Advanced Indexing Part 1 -- Purely Integer Array Indexing by killeent · Pull Request #1588 · pytorch/pytorch

killeent · 2017-05-18T19:57:22Z

This PR implements "Purely Integer Array" advanced indexing semantics in PyTorch, from NumPy, as documented here: https://docs.scipy.org/doc/numpy/reference/arrays.indexing.html#purely-integer-array-indexing. It should work with gets/sets and autograd, and allows for LongTensor indexers as well.

 x = torch.arange(0, 25).view(5, 5)

  0   1   2   3   4
  5   6   7   8   9
 10  11  12  13  14
 15  16  17  18  19
 20  21  22  23  24
[torch.FloatTensor of size 5x5]

x[[0,4,3], [1,3,0]]

  1
 23
 15
[torch.FloatTensor of size 3]

Begins to address #1080.

torch/csrc/generic/Tensor.cpp

killeent · 2017-06-12T16:38:58Z

This is now in a reviewable state cc @fmassa.

torch/csrc/generic/Tensor.cpp

kentsommer · 2017-06-14T03:49:11Z

Is this indexing supposed to work on tensors that have been pushed to the gpu with .cuda()?

If it is, I encountered this issue today while testing the indexing out:

(apytorch) kent@kbox ~/Documents/misc-git-repos $ python
Python 2.7.12 (default, Nov 19 2016, 06:48:10) 
[GCC 5.4.0 20160609] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> import numpy as np
>>> x = torch.arange(0,80).view(1,10,8)
>>> x[[0], [1, 2, 3, 4, 5, 6, 7, 8, 9], [2]]

 10
 18
 26
 34
 42
 50
 58
 66
 74
[torch.FloatTensor of size 9]

>>> y = x.cuda()
>>> y[[0], [1, 2, 3, 4, 5, 6, 7, 8, 9], [2]]
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
RuntimeError: 208 is not a device at /home/kent/Documents/misc-git-repos/pytorch/torch/lib/THC/THCGeneral.c:206
>>> y[[0], [1, 2, 3, 4, 5, 6, 7, 8, 9], [2]]
THCudaCheck FAIL file=/home/kent/Documents/misc-git-repos/pytorch/torch/lib/THC/generic/THCStorage.c line=32 error=77 : an illegal memory access was encountered
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
RuntimeError: cuda runtime error (77) : an illegal memory access was encountered at /home/kent/Documents/misc-git-repos/pytorch/torch/lib/THC/generic/THCStorage.c:32
>>>

killeent · 2017-06-14T14:04:44Z

@kentsommer Yes in theory it should work for CUDA. Let me take a look.

fmassa

I did a first pass and it looks good for the most part.
I think there is a problem with the autograd part that would be good to have fixed (in the worst case, you can have a check if the grad_output is volatile (meaning no grad of grad), and raise an error if not.
Also, for the more general case x[:, [1, 2], [2]], the linear indices might grow quite a bit, but we would need to check to see if that would have an impact or not.

torch/autograd/_functions/tensor.py

torch/csrc/generic/Tensor.cpp

killeent · 2017-06-19T22:16:17Z

@kentsommer This should now work with CUDA. The issue was that we were trying to use CPU Tensors to index into the CUDA Tensor. Now we copy the indices onto the GPU before calling the index_* functions. cc @fmassa this means that all index calculation is being done on the CPU :).

I will take a look at grad of grad tomorrow.

…for both cases (second fails)

killeent · 2017-06-22T18:01:59Z

Okay, I have now moved the advanced_index_add call into Variable. This required adding a new function to _functions/tensor.pythat handles forwards and backwards for this method, including adding a new functionadvanced_index_select` for the backward's implementation. I don't understand enough about autograd to have confidence in this implementation, but it passes the minimal tests I have added.

