---
license: apache-2.0
datasets:
- markusheimerl/socratic_dialogs
language:
- en
---

- 100 million trainable parameters
- GPT model definition from https://github.com/mistralai/mistral-src/blob/main/one_file_ref.py
- Trained for 3 hours on 4xA100 80GB.
- Repo at https://github.com/ratisbonrobotics/gpt