Compare commits

..

80 Commits

Author SHA1 Message Date
8aa878fc93 [jit] fix tuple alias analysis (#41992)
Previously when analyzing a TupleConstruct, we ignored the aliasing
information of the inputs and simply marked all elements of the returned
tuple as wildcards. But since we can fully reason about the contents of
a tuple statically, we should be able to assign them aliasing
information.

This analysis was not only incomplete but produced incorrect results,
since if `a` is not a wildcard, `a noalias wilcard`. So if we looked at
`tuple(a)` and reported the aliasing info as `tuple(wildcard)`, then
`tuple[0] noalias a`, which is...wrong.
2020-07-24 08:05:20 -07:00
7c7c9c3aa6 scatter/gather - check that inputs are of the same dimensionality (#41890)
Co-authored-by: Nikita Vedeneev <nik@quansight.com>
2020-07-22 18:33:07 -07:00
a2922f589d [1.6.0] Mark torch.set_deterministic and torch.is_deterministic as experimental (#41870)
This PR:
- renames `torch.set_deterministic` to `torch._set_deterministic`
- renames `torch.is_deterministic` to `torch._is_deterministic`
- Modifies the docstrings for both to indicate that the feature is not
yet complete.

We would like to do this because this feature is experimental and the
docstrings before this PR are misleading.

This PR does not have an accompanying change in master. That is because
there still is discussion over what the eventual state of the feature
should be: https://github.com/pytorch/pytorch/issues/15359. I expect
that there will be a better plan for this once 1.7 rolls around.

Test Plan:
- wait for CI
2020-07-22 18:32:47 -07:00
8acfecaecb [1.6] Add optimizer_for_mobile doc into python api root doc (#41491)
* Add optimizer_for_mobile doc into python api root doc

* Apply suggestions from code review

Remove all references to `optimization_blacklist` as it's missing in 1.6

Co-authored-by: Nikita Shulga <nshulga@fb.com>
2020-07-22 17:37:45 -07:00
860e18a61b Update torch.set_default_dtype doc (#41263)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/41263

Test Plan: Imported from OSS

Differential Revision: D22482989

Pulled By: anjali411

fbshipit-source-id: 2aadfbb84bbab66f3111970734a37ba74d817ffd
2020-07-22 14:50:15 -07:00
8f804baaa9 Doc note for complex (#41252)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/41252

Test Plan: Imported from OSS

Reviewed By: albanD

Differential Revision: D22553266

Pulled By: anjali411

fbshipit-source-id: f6dc409da048496d72b29b0976dfd3dd6645bc4d
2020-07-22 14:49:51 -07:00
a395e0903e Autograd Doc for Complex Numbers (#41012)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/41012

Test Plan: Imported from OSS

Differential Revision: D22476911

Pulled By: anjali411

fbshipit-source-id: 7da20cb4312a0465272bebe053520d9911475828
2020-07-22 14:40:52 -07:00
2ca55430d2 Add reference documentation for torch/library.h (#41470) (#41602)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/41470

Signed-off-by: Edward Z. Yang <ezyang@fb.com>

Test Plan: Imported from OSS

Reviewed By: zou3519

Differential Revision: D22577426

Pulled By: ezyang

fbshipit-source-id: 4bfe5806061e74181a74d161c868acb7c1ecd1e4
2020-07-22 11:10:16 -07:00
b8e77a42bd Add CUDA11 build and test (#40452) (#41543)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/40452

Differential Revision: D22316007

Pulled By: malfet

fbshipit-source-id: 94f4b4ba2a46ff3d3042ba842a615f8392cdc350

Co-authored-by: Gao, Xiang <qasdfgtyuiop@gmail.com>
2020-07-22 09:53:22 -07:00
4081fdd3df Revert "port masked_select from TH to ATen and optimize perf on CPU (#33269)" (#41829)
This reverts commit fe66bdb498efe912d8b9c437a14efa4295c04fdd.