cc @fmassa, @gchanan

apaszke

I glanced over autograd and it looks ok to me

torch/autograd/_functions/tensor.py

+        if ctx.advanced_indexing:
+            grad_input._advanced_index_add(ctx.index, grad_output)
+        else:
+            grad_input[ctx.index] = grad_output


test/test_torch.py

+        def consec(size, start=1):
+            sequence = torch.ones(int(torch.Tensor(size).prod(0)[0])).cumsum(0)
+            sequence.add_(start - 1)
+            return sequence.resize_(*size)


torch/autograd/_functions/tensor.py

+        assert not ctx.needs_input_grad[1]
+        if ctx.needs_input_grad[2]:
+            ctx.adv_index = adv_index
+        ctx.mark_dirty(tensor1)


soumith · 2017-06-22T21:23:51Z

this is now merged into master!

vadimkantorov · 2017-06-28T15:34:56Z

Just updated my PyTorch. x[[0,4,3], [1,3,0]] from the OP works, but x[[0,4,3]] doesn't. Is it ok?

x[[0,4,3], :] works as well as x[torch.LongTensor([0,4,3])].

>>> x[[0,4,3]]
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: indexing a tensor with an object of type list. The only supported types are integers, slices, numpy scalars and torch.LongTensor or torch.ByteTensor as the only argument.
>>> x[torch.LongTensor([0,4,3])]

  0   1   2   3   4
 20  21  22  23  24
 15  16  17  18  19
[torch.FloatTensor of size 3x5]

>>> x[[0,4,3], :]

  0   1   2   3   4
 20  21  22  23  24
 15  16  17  18  19
[torch.FloatTensor of size 3x5]

soumith · 2017-06-28T16:06:09Z

yes it's a todo in the code. will be fixed in a next PR

ngimel · 2017-06-29T23:41:46Z

@killeent, I have a cuda test failing when I build from master, python test_cuda.py TestCuda.test_advancedindex. It's failing in assert_set_eq when the indexer has dupes, and I've added some printouts there:

tensor
  0   1   2   3   4
  5   6   7   8   9
 10  11  12  13  14
 15  16  17  18  19
indexer [slice(None, None, None), [0, 1, 1, 2, 2]]
val
 16  12   0  14  10
 19  17   2   7   4
  5   9   6   1  15
 13  11   3   8  18
[torch.cuda.DoubleTensor of size 4x5 (GPU 0)]

pyt 
 16  12  14   3   4
 19  17   7   8   9
  5   9   1  13  14
 13  11   8  18  19
[torch.cuda.DoubleTensor of size 4x5 (GPU 0)]
 numt 
 16   0  10   3   4
 19   2   4   8   9
  5   6  15  13  14
 13   3  18  18  19

Apparently, in the correct output (numt) the second of the duped columns should be copied to tensor from val, and in my failing case (pyt) the first one is. Can you point me to where in the backend this is happening? It's passing on most architectures, so that's not an outright bug, but failing on one.

vadimkantorov · 2017-07-17T16:22:55Z

In addition to x[[0,4,3]], torch.arange(0, 25)[[0,4,3]] doesn't work either, except the workaround torch.arange(0, 25)[[0,4,3], :] isn't possible for the lack of dimensions. (I guess this one may exist on TODO as well, but since stumbled upon it, will leave it here for completeness).

