But where's the source? I just see a binary blob, what makes it open source?
The training process is fully deterministic. It's just an algorithm. Feed the same data in and you'll get the same weights out.
If you're speaking about the computational cost, it used to be that way for compilers too. Give it 20 years and you'll be able to train one of today's models on your phone.
No it is not. The training process is non-deterministic, and given exactly the same data, the same code and the same seeds you'll get different weights. Even the simplest operations like matrix multiplication will give you slightly different results depending on the hardware you're using (e.g. you'll get different results on CPU, on GPU from vendor #1 and on GPU from vendor #2, and probably on different GPUs from the same vendor, and on different CUDA versions, etc.), but also depending on the dimensions of the matrices you'll get different results (e.g. if you fuse the QKV weights from modern transformers into a single matrix and do a single multiplication instead of multiplying each separately you'll get different results), and some algorithms (e.g. backwards pass of Flash Attention) are explicitly non-deterministic to be faster.
That has everything to do with implementation, and nothing to do with algorithm. There is an important difference.
Math is deterministic. The way [random chip] implements floating point operations may not be.
Lots of scientific software has the ability to use IEEE-754 floats for speed or to flip a switch for arbitrary precision calculations. The calculation being performed remains the same.
The point is none of these models are trained with pure "math". It doesn't matter that you can describe a theoretical training process using a set of deterministic equations, because in practice it doesn't work that way. Your claim that "the training process is fully deterministic" is objectively wrong in this case because none of the non-toy models use (nor they practically can use) such a deterministic process. There is a training process which is deterministic, but no one uses it (for good reasons).
If you had infinite budget, exactly the same code, the same training data, and even the same hardware you would not be able to reproduce the weights of Deepseek R1, because it wasn't trained using a deterministic process.
torch.manual_seed(0)
random.seed(0)
np.random.seed(0)
torch.use_deterministic_algorithms(True)
Where would other non-determinism come from?
I'm open to there being another source. I'd just like to know what it would be. I haven't found one yet.
Point at the science that says that, please: Current scientific knowledge doesn't agree with you.
I'd love a citation. So far you haven't even suggested a possible source for this non-determinism you claim exists.
Training is reproducible only if, besides the pipeline and data, you also start from the same random weights.
It sounds like you're referring to something like simulated annealing. Using that as an example, the fundamental requirement is to introduce arbitrary, uncorrelated steps -- there's no requirement that the steps be random, and the only potential advantage of using a random source is that it provides independence (lack of correlation) inherently; but in exchange, it makes testing and reproduction much harder. Basically every use of simulated annealing or similar I've run into uses pseudorandom numbers for this reason.
There simply aren't that many sources of non-determinism in a modern computer.
Though I'll grant that if you've engineered your codebase for speed and not for determinism, error can creep in via floating point error, sloppy ordering of operations, etc. These are not unavoidable implementation details, however. CAD kernels and other scientific software do it every day.
When you boil down what's actually happening during training, it's just a bunch of matrix math. And math is highly repeatable. Size of the matrix has nothing to do with it.
I have little doubt that some implementations aren't deterministic, due to software engineering choices as discussed above. But the algorithms absolutely are. Claiming otherwise seems equivalent to claiming that 2 + 2 can sometimes equal 5.
Not some of them; ALL OF THEM. Engineering training pipelines for absolute determinism would be, quite frankly, extremely dumb, so no one does it. When you need millions of dollars worth of compute to train a non-toy model are you going to double or triple your cost just so that the process is deterministic, without actually making the end result perform any better?
The cost of adaptive precision floats can be negligible depending on application. One example I'm familiar with from geometry processing: https://www.cs.cmu.edu/~quake/robust.html
Integer math often carries no performance penalty compared to floating point.
I guess my takeaway from this conversation is that there's a market for fast high-precision math techniques in the AI field.
E.g. see all the specialized third-party models out there based on Qwen.
"Open-source" is the wrong word here, what they mean is "you can modify and redistribute these weights".
Also, the "redistribute" part is key here.
Fully agree, it isn't. Reverse engineering isn't necessary for modifying compiled program behaviour, so comparing it to finetuning is not applicable. Finetuning applied to program domain would be more like adding plugins or patching in some compiled routines. Reverse-engineering applied to models would be like extracting source documents from weights.
> Finetuning is a standard supported workflow for these models.
Yes, so is adding mods for some games, just put your files in a designated folder and game automatically picks it up and does required modifications.
> Also, the "redistribute" part is key here.
It is not. Redistributability and being open source is orthogonal. You can have a source for a program and not be able to redistribute source or program, or you can redistribute a compiled program, but not have it's source (freeware).
https://chatgpt.com/share/6838c070-705c-8005-9a88-83c9a5550a...