This also makes a sense to THTensorEvenMoreMath because sumall was removed, see THTensor_wrap.
2020-07-22 09:52:30 -07:00
cefb9e0cd6 Update pthreadpool to pthreadpool:029c88620802e1361ccf41d1970bd5b07fd6b7bb. (#40524) (#41190)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/40524

Reviewed By: ezyang

Differential Revision: D22215742

Pulled By: AshkanAliabadi

fbshipit-source-id: ef594e0901337a92b21ddd44e554da66c723eb7c
2020-07-10 09:11:32 -07:00
d9e9e0087a [v1.6] [RPC docs] Remove mention of TensorPipe's SHM and CMA backends as they're not built (#41229)
Summary:
In short, we messed up. The SHM and CMA backends of TensorPipe are Linux-specific and thus they are guarded by a #ifdef in the agent's code. Due to a mishap with CMake (due the fact that TensorPipe has two CMake files, one for PyTorch and a "standalone" one) we were not correctly propagating some flags and these #ifdefs were always false. This means that these two backends have always been disabled and have thus never been covered by our OSS CI. It would be irresponsible to enable them now in v1.6, so instead we remove any mention of them from the docs.

Note that this is perhaps not as bad as it sounds. These two backends were providing higher performance (latency) when the two endpoints were on the same machine. However, I suspect that most RPC users will only do transfers across machines, for which SHM and CMA wouldn't have played any role.

Original PR against master: #41200 (merged as dde3d5f4a8f713ecc4649d776565b68ca75ae5c8)

Test Plan: Docs only
2020-07-10 09:02:08 -07:00
43d746305c Preserve CUDA gencode flags (#41212)
Summary:
Add `torch._C._cuda_getArchFlags()` that returns list of architecture `torch_cuda` were compiled with
Add `torch.cuda.get_arch_list()` and `torch.cuda.get_gencode_flags()` methods that returns architecture list and gencode flags PyTorch were compiled with
Print warning if some of GPUs is not compatible with any of the CUBINs

Pull Request resolved: https://github.com/pytorch/pytorch/pull/41173

Differential Revision: D22459998

Pulled By: malfet

fbshipit-source-id: 65d40ae29e54a0ba0f3f2da11b821fdb4d452d95
2020-07-09 17:34:50 -07:00
9409e03903 [ONNX][1.6] Update interpolate recompute_scale_factor default (#41117)
* Update interpolate recompute_scale_factor default

* Update upsampling.h

* Update functional.py
2020-07-09 17:24:53 -07:00
c9a1853d2f [1.6] Make IterableDataset DataLoader.__len__ warning clearer (#41185)
* make IterableDataset DataLoader.__len__ warning clearer

* typo
2020-07-09 14:07:58 -07:00
7fa9b2923b quantizer.cpp: fix cuda memory pinning (#41139) (#41194)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/41139

Fixes the test case in https://github.com/pytorch/pytorch/issues/41115
by using PyTorch's CUDA allocator instead of the old Caffe2 one.

Test Plan:
run the test case from the issue:
https://gist.github.com/vkuzo/6d013aa1645cb986d0d4464a931c779b

let's run CI and see what it uncovers

Imported from OSS

Reviewed By: malfet

Differential Revision: D22438787

fbshipit-source-id: 0853b0115d198a99c43e6176aef34ea951bf5c2e

Co-authored-by: Vasiliy Kuznetsov <vasiliy@fb.com>
2020-07-09 14:06:11 -07:00
40bf15a8ac Remove copy_ warnings for angle and abs for complex tensors (#41152) (#41191)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/41152

fixes https://github.com/pytorch/pytorch/issues/40838

Test Plan: Imported from OSS

Differential Revision: D22444357

Pulled By: anjali411

fbshipit-source-id: 2879d0cffc0a011c624eb8e00c7b64bd33522cc3

Co-authored-by: anjali411 <chourdiaanjali123@gmail.com>
2020-07-09 13:41:15 -07:00
c164fc4d7f Patch #40883 to 1.6 release. (#41033) 2020-07-09 10:25:39 -07:00
e0b7480f34 Revert "make IterableDataset DataLoader.__len__ warning clearer (#41183)"
This reverts commit 89d7f194d8ea19f36c9afb52585a00b5b7d0ffeb.
2020-07-09 08:05:24 -07:00
89d7f194d8 make IterableDataset DataLoader.__len__ warning clearer (#41183) 2020-07-09 08:00:00 -07:00
59bb44a8e8 Add a link in RPC doc page to point to PT Distributed overview (#41108) (#41156)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/41108

