pytorch
pytorch copied to clipboard
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Stack from [ghstack](https://github.com/ezyang/ghstack): * #83079 * __->__ #83073 There's no reason why it should just apply when you import vmap Test Plan: - added a new test
I think the person who edited `mean.dim` edited the python and the associated Torchscript version manually in two different ways. This diff fixes that all up. It also fixes the...
### 🐛 Describe the bug Sometimes, _post_backward_hook will not fire if gradients were not accumulated on the FSDP managed parameter, such as if all parameters in an FSDP module were...
Stack from [ghstack](https://github.com/ezyang/ghstack): * #83119 * #83114 * #83077 * __->__ #83106 * #83105 * #83080 Reland of #83017. It was reverted because it was stacked on top of a...
Stack from [ghstack](https://github.com/ezyang/ghstack): * #83119 * #83114 * #83077 * #83106 * __->__ #83105 * #83080 The previous merge attempt failed some trunk tests. Test Plan: - wait for tests
Stack from [ghstack](https://github.com/ezyang/ghstack): * __->__ #79571 * #83092 * #82850 Somehow even with clang-format off, it was unhappy with this line >>> Lint for torch/csrc/jit/runtime/serialized_shape_function_registry.cpp: Warning (CLANGFORMAT) format See https://clang.llvm.org/docs/ClangFormat.html....
Summary: This change exposes an out-of-place `_reduce_oop` API from the ProcessGroupNCCL. It allows reducing an input tensor and placing the output in a separate output tensor. Custom collectives may be...
- Better definition of `state_dict` - Additional shallow copy warning
### 🐛 Describe the bug ```python import torch import torch.fx tracer = torch.fx.Tracer() import copy tracer_copy = copy.deepcopy(tracer) ``` ``` --------------------------------------------------------------------------- TypeError Traceback (most recent call last) [](https://localhost:8080/#) in ()...
Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * #82819 * __->__ #82214 * #82817