Man I am so hyped about DeepSeek right now! Its seriously amazing how much theyve shared with the community. Most of their core model code and architecture is totally out there on GitHub for everyone to poke around in and modify. Its such a huge win for transparency and innovation... honestly love seeing it. I wanted to dig a bit deeper into what you are trying to do though:

• Are you looking to dive into the actual model architecture for research, or are you just trying to fine-tune it for a specific task?
• Is there a specific version you are eyeing, like the Coder series or the general LLM?

DeepSeek's models are open-source and available for public review and modification, but some proprietary components are not disclosed.