Test Plan: Imported from OSS

Differential Revision: D22440751

Pulled By: mrshenli

fbshipit-source-id: 9e7b002091a3161ae385fdfcc26484ae8fc243bb
2020-07-09 07:49:10 -07:00
8f4d01d9f1 Disables unary op casting to output dtype (#41097) (#41160)
Summary:
Fixes https://github.com/pytorch/pytorch/issues/41047.

Some CPU kernel implementations don't call `cast_outputs()`, so when CPU temporaries were created to hold their outputs they weren't copied back to the out parameters correctly. Instead of fixing that issue, for simplicity this PR disables the behavior. The corresponding test in test_type_promotion.py is expanded with more operations to verify that unary ops can no longer have out arguments with different dtypes than their inputs (except in special cases like torch.abs which maps complex inputs to float outputs and torch.deg2rad which is secretly torch.mul).

Pull Request resolved: https://github.com/pytorch/pytorch/pull/41097

Differential Revision: D22422352

Pulled By: mruberry

fbshipit-source-id: 8e61d34ef1c9608790b35cf035302fd226fd9421

Co-authored-by: Mike Ruberry <mruberry@devfair044.maas>
2020-07-08 22:06:48 -07:00
77ffb25925 Add guard for non-default stream in DDP's autograd engine callback (#40115) (#41151)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40115

Closes https://github.com/pytorch/pytorch/issues/37790
Closes https://github.com/pytorch/pytorch/issues/37944

A user may wish to run DDP's forward + backwards step under a non-default CUDA stream such as those created by `with torch.cuda.Stream(stream)`. In this case, the user should be responsible for synchronizing events on this stream with other streams used in the program (per the documentation at https://pytorch.org/docs/stable/notes/cuda.html#cuda-semantics), but currently DDP has a bug which causes DDP under non-default streams to fail.

If a user does the following:
```
model = DDP(...)
loss = model(inptut).sum()
loss.backward()
grad = model.module.weight.grad()
average = dist.all_reduce(grad)
```

There is a chance that `average` and `grad` will not be equal. This is because the CUDA kernels corresponding to the  `all_reduce` call may run before `loss.backward()`'s kernels are finished. Specifically, in DDP we copy the allreduced gradients back to the model parameter gradients in an autograd engine callback, but this callback runs on the default stream. Note that this can also be fixed by the application synchronizing on the current stream, although this should not be expected, since the application is not using the current stream at all.

This PR fixes the issue by passing the current stream into DDP's callback.

Tested by adding a UT `test_DistributedDataParallel_non_default_stream` that fails without this PR
ghstack-source-id: 106481208

Differential Revision: D22073353

fbshipit-source-id: 70da9b44e5f546ff8b6d8c42022ecc846dff033e
2020-07-08 21:08:17 -07:00
af9600b1f5 [Caffe2] Move in-header virtual function implementation to .cc files (#41090)
* Move OperatorSchema default inference function implementations to .cc… (#40845)

Summary:
… file

This prevents implementation of those functions(as lambdas) to be embedded as weak symbol into every shared library that includes this header.

Combination of this and https://github.com/pytorch/pytorch/pull/40844 reduces size of `libcaffe2_module_test_dynamic.so` from 500kb to 50Kb.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40845

Differential Revision: D22334779

Pulled By: malfet

fbshipit-source-id: 64706918fc2947350a58c0877f294b1b8b085455

* Move `OperatorBase::AddRelatedBlobInfo` implementation to .cc file (#40844)

Summary:
If virtual function is implemented in header file, it's implementation will be included as a weak symbol to every shared library that includes this header along with all of it's dependencies.