…fb74b7 Summary: Previous import was 882c5283c54345d131e8fe5c859e4844dcf7ca8e Included changes: - **[45ba661](onnx/onnx@45ba661)**: Handle new types in the switch. (pytorch#1608) <Dmitri Smirnov> - **[14853b6](onnx/onnx@14853b6)**: Bump docker image version to 230 used in CircleCI (pytorch#1606) <bddppq> - **[e0993b8](onnx/onnx@e0993b8)**: [onnxifi] Make sure that backend handles run async. (pytorch#1599) <Roman Dzhabarov> - **[e6965cc](onnx/onnx@e6965cc)**: Introduce SparseTensor ML proto (pytorch#1554) <Dmitri Smirnov> - **[75b782f](onnx/onnx@75b782f)**: In driver test check the return status of onnxGetBackendIDs (pytorch#1597) <bddppq> - **[c05b364](onnx/onnx@c05b364)**: Make CI log less verbose (pytorch#1595) <bddppq> - **[fa568e4](onnx/onnx@fa568e4)**: Loop type shape inferencing (pytorch#1591) <Scott McKay> - **[937e64c](onnx/onnx@937e64c)**: add uint8 (pytorch#1590) <Lu Fang> - **[f86e951](onnx/onnx@f86e951)**: Add domain as an optional parameter for make_node function (pytorch#1588) <Young Kim> - **[ff45588](onnx/onnx@ff45588)**: Remove unreachable code in shape_inference.h (pytorch#1585) <Changming Sun> - **[f7dcad0](onnx/onnx@f7dcad0)**: Add several hyperbolic function ops. (pytorch#1499) <Sergii Dymchenko> - **[a60ac7d](onnx/onnx@a60ac7d)**: Add OneHot op to ONNX. (pytorch#1567) <Spandan Tiwari> - **[f6c3a7e](onnx/onnx@f6c3a7e)**: [compiler flag] Issue a warning if class has virtual method but missing virtual dtor. (pytorch#1583) <Roman Dzhabarov> - **[88d1784](onnx/onnx@88d1784)**: Fix MaxUnpool shape inference when output_shape is provided as input (pytorch#1578) <Spandan Tiwari> - **[20041b7](onnx/onnx@20041b7)**: Add type shape inferencing for the If operator (pytorch#1571) <Scott McKay> - **[d6c4c75](onnx/onnx@d6c4c75)**: Add a virtual destructor to GraphInferencer (pytorch#1574) <Changming Sun> - **[a339598](onnx/onnx@a339598)**: fix ConvTranspose spec (pytorch#1566) <Wenhao Hu> Differential Revision: D13049077 fbshipit-source-id: 11133f10bc6b451094d1081e4ce736b02c8b9e2a

