5 ms·
We don't actually know how big the Dataset is, right? It could be the same dataset used for Llama 2, but trained for more Epochs.
by Version467 2y ago
We don't actually know how big the Dataset is, right? It could be the same dataset used for Llama 2, but trained for more Epochs.
- GaggiX 2y agoThe dataset is 7 times bigger than the dataset used for Llama 2 as reported by Meta.
- baobabKoodaa 2y agoHas Meta disclosed how much parts of the dataset were repeated? I've only seen the "number of tokens trained" number.