This was one of the reasons why size of libcaffe2_module_test_dynamic.so  was 500Kb (AddRelatedBlobInfo implementation pulled a quarter of libprotobuf.a with it)

Combination of this and https://github.com/pytorch/pytorch/issues/40845 reduces size of `libcaffe2_module_test_dynamic.so` from 500kb to 50Kb.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40844

Differential Revision: D22334725

Pulled By: malfet

fbshipit-source-id: 836a4cbb9f344355ddd2512667e77472546616c0
2020-07-07 21:17:11 -07:00
83262b1ba1 torch._six.PY37 should be true for Python-3.8 as well (#40868) (#41091)
Summary:
Right now it is used to check whether `math.remainder` exists, which is the case for both Python-3.7 and 3.8
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40868

Differential Revision: D22343454

Pulled By: malfet

fbshipit-source-id: 6b6d4869705b64c4b952309120f92c04ac7e39fd
2020-07-07 17:15:01 -07:00
f862a6ba4d Remove unused Logger in get_matching_activations (#41023) (#41087)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/41023

Remove Logger in get_matching_activations since it's not used.
ghstack-source-id: 107237046

Test Plan:
buck test mode/dev caffe2/test:quantization -- 'test_compare_weights_lstm_dynamic'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_stub_lstm_dynamic'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_outputs_lstm_dynamic'
buck test mode/dev caffe2/test:quantization -- 'test_compare_weights_conv_static'
buck test mode/dev caffe2/test:quantization -- 'test_compare_weights_linear_static'
buck test mode/dev caffe2/test:quantization -- 'test_compare_weights_linear_dynamic'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_stub_conv_static'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_stub_linear_static'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_stub_submodule_static'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_stub_functional_static'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_stub_linear_dynamic'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_outputs_conv_static'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_outputs_linear_static'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_outputs_functional_static'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_outputs_linear_dynamic'

Differential Revision: D22394957

fbshipit-source-id: 7d59e0f35e9f4c304b8487460d48236ee6e5a872

Co-authored-by: Haixin Liu <haixin@fb.com>
2020-07-07 16:09:37 -07:00
f3c1ea7455 [PyTorch Numeric Suite] Remove unnecessary Logger in input arguments (#40890) (#41086)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40890

Remove unnecessary Logger in input arguments and simplify the API.
ghstack-source-id: 107110487

Test Plan:
buck test mode/dev caffe2/test:quantization -- 'test_compare_weights_lstm_dynamic'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_stub_lstm_dynamic'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_outputs_lstm_dynamic'
buck test mode/dev caffe2/test:quantization -- 'test_compare_weights_conv_static'
buck test mode/dev caffe2/test:quantization -- 'test_compare_weights_linear_static'
buck test mode/dev caffe2/test:quantization -- 'test_compare_weights_linear_dynamic'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_stub_conv_static'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_stub_linear_static'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_stub_submodule_static'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_stub_functional_static'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_stub_linear_dynamic'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_outputs_conv_static'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_outputs_linear_static'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_outputs_functional_static'
buck test mode/dev caffe2/test:quantization -- 'test_compare_model_outputs_linear_dynamic'

Differential Revision: D22345477

fbshipit-source-id: d8b4eb3d6cb3049aa3296dead8ba29bf5467bd1c

Co-authored-by: Haixin Liu <haixin@fb.com>
2020-07-07 16:09:11 -07:00
2ed3ad2891 fix autodoc for torch.distributed.launch (#40963) (#41089)
Summary:
The doc for `torch.distributed.launch` is missing since v1.2.0 (see issue https://github.com/pytorch/pytorch/issues/36386) because PR https://github.com/pytorch/pytorch/issues/22501 added some imports at the first line.
542ac74987/torch/distributed/launch.py (L1-L5)
I move it below the docstring to make the autodoc in Sphinx work normally.

