|
|
--- |
|
|
license: mit |
|
|
language: |
|
|
- en |
|
|
tags: |
|
|
- reasoning, |
|
|
- chat, |
|
|
- instruct, |
|
|
- base, |
|
|
pretty_name: Dual Mode Reasoning Example |
|
|
size_categories: |
|
|
- n<1K |
|
|
--- |
|
|
|
|
|
 |
|
|
|
|
|
This is an example of a dataset that can be used to finetune a base llm to have dual thinking modes by adding flags to the system prompt. Here is an example of a chat template (https://huggingface.co/AIGym/smollm3-base-3b/blob/main/chat_template.jinja). |
|
|
|
|
|
|
|
|
If you are interested in pretraining a base model checkout this example (https://github.com/vibe-coding-projects/build-nanogpt) |