…002d19 Summary: Previous import was 882c5283c54345d131e8fe5c859e4844dcf7ca8e Included changes: - **[f461f7a](onnx/onnx@f461f7a)**: Show the op's type and name when the shape inference is failed. (pytorch#1623) <Jerry> - **[ab8aaf9](onnx/onnx@ab8aaf9)**: Add scan test case (pytorch#1586) <G. Ramalingam> - **[c95357e](onnx/onnx@c95357e)**: link the tutorial (pytorch#1650) <Lu Fang> - **[d7e2420](onnx/onnx@d7e2420)**: Upgrade label encoder to support more input types (pytorch#1596) <Wei-Sheng Chin> - **[6425108](onnx/onnx@6425108)**: Add Doc about Adding New Operator into ONNX (pytorch#1647) <Lu Fang> - **[295889c](onnx/onnx@295889c)**: use an empty initializer to create map (pytorch#1643) <Lu Fang> - **[e38f3ec](onnx/onnx@e38f3ec)**: Remove redundant const (pytorch#1639) <daquexian> - **[ea694bf](onnx/onnx@ea694bf)**: implement fuse reduce->unsqueeze + fix assumption in nop_dropout pass (pytorch#1565) <Armen> - **[6db386e](onnx/onnx@6db386e)**: make output shape clear enough for Softmax family (pytorch#1634) <Lu Fang> - **[2b67c6e](onnx/onnx@2b67c6e)**: fix batchnorm doc (pytorch#1633) <Lu Fang> - **[c901784](onnx/onnx@c901784)**: remove inappropriate consts (pytorch#1632) <Lu Fang> - **[de82119](onnx/onnx@de82119)**: Shape inference fix for broadcast, concat and scan (pytorch#1594) <KeDengMS> - **[d7ffe3b](onnx/onnx@d7ffe3b)**: Update Optimizer Docs (pytorch#1607) <Armen> - **[d09d139](onnx/onnx@d09d139)**: mark PROTOBUF_INCLUDE_DIRS as BUILD_INTERFACE (pytorch#1466) <Yuta Okamoto> - **[eb4b7c2](onnx/onnx@eb4b7c2)**: allow variadic parameters of different types (pytorch#1615) <G. Ramalingam> - **[4166246](onnx/onnx@4166246)**: Fix onnxifi test (pytorch#1617) <Yinghai Lu> - **[6706a4d](onnx/onnx@6706a4d)**: Fix a bug in vector address access (pytorch#1598) <Raymond Yang> - **[ae39866](onnx/onnx@ae39866)**: Separate types of inputs 1 and 2 in OneHot op. (pytorch#1610) <Spandan Tiwari> - **[45ba661](onnx/onnx@45ba661)**: Handle new types in the switch. (pytorch#1608) <Dmitri Smirnov> - **[14853b6](onnx/onnx@14853b6)**: Bump docker image version to 230 used in CircleCI (pytorch#1606) <bddppq> - **[e0993b8](onnx/onnx@e0993b8)**: [onnxifi] Make sure that backend handles run async. (pytorch#1599) <Roman Dzhabarov> - **[e6965cc](onnx/onnx@e6965cc)**: Introduce SparseTensor ML proto (pytorch#1554) <Dmitri Smirnov> - **[75b782f](onnx/onnx@75b782f)**: In driver test check the return status of onnxGetBackendIDs (pytorch#1597) <bddppq> - **[c05b364](onnx/onnx@c05b364)**: Make CI log less verbose (pytorch#1595) <bddppq> - **[fa568e4](onnx/onnx@fa568e4)**: Loop type shape inferencing (pytorch#1591) <Scott McKay> - **[937e64c](onnx/onnx@937e64c)**: add uint8 (pytorch#1590) <Lu Fang> - **[f86e951](onnx/onnx@f86e951)**: Add domain as an optional parameter for make_node function (pytorch#1588) <Young Kim> - **[ff45588](onnx/onnx@ff45588)**: Remove unreachable code in shape_inference.h (pytorch#1585) <Changming Sun> - **[f7dcad0](onnx/onnx@f7dcad0)**: Add several hyperbolic function ops. (pytorch#1499) <Sergii Dymchenko> - **[a60ac7d](onnx/onnx@a60ac7d)**: Add OneHot op to ONNX. (pytorch#1567) <Spandan Tiwari> - **[f6c3a7e](onnx/onnx@f6c3a7e)**: [compiler flag] Issue a warning if class has virtual method but missing virtual dtor. (pytorch#1583) <Roman Dzhabarov> - **[88d1784](onnx/onnx@88d1784)**: Fix MaxUnpool shape inference when output_shape is provided as input (pytorch#1578) <Spandan Tiwari> - **[20041b7](onnx/onnx@20041b7)**: Add type shape inferencing for the If operator (pytorch#1571) <Scott McKay> - **[d6c4c75](onnx/onnx@d6c4c75)**: Add a virtual destructor to GraphInferencer (pytorch#1574) <Changming Sun> - **[a339598](onnx/onnx@a339598)**: fix ConvTranspose spec (pytorch#1566) <Wenhao Hu> Differential Revision: D13263831 fbshipit-source-id: 0c158dd12c45d704b6f37f63f3d74ed34ef2f534

