 as reported by Meta.
- baobabKoodaa 3y agoHas Meta disclosed how much parts of the dataset were repeated? I've only seen the "number of tokens trained" number.