FareedKhan-dev/train-deepseek-r1
A Jupyter notebook and guide walking through the step-by-step implementation of DeepSeek R1's training process using GRPO reinforcement learning.
Not currently ranked — collecting fresh signals.
star history
The repository provides a hands-on implementation of DeepSeek R1’s training methodology, covering reinforcement learning fundamentals, the GRPO algorithm, reward functions for accuracy and format validation, and policy model setup. It includes explanatory markdown documents with hand-drawn diagrams to help non-technical audiences understand LLM training concepts.
Frequently asked
- What is FareedKhan-dev/train-deepseek-r1?
- A Jupyter notebook and guide walking through the step-by-step implementation of DeepSeek R1's training process using GRPO reinforcement learning.
- Is train-deepseek-r1 open source?
- Yes — FareedKhan-dev/train-deepseek-r1 is open source, released under the MIT license.
- What language is train-deepseek-r1 written in?
- FareedKhan-dev/train-deepseek-r1 is primarily written in Jupyter Notebook.
- How popular is train-deepseek-r1?
- FareedKhan-dev/train-deepseek-r1 has 782 stars on GitHub.
- Where can I find train-deepseek-r1?
- FareedKhan-dev/train-deepseek-r1 is on GitHub at https://github.com/FareedKhan-dev/train-deepseek-r1.