Pull Request resolved: https://github.com/pytorch/pytorch/pull/40963

Differential Revision: D22380816

Pulled By: mrshenli

fbshipit-source-id: ee8406785b9a198bbf3fc65e589854379179496f

Co-authored-by: Xin Yao <yaox12@outlook.com>
2020-07-07 14:23:31 -07:00
a857af50a4 [quant][graphmode][fix] cloning schema in insert_observers (#40624) (#40934)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40624

Previously we didn't clone schema, so the default schema is used, this is
causing issue for some models

Test Plan: Imported from OSS

Differential Revision: D22259519

fbshipit-source-id: e2a393a54cb18f55da0c7152a74ddc22079ac350
2020-07-07 13:27:36 -07:00
d0045e5520 Some fixes for graph mode quantization (#40935)
* [quant] aten::repeat work for quantized tensor (#40644)

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/40644

Test Plan: Imported from OSS

Differential Revision: D22268558

fbshipit-source-id: 3bc9a129bece1b547c519772ecc6b980780fb904

* [quant][graphmode][fix] remove unsupported ops in the list (#40653)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40653

(Note: this ignores all push blocking failures!)

Test Plan: Imported from OSS

Differential Revision: D22271413

fbshipit-source-id: a01611b5d90849ac673fa5a310f910c858e907a3
2020-07-07 13:26:27 -07:00
0406b69b79 [quant][graphmode][fix] Fold conv bn (#40865) (#40970)
* [quant][graphmode][fix] Fold conv bn (#40865)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40865

1. applied filter for the module types
2. removed the assumption that the conv bn are immediate child of parent module

Test Plan:
python test/test_quantization.py TestQuantizeJitPasses

Imported from OSS

Differential Revision: D22338074

fbshipit-source-id: 64739a5e56c0a74249a1dbc2c8454b88ec32aa9e

* [quant][graphmode][fix] Print the node in error message (#40889)

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/40889

Test Plan: Imported from OSS

Differential Revision: D22348266

fbshipit-source-id: eed2ece5c94fcfaf187d6770bed4a7109f0c0b4a
2020-07-07 13:25:39 -07:00
6220cc4380 [quant][graphmode][fix] dequantize propagation for {add/mul}_scalar + aten::repeat (#40933)
* [quant][graphmode][fix] dequantize propagation for {add/mul}_scalar (#40596)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40596

Previously the fusion patterns for {add/mul}_scalar is inconsistent since the op pattern
produces a non-quantized tensor and the op replacement graph produces a quantized tensor

Test Plan: Imported from OSS

Differential Revision: D22251072

fbshipit-source-id: e16eb92cf6611578cca1ed8ebde961f8d0610137

* [quant][graphmode] Support quantization for `aten::apend` (#40743)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40743

`aten::append` modifies input inplace and the output is ignored, these ops are not
supported right now, so we'll need to first make `aten::append` non-inplace
by change
```
ignored = aten::append(list, x)
```
to
```
x_list = aten::ListConstruct(x)
result = aten::add(list, x_list)
```
and then quantize the aten::add instead.

Test Plan:
TestQuantizeJitOps.test_general_shape_ops

Imported from OSS

Differential Revision: D22302151

fbshipit-source-id: 931000388e7501e9dd17bec2fad8a96b71a5efc5
2020-07-07 13:25:02 -07:00
eaf3f2fd34 Added index_put to promotelist (#41036)
* Added index_put to promotelist

* docstring

Co-authored-by: Michael Carilli <mcarilli@nvidia.com>
2020-07-07 13:00:32 -07:00
c35b4c770b Bucket of shape analysis fixes (#41044)
* [JIT] fix unfold shape analysis (#40749)

Summary:
unfold on a 0-dimensioned tensor returns a 1-dim tensor
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40749

Differential Revision: D22361481

Pulled By: eellison

fbshipit-source-id: 621597e5f97f6e39953eb86f8b85bb4142527a9f

* shape analysis fix for default dtype'

ghstack-source-id: 723aa27c2685417715a0891f5ca1ae885d4c9832
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40938

* fix grad thrashing of shape analysis

ghstack-source-id: dd8742b1da52d17e9d6ab6c81ff0b27520b09417
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40939

Co-authored-by: Elias Ellison <eellison@fb.com>
2020-07-07 12:59:47 -07:00
11baccf1b5 [release/1.6] .circleci: Output binary sizes, store binaries (#41075)
We need an easy to way to quickly visually grep binary sizes from builds
and then have a way to test out those binaries quickly.

