In preparation of adopting future rocblas library options, it is necessary to track when the backward pass of training is executing. The scope-based helper class `BackwardPassGuard` is provided to toggle state.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/71881
Approved by: https://github.com/albanD
This PR modifies `lu_unpack` by:
- Using less memory when unpacking `L` and `U`
- Fuse the subtraction by `-1` with `unpack_pivots_stub`
- Define tensors of the correct types to avoid copies
- Port `lu_unpack` to be a strucutred kernel so that its `_out` version
does not incur on extra copies
Then we implement `linalg.lu` as a structured kernel, as we want to
compute its derivative manually. We do so because composing the
derivatives of `torch.lu_factor` and `torch.lu_unpack` would be less efficient.
This new function and `lu_unpack` comes with all the things it can come:
forward and backward ad, decent docs, correctness tests, OpInfo, complex support,
support for metatensors and support for vmap and vmap over the gradients.
I really hope we don't continue adding more features.
This PR also avoids saving some of the tensors that were previously
saved unnecessarily for the backward in `lu_factor_ex_backward` and
`lu_backward` and does some other general improvements here and there
to the forward and backward AD formulae of other related functions.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/67833
Approved by: https://github.com/IvanYashchuk, https://github.com/nikitaved, https://github.com/mruberry
Fixes#73298
I don't know whether `where` kernel actually supports type promotion, nor am I in the mood to find out, so it's manual type promotion.
Edit: nah, i can't tell TI to "promote to common dtype" because of bool condition, so manual type promotion is our only option.
I'll see what tests start failing and fix.
Uses some parts from #62084
Pull Request resolved: https://github.com/pytorch/pytorch/pull/76691
Approved by: https://github.com/mruberry
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/74226
Update signature of `scatter_reduce_` to match `scatter_/scatter_add_`
`Tensor.scatter_reduce_(int64 dim, Tensor index, Tensor src, str reduce)`
- Add new reduction options in ScatterGatherKernel.cpp and update `scatter_reduce` to call into the cpu kernel for `scatter.reduce`
- `scatter_reduce` now has the same shape constraints as `scatter_` and `scatter_add_`
- Migrate `test/test_torch.py:test_scatter_reduce` to `test/test_scatter_gather_ops.py`
Test Plan: Imported from OSS
Reviewed By: ngimel
Differential Revision: D35222842
Pulled By: mikaylagawarecki
fbshipit-source-id: 84930add2ad30baf872c495251373313cb7428bd
(cherry picked from commit 1b45139482e22eb0dc8b6aec2a7b25a4b58e31df)
The doc was indicating "If a dimension is not specified, the tensor will
be flattened", whereas the actual behavior is that the input tensor is
flattened only if the `dims` argument is not provided at all.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/74880
Approved by: https://github.com/albanD
Summary:
This PR ports `index_copy` implementation to structured kernels, also adds an `out` variant.
~Note to the reviewers: This is in draft mode, waiting for the tests from the CI, and I'll give a final look before requesting the review.~
Issue tracker: https://github.com/pytorch/pytorch/issues/55070
cc: bdhirsh ysiraichi
Pull Request resolved: https://github.com/pytorch/pytorch/pull/67329
Reviewed By: ejguan
Differential Revision: D34077219
Pulled By: bdhirsh
fbshipit-source-id: 6accda33957f654b753261c5c3d765a27a64d2c0
(cherry picked from commit f3ac83217a)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/71486
This PR adds upgraders for linspace and linspace.out as the optional step size will be deprecated soon. Old models will be using steps size of 100 when nothing is provided.
Test Plan: buck-out/gen/caffe2/test/jit#binary.par -r TestUpgraders.test_aten_linspace
Reviewed By: cccclai, mruberry
Differential Revision: D33654308
fbshipit-source-id: 0e0138091da0b11d4f49156eeb6bcd7e46102a5b
(cherry picked from commit 931ae4af32)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/69908
I also took this chance to clarify a bit the documentation of these
functions.
cc brianjo mruberry
Test Plan: Imported from OSS
Reviewed By: anjali411
Differential Revision: D33774417
Pulled By: mruberry
fbshipit-source-id: ab4a9014006783d1f87d432ecb959c854374c2d4
(cherry picked from commit f319a75d78)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/69827
In general, the current pattern allows for implementing optimisations
for all the backends in a common place (see for example the optimisation
for empty matrices).
After this PR, `torch.svd` is implemented in terms of `linalg.svd` and
`linalg.svdvals`, as expected. This makes it differentiable in the case
when `compute_uv=False`, although this is not particularly important, as
`torch.svd` will eventually be deprecated.
This PR also instantiates smaller `U` / `V` when calling cusolver_gesvdj
in the cases when `full_matrices=False` or `compute_uv=False`.
The memory for auxiliary `U` and `V` in the cases above, needed for some
cuSOLVER routines is allocated raw allocators rather than through fully
fledged tensors, as it's just a blob of memory the algorithm requests.
As the code is better structured now, it was easier to see that `U` and
`Vh` needn't be allocated when calling `svd_cusolver_gesvd`.
Now `linalg.svdvals` work as expected wrt the `out=` parameter.
Note that in the test `test_svd_memory_allocation` we were
passing a tensor of the wrong size and dtype and the test seemed to
pass...
This PR also changes the backward formula to avoid saving the input
matrix, as it's not necessary. In a follow up PR, I will clean the
backward formula and make it more numerically stable and efficient.
This PR also does a number of memory optimisations here and there, and fixes
the call to cusolver_gesvd, which were incorrect for m <= n. To test
this path, I compiled the code with a flag to unconditionally execute
the `if (!gesvdj_convergence_check.empty())` branch, and all the tests
passed.
I also took this chance to simplify the tests for these functions in
`test_linalg.py`, as we had lots of tests that were testing some
functionality that is already currently tested in the corresponding
OpInfos. I used xwang233's feature to test both MAGMA and CUDA
backends. This is particularly good for SVD, as cuSOLVER is always
chosen over MAGMA when available, so testing MAGMA otherwise would be
tricky.
cc jianyuh nikitaved pearu mruberry walterddr IvanYashchuk xwang233 Lezcano
Test Plan: Imported from OSS
Reviewed By: mikaylagawarecki
Differential Revision: D33751983
Pulled By: mruberry
fbshipit-source-id: 11d48d977946345583d33d14fb11a170a7d14fd2
(cherry picked from commit a1860bd567)
Summary:
Fixes https://github.com/pytorch/pytorch/issues/65908
Added a new overload instead of updating the current signature. (Had issues with JIT and **maybe** it would have been FC breaking)
TODO:
* [x] Don't compute `std::pow(10, decimals)` for each element.
* [x] Update docs (https://docs-preview.pytorch.org/66195/generated/torch.round.html?highlight=round#torch.round)
* [x] Add tests
* ~~Should we try to make it composite?~~
* ~~Should we add specialized test with more values of `decimals` outside of OpInfo with larger range of values in input tensor?~~
cc mruberry rgommers
Pull Request resolved: https://github.com/pytorch/pytorch/pull/66195
Reviewed By: anjali411
Differential Revision: D31821385
Pulled By: mruberry
fbshipit-source-id: 9a03fcb809440f0c83530108284e69c345e1850f
(cherry picked from commit 50b67c6968)