…002d19 (#14568) Summary: Pull Request resolved: #14568 Previous import was 882c5283c54345d131e8fe5c859e4844dcf7ca8e Included changes: - **[f461f7a](onnx/onnx@f461f7a)**: Show the op's type and name when the shape inference is failed. (#1623) <Jerry> - **[ab8aaf9](onnx/onnx@ab8aaf9)**: Add scan test case (#1586) <G. Ramalingam> - **[c95357e](onnx/onnx@c95357e)**: link the tutorial (#1650) <Lu Fang> - **[d7e2420](onnx/onnx@d7e2420)**: Upgrade label encoder to support more input types (#1596) <Wei-Sheng Chin> - **[6425108](onnx/onnx@6425108)**: Add Doc about Adding New Operator into ONNX (#1647) <Lu Fang> - **[295889c](onnx/onnx@295889c)**: use an empty initializer to create map (#1643) <Lu Fang> - **[e38f3ec](onnx/onnx@e38f3ec)**: Remove redundant const (#1639) <daquexian> - **[ea694bf](onnx/onnx@ea694bf)**: implement fuse reduce->unsqueeze + fix assumption in nop_dropout pass (#1565) <Armen> - **[6db386e](onnx/onnx@6db386e)**: make output shape clear enough for Softmax family (#1634) <Lu Fang> - **[2b67c6e](onnx/onnx@2b67c6e)**: fix batchnorm doc (#1633) <Lu Fang> - **[c901784](onnx/onnx@c901784)**: remove inappropriate consts (#1632) <Lu Fang> - **[de82119](onnx/onnx@de82119)**: Shape inference fix for broadcast, concat and scan (#1594) <KeDengMS> - **[d7ffe3b](onnx/onnx@d7ffe3b)**: Update Optimizer Docs (#1607) <Armen> - **[d09d139](onnx/onnx@d09d139)**: mark PROTOBUF_INCLUDE_DIRS as BUILD_INTERFACE (#1466) <Yuta Okamoto> - **[eb4b7c2](onnx/onnx@eb4b7c2)**: allow variadic parameters of different types (#1615) <G. Ramalingam> - **[4166246](onnx/onnx@4166246)**: Fix onnxifi test (#1617) <Yinghai Lu> - **[6706a4d](onnx/onnx@6706a4d)**: Fix a bug in vector address access (#1598) <Raymond Yang> - **[ae39866](onnx/onnx@ae39866)**: Separate types of inputs 1 and 2 in OneHot op. (#1610) <Spandan Tiwari> - **[45ba661](onnx/onnx@45ba661)**: Handle new types in the switch. (#1608) <Dmitri Smirnov> - **[14853b6](onnx/onnx@14853b6)**: Bump docker image version to 230 used in CircleCI (#1606) <bddppq> - **[e0993b8](onnx/onnx@e0993b8)**: [onnxifi] Make sure that backend handles run async. (#1599) <Roman Dzhabarov> - **[e6965cc](onnx/onnx@e6965cc)**: Introduce SparseTensor ML proto (#1554) <Dmitri Smirnov> - **[75b782f](onnx/onnx@75b782f)**: In driver test check the return status of onnxGetBackendIDs (#1597) <bddppq> - **[c05b364](onnx/onnx@c05b364)**: Make CI log less verbose (#1595) <bddppq> - **[fa568e4](onnx/onnx@fa568e4)**: Loop type shape inferencing (#1591) <Scott McKay> - **[937e64c](onnx/onnx@937e64c)**: add uint8 (#1590) <Lu Fang> - **[f86e951](onnx/onnx@f86e951)**: Add domain as an optional parameter for make_node function (#1588) <Young Kim> - **[ff45588](onnx/onnx@ff45588)**: Remove unreachable code in shape_inference.h (#1585) <Changming Sun> - **[f7dcad0](onnx/onnx@f7dcad0)**: Add several hyperbolic function ops. (#1499) <Sergii Dymchenko> - **[a60ac7d](onnx/onnx@a60ac7d)**: Add OneHot op to ONNX. (#1567) <Spandan Tiwari> - **[f6c3a7e](onnx/onnx@f6c3a7e)**: [compiler flag] Issue a warning if class has virtual method but missing virtual dtor. (#1583) <Roman Dzhabarov> - **[88d1784](onnx/onnx@88d1784)**: Fix MaxUnpool shape inference when output_shape is provided as input (#1578) <Spandan Tiwari> - **[20041b7](onnx/onnx@20041b7)**: Add type shape inferencing for the If operator (#1571) <Scott McKay> - **[d6c4c75](onnx/onnx@d6c4c75)**: Add a virtual destructor to GraphInferencer (#1574) <Changming Sun> - **[a339598](onnx/onnx@a339598)**: fix ConvTranspose spec (#1566) <Wenhao Hu> Reviewed By: zrphercule Differential Revision: D13263831 fbshipit-source-id: a2ff22c6454e2430429e5a7d18d21661a7ffb0cb

