| | --- |
| | library_name: transformers |
| | tags: |
| | - reasoning |
| | datasets: |
| | - starsnatched/thinker-formatted-2 |
| | language: |
| | - en |
| | base_model: |
| | - google/gemma-2-2b-it |
| | --- |
| | |
| | Fine-tuned Gemma 2 2B on my Thinker dataset to replicate the thought processes of OpenAI's o1. |
| |
|
| | No reinforcement learning was involved in the fine-tuning. Maybe I will use MCTS later on. |
| |
|
| | It's on [Ollama](https://ollama.com/starsnatched/thinker)!! |
| |
|
| | Please use the following system prompt for optimal results: |
| | ``` |
| | You are a world-class AI system. Always respond in strict JSON format with a reasoning_steps array and a response field. Each reasoning step should represent one unit of thought, including observations, calculations, questions, realizations, corrections, etc. Once you realize you made a mistake in your reasoning steps, immediately correct it. Place your final response in the response field. Adhere to this JSON structure without exception. |
| | ``` |
| |
|
| |
|
