🐛 Describe the bug
Currently, the Python pytree does not guarantee "equal dicts imply equal leaves and equal treespecs".
In [1]: import torch.utils._pytree as pytree
In [2]: {'a': 1, 'b': 2} == {'b': 2, 'a': 1}
Out[2]: True
In [3]: pytree.tree_leaves({'a': 1, 'b': 2})
Out[3]: [1, 2]
In [4]: pytree.tree_leaves({'b': 2, 'a': 1})
Out[4]: [2, 1]
In [5]: pytree.tree_structure({'a': 1, 'b': 2})
Out[5]:
TreeSpec(dict, ['a', 'b'], [*,
*])
In [6]: pytree.tree_structure({'b': 2, 'a': 1})
Out[6]:
TreeSpec(dict, ['b', 'a'], [*,
*])
In [7]: pytree.tree_leaves({'a': 1, 'b': 2}) == pytree.tree_leaves({'b': 2, 'a': 1})
Out[7]: False
In [8]: pytree.tree_structure({'a': 1, 'b': 2}) == pytree.tree_structure({'b': 2, 'a': 1})
Out[8]: FalseSee #114947 (comment) for more explanation.
Versions
Versions
Collecting environment information...
PyTorch version: 2.2.0.dev20231121
Is debug build: False
CUDA used to build PyTorch: None
ROCM used to build PyTorch: N/A
OS: macOS 12.7.1 (x86_64)
GCC version: Could not collect
Clang version: 17.0.5
CMake version: version 3.27.8
Libc version: N/A
Python version: 3.11.5 (main, Sep 11 2023, 08:19:27) [Clang 14.0.6 ] (64-bit runtime)
Python platform: macOS-10.16-x86_64-i386-64bit
Is CUDA available: False
CUDA runtime version: No CUDA
CUDA_MODULE_LOADING set to: N/A
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True
CPU:
Intel(R) Core(TM) i7-4980HQ CPU @ 2.80GHz
Versions of relevant libraries:
[pip3] flake8==6.0.0
[pip3] flake8-bugbear==23.3.23
[pip3] flake8-comprehensions==3.12.0
[pip3] flake8-executable==2.1.3
[pip3] flake8-logging-format==0.9.0
[pip3] flake8-pyi==23.3.1
[pip3] flake8-simplify==0.19.3
[pip3] mypy==1.6.0
[pip3] mypy-extensions==1.0.0
[pip3] numpy==1.26.0
[pip3] optree==0.10.0
[conda] blas 1.0 mkl
[conda] mkl 2023.1.0 h8e150cf_43560
[conda] numpy 1.26.0 pypi_0
[conda] optree 0.10.0 pypi_0
[conda] torchfix 0.2.0 pypi_0
cc @zou3519
🐛 Describe the bug
Currently, the Python pytree does not guarantee "equal
dicts imply equal leaves and equal treespecs".See #114947 (comment) for more explanation.
Versions
Versions
cc @zou3519