3 ms·
This is sort of the polar opposite of how modern high performance ML frameworks are built. Skimming the code, there's a ton of boilerplate for the various oper
by dekhn 4y ago
This is sort of the polar opposite of how modern high performance ML frameworks are built. Skimming the code, there's a ton of boilerplate for the various operations that could be library-ized and generified, if that makes sense.
I actually really like minimal implementations of state-of-the-art systems because the code is much easier to understand (modern frameworks are super-complex) but I wonder what it means long-term if you don't need frameworks.
- SCLeo 4y agoThis is just inference. The core at most ML library is the auto differentiation capability. It will be extremely tedious if you are to calculate the gradients manually. Or, if you implemented your own AD, then it is effectively a minified version of a ML library.
- version_five 4y agollama.cpp/ggml.h // GGML Tensor Library ... // This library implements: // - a set of tensor operations // - automatic differentiation // - basic optimization algorithms
- mhh__ 4y ago> autodiff There's a lot of // TODO: implement backward in there
- version_five 4y agoOk thanks for digging deeper, I didn't realize that, and obviously that invalidates the excerpt I posted above