storch
storch copied to clipboard
sbrunk
GPU accelerated deep learning and numeric computing for Scala 3.
Response to request in issue https://github.com/sbrunk/storch/issues/44. Attempt to rewrite the "pico" example from Karpathy's ["Let's build GPT: from scratch, in code, spelled out"](https://www.youtube.com/watch?v=kCc8FmEb1nY) in storch.
With compile-time tracked tensor shapes as discussed #63 we'll probably need three type parameters like `Tensor[DType, Shape, Device]`. While this is great for type-safety, it makes the Tensor type a...
Compile-time tracked tensor shapes means that the shape of a tensor, or parts of it, is statically known at compile time. ### Why? Statically know shapes can be useful for...
Add a weight and bias initialization method to the `nn.Module` so we can set these values via an [`apply` method like PyTorch](https://stackoverflow.com/questions/49433936/how-do-i-initialize-weights-in-pytorch) that does this. Reference to Python documentation [here](https://pytorch.org/docs/stable/generated/torch.nn.Module.html#torch.nn.Module.apply)....
The `Promoted` type can cause a mismatch with the runtime promotion in certain cases when combining tensors with scalar values. The reason is, that the type promotion , somewhat unintuitively,...
It would be great to add a few transformer architectures. This will also help us to prioritize which op and module implementations to add next. It probably makes sense to...
Currently [`TensorModule`](https://github.com/sbrunk/storch/blob/main/core/src/main/scala/torch/nn/modules/Module.scala#L150-L151) is parametrized on a single type, so keeps the transformation within the same DType: ```scala trait TensorModule[D Tensor[D]): override def toString(): String = "TensorModule" ``` However there are...
Hello! I stumbled upon a fatal error while using `torch.split` + reassignment of a tensor - not sure how to even start debugging this, but I am documenting it here...
We should provide a few examples in the docs and perhaps also as subproject showing things like data loading and a complete training loop. Some interesting examples could be: -...
Hello I've been playing around with storch for the past few days. It was a very pleasant experience so far - thanks a lot for this great effort. Today I...
