Add memory format support to the `resize_` op. #28292

VitalyFedyunin · 2019-10-18T17:01:59Z

Rationale to add optional memory format to resize_as_ and resize_ operators is the fact that they are frequently used as inplace empty_like and empty operators inside of our code base. So having them to accept memory format similarly to empty_like and empty seems to be logical.

We could also add two new operators, but it will be more confusing (taking into account that renaming existing operators in not an option).

Stack from ghstack:

[WIP] Experiments #29461 [WIP] Experiments
[WIP] Switch default of clone, to, *_like ops #29460 [WIP] Switch default of clone, to, *_like ops
[WIP] Prepare copy op to switch memory format preservation defaults #29459 [WIP] Prepare copy op to switch memory format preservation defaults
Adding function to convert Module to channels last #28991 Adding function to convert Module to channels last
Add torch.memory_format support to the TorchScript #28544 Add torch.memory_format support to the TorchScript
Add memory format support to the resize_ op. #28292 Add memory format support to the resize_ op.
TensroIterator preserve format for binary, ternary operators. #28291 TensroIterator preserve format for binary, ternary operators.
Add channels last support to cuda.comm.scatter and gather #28077 Add channels last support to cuda.comm.scatter and gather
Kill operator== of TensorOptions as confusing one #28076 Kill operator== of TensorOptions as confusing one
Add memory format support to resize_as_ operator #27979 Add memory format support to resize_as_ operator

Allows to simplify patterns like:

output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeCosizeHosizeW, 1, osizeW*sizeC, sizeC});
output.resize_({nbatch, nInputPlane, outputHeight, outputWidth});
indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth});
output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format);
indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format);
gradInput.resize_as_(input);
gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format);

Differential Revision: D18044978

Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); [ghstack-poisoned]

VitalyFedyunin · 2019-10-18T17:03:00Z

CC @jjsjann123

Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); [ghstack-poisoned]

Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); ghstack-source-id: de38e28 Pull Request resolved: #28292

Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); [ghstack-poisoned]

Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); ghstack-source-id: f373cbe Pull Request resolved: #28292

…op." Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); [ghstack-poisoned]

Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); ghstack-source-id: af48d69 Pull Request resolved: #28292

Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); Differential Revision: [D18044978](https://our.internmc.facebook.com/intern/diff/D18044978) [ghstack-poisoned]

Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); ghstack-source-id: c548695 Pull Request resolved: #28292

Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); Differential Revision: [D18044978](https://our.internmc.facebook.com/intern/diff/D18044978) [ghstack-poisoned]

Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); ghstack-source-id: ebad22f Pull Request resolved: #28292

…to the `resize_` op." Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); Differential Revision: [D18044978](https://our.internmc.facebook.com/intern/diff/D18044978) [ghstack-poisoned]

…"Add memory format support to the `resize_` op." Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); Differential Revision: [D18044978](https://our.internmc.facebook.com/intern/diff/D18044978) [ghstack-poisoned]

Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); Differential Revision: [D18044978](https://our.internmc.facebook.com/intern/diff/D18044978) [ghstack-poisoned]

…op." Rationale to add optional memory format to `resize_as_` and `resize_` operators is the fact that they are frequently used as inplace `empty_like` and `empty` operators inside of our code base. So having them to accept memory format similarly to `empty_like` and `empty` seems to be logical. We could also add two new operators, but it will be more confusing (taking into account that renaming existing operators in not an option). --- Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); Differential Revision: [D18044978](https://our.internmc.facebook.com/intern/diff/D18044978) [ghstack-poisoned]

…esize_` op." Rationale to add optional memory format to `resize_as_` and `resize_` operators is the fact that they are frequently used as inplace `empty_like` and `empty` operators inside of our code base. So having them to accept memory format similarly to `empty_like` and `empty` seems to be logical. We could also add two new operators, but it will be more confusing (taking into account that renaming existing operators in not an option). --- Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); Differential Revision: [D18044978](https://our.internmc.facebook.com/intern/diff/D18044978) [ghstack-poisoned]

