4 ms·
So, rckrd wrote a (rather short) article about the license of Llama 2 but got it all wrong: Even though the press calls it open-source, it's not. Open Source h
by Tepix 3y ago
So, rckrd wrote a (rather short) article about the license of Llama 2 but got it all wrong: Even though the press calls it open-source, it's not.
Open Source has a very clear definition. Llama 2 fails in multiple regards.
First, the license by itself is not an open source license. It has important restrictions that make it non-open source.
Second, the distribution. You have to apply for the download with a web form and you are not allowed to redistribute the model.
Third, the source code, i.e. the data used to train the model. Meta is not telling us about it. One of the core features of open source is that you can recreate the binary (in this case, the model weights) by yourself. You can't do that here.
- charcircuit 3y agoYou can redistribute the model. A model is not a binary. Training code is not source code for the model. The training code is closer to an editor. You can release open source software created by a closed source IDE.
- Tepix 3y ago> You can redistribute the model. That's great. I disagree about the other part.
- catchnear4321 3y agothe source of the model is code and data. you cannot generate the model without the same code and training data. it is not open source. you are conflating editing the model with the capability to edit a thing. using your source code example, altering source code changes the binary. as if one were editing the binary. the source code isn’t called an IDE.
- fragmede 3y agoWhat's the definition of binary you're using here? Because a model is an inscrutable blob of bytes that is used (with the help of libaries and other glue code) to perform a function.
- taneq 3y agoAre we gonna have to do the FOSS vs. 'source available' thing again? Also, while I think your 'source=training set+model, binary=weights' analysis is correct, I'm not sure we have 100% consensus on this yet?