f45f6e86b9
Written with Claude Code. Fixes https://github.com/pytorch/pytorch/issues/157569 Fixes https://github.com/pytorch/pytorch/issues/158134 NumPy and PyTorch handle advanced indexing differently when advanced indices are separated by slices (e.g., arr[:, [0], :, 0]). PyTorch uses "outer" indexing placing result dimensions in original positions, while NumPy uses "vectorized" indexing moving advanced index dimensions to the front. This adds _numpy_style_advanced_indexing() to detect separated advanced indices and transpose results to match NumPy's dimension ordering, ensuring torch._numpy maintains compatibility with NumPy's indexing behavior. Fixes cases like: - arr[:, [0], :, 0] now returns shape (1, 5, 7) instead of (5, 1, 7) - arr[:, [0, 1], :, 0] now returns shape (2, 5, 7) instead of (5, 2, 7) Pull Request resolved: https://github.com/pytorch/pytorch/pull/157676 Approved by: https://github.com/manuelcandales Co-authored-by: Claude <noreply@anthropic.com>
NumPy <> PyTorch Compat Layer
This folder contains an implementation of (most of) the NumPy public API using PyTorch tensors. Note that this folder does not depend on NumPy in any way. This is a standalone implementation.
This implementation is used by Dynamo to through NumPy code and lower it into PyTorch code.
To see design decisions that went into this implementation, please see the rfc.
Structure of the code
This folder exports a drop-in replacement for the NumPy namespace and its modules linalg, fft and random via its __init__.py.
The implementation is split into files that work with PyTorch objects (PyTorch Tensors, dtypes, etc) and files that
use these PyTorch-only files and convert them into functions/objects that can process all the types that the NumPy functions
accept. In particular, they accept torch._numpy.dtypes or torch._numpy.ndarrays.
The PyTorch-only files are the *_impl.py files, while the wrapper files are those that do not have an *_impl.py. This creates a
hierarchy, wherein, for example, _dtypes.py will import _dtypes_impl.py, but not the other way around. In particular, *_impl.py
will only depend on other *_impl.py files.
As discussed in the rfc, we use types as tags in our PyTorch implementations. We then use
a decorator called normalizer that will inspect these types and preprocess the inputs before sending them to the function. This
preprocessing is the one in charge of mapping array-like objects into Tensors, dtype-like objects into PyTorch dtypes, implement
the out= behaviour and so on.
In the files _funcs.py and _ufuncs.py we use register the normalizer decorator to all the *_impl.py functions.
In the file _ndarray.py we define the ndarray class, which is just a thin wrapper around a PyTorch tensor. We use the free functions
and a bit of metaprogramming to implement many of the methods.
Adding a new function
You just need to add a function in the relevant *_impl.py file. You will need to tag the inputs with the relevant Types. After that, you
can assume that the inputs are all PyTorch objects. Your function should return PyTorch tensors. The normalizer will make sure that you
always get PyTorch objects. If in doubt, you can see the implementation of the normalization attached to each type annotation in the file
_normalizations.py.
Debugging
It may be useful to figure out whether a given bug is caused by dynamo or the compatibility layer. You may use the compat layer in eager mode
simply by changing import numpy as np by import torch._numpy as np in your program, without having to call torch.compile at all.
Note that torch._numpy will be quite slow when used in eager mode, and it is in no way a replacement or an alternative to the regular PyTorch API.
This should only be used as a debugging tool.