Rationale to add optional memory format to `resize_as_` and `resize_` operators is the fact that they are frequently used as inplace `empty_like` and `empty` operators inside of our code base. So having them to accept memory format similarly to `empty_like` and `empty` seems to be logical. We could also add two new operators, but it will be more confusing (taking into account that renaming existing operators in not an option). --- Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); Differential Revision: [D18044978](https://our.internmc.facebook.com/intern/diff/D18044978) [ghstack-poisoned]

…size_` op." Rationale to add optional memory format to `resize_as_` and `resize_` operators is the fact that they are frequently used as inplace `empty_like` and `empty` operators inside of our code base. So having them to accept memory format similarly to `empty_like` and `empty` seems to be logical. We could also add two new operators, but it will be more confusing (taking into account that renaming existing operators in not an option). --- Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); Differential Revision: [D18044978](https://our.internmc.facebook.com/intern/diff/D18044978) [ghstack-poisoned]

Rationale to add optional memory format to `resize_as_` and `resize_` operators is the fact that they are frequently used as inplace `empty_like` and `empty` operators inside of our code base. So having them to accept memory format similarly to `empty_like` and `empty` seems to be logical. We could also add two new operators, but it will be more confusing (taking into account that renaming existing operators in not an option). --- Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); Differential Revision: [D18044978](https://our.internmc.facebook.com/intern/diff/D18044978) [ghstack-poisoned]

VitalyFedyunin · 2019-11-12T00:48:34Z

CC @gchanan similar to resize_as_ PR.

gchanan · 2019-11-13T22:21:40Z

aten/src/ATen/native/Resize.cpp

  self_->maybe_zero_dim(size.size() == 0);
+  if (optional_memory_format.has_value()) {
+    auto memory_format =
+        optional_memory_format.value_or(MemoryFormat::Contiguous);


you already know the tensor has a value, you don't need to use value_or.

gchanan · 2019-11-13T22:24:43Z

aten/src/ATen/native/ResizeCommon.h

      "). This may be caused by passing a named tensor ",
      "as an `out=` argument; please ensure that the sizes are the same. ");
+  TORCH_CHECK(
+      optional_memory_format.value_or(MemoryFormat::Contiguous) ==


shouldn't this just checking if optional_memory_format has a value?

The behavior is we only restride if there's a value, and here we don't restride.

gchanan · 2019-11-13T22:25:07Z

aten/src/ATen/native/cuda/Resize.cu

  self_->maybe_zero_dim(size.size() == 0);
+  if (optional_memory_format.has_value()) {
+    auto memory_format =
+        optional_memory_format.value_or(MemoryFormat::Contiguous);


same as CPU, this should just use value.

gchanan · 2019-11-13T22:26:19Z

aten/src/ATen/native/quantized/cpu/tensor_operators.cpp

+    Tensor& self,
+    IntArrayRef size,
+    c10::optional<MemoryFormat> optional_memory_format) {
+  auto memory_format =


same as the named tensor case, shouldn't this check for not having a value?

Because this behavior differs from non-quantized tensor, where we will actually make something contiguous even if the size matches.

Rationale to add optional memory format to `resize_as_` and `resize_` operators is the fact that they are frequently used as inplace `empty_like` and `empty` operators inside of our code base. So having them to accept memory format similarly to `empty_like` and `empty` seems to be logical. We could also add two new operators, but it will be more confusing (taking into account that renaming existing operators in not an option). --- Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); Differential Revision: [D18044978](https://our.internmc.facebook.com/intern/diff/D18044978) [ghstack-poisoned]

gchanan · 2019-11-16T00:04:23Z

aten/src/ATen/native/quantized/cpu/tensor_operators.cpp

+    IntArrayRef size,
+    c10::optional<MemoryFormat> optional_memory_format) {
+  TORCH_CHECK(
+      !optional_memory_format.has_value(),


are you sure we need this? quantized tensors should just be able to resized in the same way.

gchanan

I'm okay with this but you should get someone from quantization to test/okay the quantization parts.

facebook-github-bot · 2019-11-18T14:03:38Z

@VitalyFedyunin merged this pull request in cb43170.