@kevinstephano

…er (#1588) Fixes #75622 1. Instead of getting max/min_value for reduction init value, we go with (-)infinity instead so we can properly preserve inf inputs; 2. Adding inf/(-)inf/nan for float value. 3. Adding aten::amin in nvfuser (@kevinstephano @rdspring1 for review) Pull Request resolved: #75646 Approved by: https://github.com/rdspring1, https://github.com/kevinstephano, https://github.com/ngimel

@carlobertolli

pytorch#1588) …d in reduce config (pytorch#135397) Fixes pytorch#132964 This change is to optimize torch.sum() performance by increasing max_values_per_thread in setReduceConfig() for ROCm platform. By increasing this parameter, it uses fewer threadblocks and improved the performance. Test: Tested on MI300x and H100, and now the MI300x perf improved to 3205GByte/s from ~1690GByte/s for the test case and is slightly better than H100 (3136GByte/s). Also tested with other different sizes of tensors and also see perf improvement. ```python import torch from triton.testing import do_bench x = torch.randn(2**30, device='cuda') ms = do_bench(lambda: x.sum(dim=-1)) bandwidth_gbyte = x.numel() * x.dtype.itemsize / (10**9) time_s = ms / 1000 bw_per_second = bandwidth_gbyte / time_s print(bw_per_second) ``` Co-author: @carlobertolli Pull Request resolved: pytorch#135397 Approved by: https://github.com/eqy, https://github.com/malfet Fixes #ISSUE_NUMBER Co-authored-by: hongxyan <hongxyan@amd.com>

@carlobertolli

pytorch#1588) …d in reduce config (pytorch#135397) Fixes pytorch#132964 This change is to optimize torch.sum() performance by increasing max_values_per_thread in setReduceConfig() for ROCm platform. By increasing this parameter, it uses fewer threadblocks and improved the performance. Test: Tested on MI300x and H100, and now the MI300x perf improved to 3205GByte/s from ~1690GByte/s for the test case and is slightly better than H100 (3136GByte/s). Also tested with other different sizes of tensors and also see perf improvement. ```python import torch from triton.testing import do_bench x = torch.randn(2**30, device='cuda') ms = do_bench(lambda: x.sum(dim=-1)) bandwidth_gbyte = x.numel() * x.dtype.itemsize / (10**9) time_s = ms / 1000 bw_per_second = bandwidth_gbyte / time_s print(bw_per_second) ``` Co-author: @carlobertolli Pull Request resolved: pytorch#135397 Approved by: https://github.com/eqy, https://github.com/malfet Fixes #ISSUE_NUMBER Co-authored-by: hongxyan <hongxyan@amd.com> (cherry picked from commit 4360582)

fmassa reviewed May 22, 2017

View reviewed changes

torch/csrc/generic/Tensor.cpp Outdated

This comment was marked as off-topic.

Sign in to view

This comment was marked as off-topic.

Sign in to view

killeent force-pushed the advanced-indexing branch from ccfbaa6 to 9888e9a Compare May 23, 2017 15:46

killeent force-pushed the advanced-indexing branch from 9888e9a to 0518639 Compare June 12, 2017 16:32

killeent changed the title ~~WIP: a subset of advanced indexing~~ Advanced Indexing Part 1 -- Purely Integer Array Indexing Jun 12, 2017

fmassa reviewed Jun 12, 2017

View reviewed changes

torch/csrc/generic/Tensor.cpp Outdated

This comment was marked as off-topic.

Sign in to view

fmassa reviewed Jun 12, 2017

View reviewed changes

torch/csrc/generic/Tensor.cpp Outdated

This comment was marked as off-topic.