Signed-off-by: Eli Uriegas <eliuriegas@fb.com>
(cherry picked from commit 66813515d4dec66f319442ba967c64b87c0286cd)
Signed-off-by: Eli Uriegas <eliuriegas@fb.com>
2020-07-07 11:27:00 -07:00
f0f0cbdd4a Docstring changes for dynamic quantized classes (#40931) (#41032)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40931

Fix docstrings for dynamic quantized Linear/LSTM and associated classes
ghstack-source-id: 107064446

Test Plan: Docs show up in correctly

Differential Revision: D22360787

fbshipit-source-id: 8e357e081dc59ee42fd7f12ea5079ce5d0cc9df2
2020-07-06 21:37:53 -07:00
11b70b0041 [JIT] Switch executor from Simple to Legacy. (#41017)
* properly skip legacy tests regardless of the default executor (#40381)

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/40381

Differential Revision: D22173938

Pulled By: Krovatkin

fbshipit-source-id: 305fc4484977e828cc4cee6e053a1e1ab9f0d6c7

* [JIT] Switch executor from Simple to Legacy.

This is done for 1.6 only in order to recover performance regressions
caused by the Legacy->Simple switch that was done in 1.5. On master we
still plan to use Simple executor and fix the performance issues in 1.7
without falling back to the Legacy executor.

Co-authored-by: Nikolay Korovaiko <korovaikon@gmail.com>
2020-07-06 21:35:02 -07:00
01e9562313 [1.6 cherrypick] Fix delegating to jit.load from torch.load (#41013) 2020-07-06 16:55:00 -07:00
3f13c9a2c8 infer tensor properties based on an input tensor rather than defaults for xxx_like ctors (#40895) (#41016)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/40895

Reviewed By: eellison

Differential Revision: D22358878

Pulled By: Krovatkin

fbshipit-source-id: 2db2429aa89c180d8e52a6bb1265308483da46a2
2020-07-06 16:52:59 -07:00
63a94c021a shape inference of undefined for prim::grad (#40866) (#41015)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/40866

Reviewed By: pbelevich

Differential Revision: D22358988

Pulled By: Krovatkin

fbshipit-source-id: 7118d7f8d4eaf056cfb71dc0d588d38b1dfb0fc7
2020-07-06 16:51:37 -07:00
2b175ba909 update requires_gard on loop inputs correctly (master) (#40926) (#41014)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/40926

Reviewed By: eellison

Differential Revision: D22359471

Pulled By: Krovatkin

fbshipit-source-id: 823e87674e2d2917f075255ec926e0485972f4e2
2020-07-06 16:30:14 -07:00
8c3f662224 Update FP16 to FP16:4dfe081cf6bcd15db339cf2680b9281b8451eeb3. (#40956) 2020-07-06 06:59:41 -07:00
0ffdd5aa1d Update cpuinfo to cpuinfo:63b254577ed77a8004a9be6ac707f3dccc4e1fd9. (#40955) 2020-07-06 06:59:30 -07:00
d53427c541 Update FXdiv to FXdiv:b408327ac2a15ec3e43352421954f5b1967701d1. (#40954) 2020-07-06 06:59:17 -07:00
b44b1d868e Update psimd to psimd:072586a71b55b7f8c584153d223e95687148a900 (#40953) 2020-07-06 06:59:01 -07:00
9184c9832e Re-apply PyTorch pthreadpool changes (#40951)
* Re-apply PyTorch pthreadpool changes

Summary:
This re-applies D21232894 (b9d3869df3) and D22162524, plus updates jni_deps in a few places
to avoid breaking host JNI tests.