Summary: Pull Request resolved: pytorch/pytorch#28292 Allows to simplify patterns like: 1. output.resize_({sizeB, sizeC, osizeH, osizeW}).as_strided_({sizeB, sizeC, osizeH, osizeW}, {sizeC*osizeH*osizeW, 1, osizeW*sizeC, sizeC}); 2. output.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); indices.resize_({nbatch, nInputPlane, outputHeight, outputWidth}); output.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); indices.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); 3. gradInput.resize_as_(input); gradInput.unsafeGetTensorImpl()->empty_tensor_restride(memory_format); Test Plan: Imported from OSS Differential Revision: D18044978 Pulled By: VitalyFedyunin fbshipit-source-id: bbf67c25f9cf88bc6e949089a3b247df50f86dc4

VitalyFedyunin requested a review from apaszke as a code owner October 18, 2019 17:01

VitalyFedyunin requested a review from ifedan October 18, 2019 17:02

VitalyFedyunin mentioned this pull request Oct 23, 2019

Add torch.memory_format support to the TorchScript #28544

Closed

This was referenced Nov 8, 2019

[WIP] Prepare copy op to switch memory format preservation defaults #29459

Closed

[WIP] Switch default of clone, to, *_like ops #29460

Closed

[WIP] Experiments #29461

Closed

VitalyFedyunin added 6 commits November 8, 2019 10:23

VitalyFedyunin requested a review from gchanan November 11, 2019 22:15

VitalyFedyunin added 3 commits November 11, 2019 16:38

gchanan requested changes Nov 13, 2019

View reviewed changes

VitalyFedyunin added 2 commits November 14, 2019 09:40

VitalyFedyunin requested a review from gchanan November 14, 2019 18:21

VitalyFedyunin added 2 commits November 14, 2019 19:30

gchanan reviewed Nov 16, 2019

View reviewed changes

gchanan approved these changes Nov 16, 2019

View reviewed changes

facebook-github-bot closed this in cb43170 Nov 18, 2019

facebook-github-bot added the merged label Nov 18, 2019

facebook-github-bot deleted the gh/VitalyFedyunin/18/head branch November 21, 2019 15:16

mruberry added the Merged label Oct 28, 2020

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Add memory format support to the `resize_` op. #28292

Add memory format support to the `resize_` op. #28292

Uh oh!

VitalyFedyunin commented Oct 18, 2019 •

edited

Loading

Uh oh!

VitalyFedyunin commented Oct 18, 2019

Uh oh!

VitalyFedyunin commented Nov 12, 2019

Uh oh!

gchanan Nov 13, 2019

Uh oh!

gchanan Nov 13, 2019

Uh oh!

gchanan Nov 13, 2019

Uh oh!

gchanan Nov 13, 2019

Uh oh!

gchanan Nov 16, 2019

Uh oh!

gchanan left a comment

Uh oh!

facebook-github-bot commented Nov 18, 2019

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Add memory format support to the resize_ op. #28292

Add memory format support to the resize_ op. #28292

Uh oh!

Conversation

VitalyFedyunin commented Oct 18, 2019 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

VitalyFedyunin commented Oct 18, 2019

Uh oh!

VitalyFedyunin commented Nov 12, 2019

Uh oh!

gchanan Nov 13, 2019

Choose a reason for hiding this comment

Uh oh!

gchanan Nov 13, 2019

Choose a reason for hiding this comment

Uh oh!

gchanan Nov 13, 2019

Choose a reason for hiding this comment

Uh oh!

gchanan Nov 13, 2019

Choose a reason for hiding this comment

Uh oh!

gchanan Nov 16, 2019

Choose a reason for hiding this comment

Uh oh!

gchanan left a comment

Choose a reason for hiding this comment

Uh oh!

facebook-github-bot commented Nov 18, 2019

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Add memory format support to the `resize_` op. #28292

Add memory format support to the `resize_` op. #28292

VitalyFedyunin commented Oct 18, 2019 •

edited

Loading