Sign in to view

fmassa requested changes Jun 14, 2017

View reviewed changes

killeent force-pushed the advanced-indexing branch from 4cfa35e to a808771 Compare June 22, 2017 17:20

killeent added 18 commits June 22, 2017 10:57

temp commit advanced indexing

15aaf85

basic advanced indexing with ndim 1d long tensors

1f71cff

support arbitrary indexing shapes for output

2320b68

arbitrary sequence indexers

cbed223

use strides directly for linear index calculation

c58b9f9

tests in passing state minus broadcasting

5ee065e

clean up checkAdvancedIndexing

8a8cf12

clean up code, add memory frees/decrefs

16f570b

support empty indexer

c561c5f

hacky commit broadcast semantics

4b43552

support transposed tensors

ade7d24

move advanced indexing into getvalue

169ab74

refactor out argument conversion code into separate method

6a87fcf

refactor out linear index calculation

906429e

advanced index set + unit tests

2721ba4

fix minor mem leak

493fdd5

check available at interpreter level

3d1d78d

get advanced indexing autograd working for non duplicates, unit test …

dc212af

…for both cases (second fails)

killeent added 4 commits June 22, 2017 10:59

call newcontiguous where appropriate

09d4180

index bounds checking

e325238

make work with cuda

eb912b6

make adv indexing work with variable, gradgrad

7794c92

killeent force-pushed the advanced-indexing branch from a808771 to 7794c92 Compare June 22, 2017 17:59

apaszke reviewed Jun 22, 2017

View reviewed changes

killeent added 2 commits June 22, 2017 12:17

use .view() instead of .resize() in indexing test

6e5d26c

smaller tensors for advancedindex_big, run in cuda as well

99ce824

soumith reviewed Jun 22, 2017

View reviewed changes

torch/autograd/_functions/tensor.py

assert not ctx.needs_input_grad[1]

if ctx.needs_input_grad[2]:

ctx.adv_index = adv_index

ctx.mark_dirty(tensor1)

This comment was marked as off-topic.

Sign in to view

This comment was marked as off-topic.

Sign in to view

soumith approved these changes Jun 22, 2017

View reviewed changes

soumith closed this Jun 22, 2017

killeent mentioned this pull request Jun 23, 2017

[READY FOR REVIEW] Advanced Indexing 2A - Colons + Adjacent Adv Indexers #1890

Merged

houseroad mentioned this pull request Nov 14, 2018

Automatic update of fbcode/onnx to 45ba661d278bc13d504829bfa6bfb48f15fb74b7 #13955

Closed

houseroad mentioned this pull request Nov 29, 2018

Automatic update of fbcode/onnx to f461f7aad9987635b4aff108620ed7918f002d19 #14568

Closed

ezyang added the open source label Jun 24, 2019

Conversation

killeent commented May 18, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment was marked as off-topic.

Uh oh!

This comment was marked as off-topic.

Uh oh!

killeent commented Jun 12, 2017

Uh oh!

This comment was marked as off-topic.

Uh oh!

This comment was marked as off-topic.

Uh oh!

kentsommer commented Jun 14, 2017

Uh oh!

killeent commented Jun 14, 2017

Uh oh!

fmassa left a comment

Choose a reason for hiding this comment

Uh oh!

This comment was marked as off-topic.

Uh oh!

This comment was marked as off-topic.

Uh oh!

This comment was marked as off-topic.

Uh oh!

This comment was marked as off-topic.

Uh oh!

killeent commented Jun 19, 2017

Uh oh!

killeent commented Jun 22, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

apaszke left a comment

Choose a reason for hiding this comment

Uh oh!

This comment was marked as off-topic.

Uh oh!

This comment was marked as off-topic.

Uh oh!

This comment was marked as off-topic.

Uh oh!

This comment was marked as off-topic.

Uh oh!

This comment was marked as off-topic.

Uh oh!

This comment was marked as off-topic.

Uh oh!

This comment was marked as off-topic.

Uh oh!

soumith commented Jun 22, 2017

Uh oh!

vadimkantorov commented Jun 28, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

soumith commented Jun 28, 2017

Uh oh!

ngimel commented Jun 29, 2017

Uh oh!

vadimkantorov commented Jul 17, 2017

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

8 participants

killeent commented May 18, 2017 •

edited

Loading

killeent commented Jun 22, 2017 •

edited

Loading

vadimkantorov commented Jun 28, 2017 •

edited

Loading