Test Plan: `buck test @//fbandroid/mode/server //fbandroid/instrumentation_tests/com/facebook/caffe2:host-test`

Reviewed By: xcheng16

Differential Revision: D22199952

fbshipit-source-id: df13eef39c01738637ae8cf7f581d6ccc88d37d5

* Enable XNNPACK ops on iOS and macOS.

Test Plan: buck run aibench:run_bench -- -b aibench/specifications/models/pytorch/pytext/pytext_mobile_inference.json --platform ios --framework pytorch --remote --devices D221 (9788a74da8)AP-12.0.1

Reviewed By: xta0

Differential Revision: D21886736

fbshipit-source-id: ac482619dc1b41a110a3c4c79cc0339e5555edeb

* Respect user set thread count. (#40707)

Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/40707

Test Plan: Imported from OSS

Differential Revision: D22318197

Pulled By: AshkanAliabadi

fbshipit-source-id: f11b7302a6e91d11d750df100d2a3d8d96b5d1db

* Fix and reenable threaded QNNPACK linear (#40587)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40587

Previously, this was causing divide-by-zero only in the multithreaded
empty-batch case, while calculating tiling parameters for the threads.
In my opinion, the bug here is using a value that is allowed to be zero
(batch size) for an argument that should not be zero (tile size), so I
fixed the bug by bailing out right before the call to
pthreadpool_compute_4d_tiled.

Test Plan: TestQuantizedOps.test_empty_batch

Differential Revision: D22264414

Pulled By: dreiss

fbshipit-source-id: 9446d5231ff65ef19003686f3989e62f04cf18c9

* Fix batch size zero for QNNPACK linear_dynamic (#40588)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40588

Two bugs were preventing this from working.  One was a divide by zero
when multithreading was enabled, fixed similarly to the fix for static
quantized linear in the previous commit.  The other was computation of
min and max to determine qparams.  FBGEMM uses [0,0] for [min,max] of
empty input, do the same.

Test Plan: Added a unit test.

Differential Revision: D22264415

Pulled By: dreiss

fbshipit-source-id: 6ca9cf48107dd998ef4834e5540279a8826bc754

Co-authored-by: David Reiss <dreiss@fb.com>
2020-07-06 06:58:25 -07:00
e89c4f0dec [quant] Fix fuse linear pass (#40549) (#40751)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40549

Currently we didn't check if %weight_t is produced by `aten::t`, this will fuse some `matmul`/`addmm` that is
not 2d to `aten::linear`, which is incorrect

Test Plan: Imported from OSS

Differential Revision: D22225921

fbshipit-source-id: 9723e82fdbac6d8e1a7ade22f3a9791321ab12b6
2020-07-02 10:23:22 -07:00
ea273c68f9 Inplace construct of TorchScript Module and inplace option for quantization (#40750)
* [WIP][JIT] Add ScriptModule._reconstruct (#39979)

Summary:
**Summary**
This commit adds an instance method `_reconstruct` that permits users
to reconstruct a `ScriptModule` from a given C++ `Module` instance.

**Testing**
This commit adds a unit test for `_reconstruct`.

**Fixes**
This pull request fixes https://github.com/pytorch/pytorch/issues/33912.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/39979

Differential Revision: D22172323

Pulled By: SplitInfinity

fbshipit-source-id: 9aa6551c422a5a324b822a09cd8d7c660f99ca5c

* [quant][graphmode] Enable inplace option for top level API (#40414)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40414

after `_reconstruct` is supported in RecursiveScriptModule: https://github.com/pytorch/pytorch/pull/39979
we can support inplace option in quantization API

Test Plan: Imported from OSS

Differential Revision: D22178326

fbshipit-source-id: c78bc2bcf2c42b06280c12262bb31aebcadc6c32

Co-authored-by: Meghan Lele <meghanl@fb.com>
2020-07-02 10:22:45 -07:00
4dd37bfbf7 [jit] Remove unnecessary clone APIs for script::Module and RecursiveScriptModule (#40297) (#40748)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/40297

