Oh I thought we just did Executive Orders these days. Interesting for Congress to want to be in the pocket of the three richest men in the United States, too.
They’re gonna have a hard time making it illegal to download a completely open source software lol
Unfortunately, as I’ve learned recently, it doesn’t look like Deepseek is actually open source.
You can download the model, but unless I’m misunderstanding, that feels comparable to calling Photoshop open source because you can download the .exe file on your computer.
Its MIT licensed. Meaning the code is open but the license is permissible in that copy’s can be subsequently closed. This is unlike with the GPL most generally associated with open source code.
What’s this? https://github.com/deepseek-ai/DeepSeek-V3
This comment here seems to summarize it well: https://github.com/deepseek-ai/DeepSeek-V3/issues/457#issuecomment-2627016777
It’s more open-sourced than I thought, but also seems debatable. I don’t know enough about LLMs to properly judge. I would probably stay away from calling it “completely open-sourced” though.
Well now I’m just gonna download it even harder.
I had zero interest in downloading this shit before because LLMs are just lying slop machines, but if it becomes illegal, I will go out of my way to.
I’m with you on principle, but the thing is also like 600 GB of data. Not sure I have the disk space to take a stand on this one.
There are no lite versions? I was trying to find a small LLM version I can run on an old machine and take it off the internet (or just firewall it) and play around with it to see if there is anything worth learning there for me. I was looking at the lite version of llama but when I tried to run the install on mint I ran into some issues and then had to many drinks to focus on it so I went back to something else. Maybe next weekend. If you have any recommendations I’m all ears
There are finetunes of Llama, Qwen, etc., based on DeepSeek that implement the same pre-response thinking logic, but they are ultimately still the smaller models with some tuning. If you want to run locally and don’t have tens of thousands to throw at datacenter-scale GPUs, those are your best option, but they differ from what you’d get in the Deepseek app.
No worries. I’ve got some 20TB drives i can throw these on 😁
Now, running a model that large… Well, I’ll just have to stick with the 8-13b params.
These models are mostly giant tables of weights that run on standardized framework software, right?
We’re talking about making illegal numbers again, aren’t we?
