If you are here because you ran into a runtime error due to a missing feature or some kind of bug, please open an issue and fill in the appropiate template. If you have general feedback about this prototype you can use our suggested template or just open a free-form issue if you like. Thank you for contributing to this project!
If you are new to this project, we recommend you take a look at our whirlwind introduction to get started.
Due to missing extensibility features of PyTorch nestedtensor currently lacks autograd support. We're actively working on this and recognize that it severely limits the applicability of the project. Please run nestedtensor operations within the inference mode context to prevent any adverse interactions with the autograd system.
sentences = [torch.randn(10, 5), torch.randn(5, 5), torch.randn(9, 5)] with torch.inference_mode(): nt = nestedtensor.nested_tensor(sentences) nt.sum(1)
Due to the development velocity of PyTorch the nestedtensor project is built on top of and dependent on a fixed, recent PyTorch nightly.
When installing a binary please specify the corresponding torch nightly link archive to automatically pull in the correct PyTorch nightly.
pip install https://download.pytorch.org/nestedtensor/whl/nightly/cpu/py3.7/nestedtensor-0.1.1_cpu-cp37-cp37m-linux_x86_64.whl -f https://download.pytorch.org/whl/nightly/cpu/torch_nightly.html
pip install https://download.pytorch.org/nestedtensor/whl/nightly/cu102/py3.7/nestedtensor-0.1.1_cu102-cp37-cp37m-linux_x86_64.whl -f https://download.pytorch.org/whl/nightly/cu102/torch_nightly.html
In general we batch data for efficiency, but usually batched kernels need, or greatly benefit from, regular, statically-shaped data.
One way of dealing with dynamic shapes then, is via padding and masking. Various projects construct masks that, together with a data Tensor, are used as a representation for lists of dynamically shaped Tensors.
Obviously this is inefficient from a memory and compute perspective if the Tensors within this list are sufficiently diverse.
You can also trace through the codebase where these masks are used and observe the kind of code this approach often leads to. See for example universal_sentence_embedding.
Otherwise we also have one-off operator support in PyTorch that aims to support dynamic shapes via extra arguments such as a padding index. Of course, while these functions are fast and sometimes memory efficient, they don't provide a consistent interface.
Other users simply gave up and started writing for-loops, or discovered that batching didn't help.
We want to have a single abstraction that is consistent, fast, memory efficient and readable and the nestedtensor project aims to provide that.
NestedTensors are a generalization of torch Tensors which eases working with data of different shapes and lengths. In a nutshell, Tensors have scalar entries (e.g. floats) and NestedTensors have Tensor entries. However, note that a NestedTensor is still a Tensor. That means it needs to have a single dimension, single dtype, single device and single layout.
Tensor entry constraints:
The nestedtensor package is a prototype intended for early stage feedback and testing. It is on the road to a beta classification, but there is no definitive timeline yet. See PyTorch feature classification for what prototype, beta and stale means.
The project is under active development. If you have a suggestions or found a bug, please file an issue!