Vara Sera
Vara Sera
### Proposal It would be nice to include OLMo (1B and 7B) and their checkpoints as available compatible models for HookedTransformer. ### Motivation OLMo-1B would be a great model to...
### Make sure you can reproduce the issue with the latest version available ``` pip install milatools --upgrade ``` ### What command did you run? `mila init` ### Describe the...
### 🐛 Describe the bug When trying to train olmo2-1B from a checkpoint, I've begun to see very poor/inconsistent connection to olmo-data.org in the last few weeks. I wasn't sure...