OP, you should not have written this in Rust. It should be in PyTorch, which is by far the most popular. We can’t tell if this architecture is good or whether there is a problem in your implementation.
You can test the whole thing for free on a GPU with Google Colab. Test both the transformer and your new architecture on a larger dataset. Something that maxes out the GPU for an hour each run.
Also, the readme mentions keeping the same optimizer schedule which sounds nice at first but they are completely different architectures. The loss is high on the transformer, did you try raising the learning rate on it?
In general I’m interested in parameter efficient architectures. I don’t think transformers are optimal, and indeed many improvements have been made to vanilla transformers. But if you have an idea for something better you need to show it.
Apparently op cares a lot about speed, which is fine, but ML researchers care about correctness first, speed second. And it makes sense, because they are not as resource constrained as OP.
Most PyTorch tensor operations are cython not python. So imo rewriting in rust is not going to have an enormous speed up. If that really was the concern we should see a throughput comparison vs PyTorch or something.
Rust has its place but python is default for this kind of thing. By doing it in rust, they have now changed two things: the implementation of the transformer, and this new model.
Yeah I read the title and did a double-take. It's an uninteresting choice for systems like these. The more pressing concerns, which go undescribed, are the exact mathematical choices behind the actual model. Rust provides almost zero value here because tensor stuff is all just 2d-arrays of floats for the most part.
OP, you should not have written this in Rust. It should be in PyTorch, which is by far the most popular. We can’t tell if this architecture is good or whether there is a problem in your implementation.
You can test the whole thing for free on a GPU with Google Colab. Test both the transformer and your new architecture on a larger dataset. Something that maxes out the GPU for an hour each run.
Also, the readme mentions keeping the same optimizer schedule which sounds nice at first but they are completely different architectures. The loss is high on the transformer, did you try raising the learning rate on it?
In general I’m interested in parameter efficient architectures. I don’t think transformers are optimal, and indeed many improvements have been made to vanilla transformers. But if you have an idea for something better you need to show it.
I dunno, I could probably be convinced to try a new tool purely on the basis of not having to deal with installing pytorch
Apparently op cares a lot about speed, which is fine, but ML researchers care about correctness first, speed second. And it makes sense, because they are not as resource constrained as OP.
Most PyTorch tensor operations are cython not python. So imo rewriting in rust is not going to have an enormous speed up. If that really was the concern we should see a throughput comparison vs PyTorch or something.
You can just tell when the idea itself was generated
How would it compare to current state of the art tested state space layers like Kimi Delta Attention?
Homie just had some tokens to burn at the end of the month and “in Rust” is pure HN clickbait.
Isn't this just RNN and nearly nothing about this is actually new?
“In Rust” man of all the cliche title baits, I hate this one the most.
Why did you choose Rust? Why does that matter?
Nothing wrong with Rust. Lots wrong with the bandwagon that “in Rust” somehow adds value.
Rust has its place but python is default for this kind of thing. By doing it in rust, they have now changed two things: the implementation of the transformer, and this new model.
Counterpoint: I only clicked on it because it said "in Rust".
Yeah I read the title and did a double-take. It's an uninteresting choice for systems like these. The more pressing concerns, which go undescribed, are the exact mathematical choices behind the actual model. Rust provides almost zero value here because tensor stuff is all just 2d-arrays of floats for the most part.
Seems similar to a neural turing machine.
vibe coded vibe coder!