Test Plan: Imported from OSS

Differential Revision: D22191660

fbshipit-source-id: 4b338ca82caaca04784bffe01fdae3d180c192f4
2020-07-02 10:22:27 -07:00
2533b9da83 Fix complex printing for sci_mode=True (#40513) (#40919)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40513

This PR makes the following changes:
1. Complex Printing now uses print formatting for it's real and imaginary values and they are joined at the end.
2. Adding 1. naturally fixes the printing of complex tensors in sci_mode=True

```
>>> torch.tensor(float('inf')+float('inf')*1j)
tensor(nan+infj)
>>> torch.randn(2000, dtype=torch.cfloat)
tensor([ 0.3015-0.2502j, -1.1102+1.2218j, -0.6324+0.0640j,  ...,
        -1.0200-0.2302j,  0.6511-0.1889j, -0.1069+0.1702j])
>>> torch.tensor([1e-3, 3+4j, 1e-5j, 1e-2+3j, 5+1e-6j])
tensor([1.0000e-03+0.0000e+00j, 3.0000e+00+4.0000e+00j, 0.0000e+00+1.0000e-05j,
        1.0000e-02+3.0000e+00j, 5.0000e+00+1.0000e-06j])
>>> torch.randn(3, dtype=torch.cfloat)
tensor([ 1.0992-0.4459j,  1.1073+0.1202j, -0.2177-0.6342j])
>>> x = torch.tensor([1e2, 1e-2])
>>> torch.set_printoptions(sci_mode=False)
>>> x
tensor([  100.0000,     0.0100])
>>> x = torch.tensor([1e2, 1e-2j])
>>> x
tensor([100.+0.0000j,   0.+0.0100j])
```

Test Plan: Imported from OSS

Differential Revision: D22309294

Pulled By: anjali411

fbshipit-source-id: 20edf9e28063725aeff39f3a246a2d7f348ff1e8

Co-authored-by: anjali411 <chourdiaanjali123@gmail.com>
2020-07-02 09:45:35 -07:00
c5c8a85a82 If ninja is being used, force build_ext to run. (#40881)
As ninja has accurate dependency tracking, if there is nothing to do,
then we will very quickly noop.  But this is important for correctness:
if a change was made to a header that is not listed explicitly in
the distutils Extension, then distutils will come to the wrong
conclusion about whether or not recompilation is needed (but Ninja
will work it out.)

This caused https://github.com/pytorch/vision/issues/2367

Signed-off-by: Edward Z. Yang <ezyang@fb.com>

ghstack-source-id: 6409595c8ac091f3863f305c123266b9d3a167ad
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40837
2020-07-02 08:05:25 -07:00
b4b8f5b9d4 Release GIL during DDP construction. (#40877)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40495

As part of debugging flaky ddp_under_dist_autograd tests, I realized
we were running into the following deadlock.

1) Rank 0 would go into DDP construction, hold GIL and wait for broadcast in
DDP construction.
2) Rank 3 is a little slower and performs an RRef fetch call before the DDP
construction.
3) The RRef fetch call is done on Rank 0 and tries to acquire GIL.
4) We now have a deadlock since Rank 0 is waiting for Rank 3 to enter the
collective and Rank 3 is waiting for Rank 0 to release GIL.
ghstack-source-id: 106534442

Test Plan:
1) Ran ddp_under_dist_autograd 500 times.
2) waitforbuildbot

Differential Revision: D22205180

fbshipit-source-id: 6afd55342e801b9edb9591ff25158a244a8ea66a

Co-authored-by: Pritam Damania <pritam.damania@fb.com>
2020-07-01 13:36:50 -07:00
41816dc97f [1.6] Fix dictConstruct ordering and enable dict mix (#40797)
A combination of https://github.com/pytorch/pytorch/pull/39601 and
https://github.com/pytorch/pytorch/pull/40424 both are approved and
merged in master
2020-07-01 09:30:16 -07:00
31d9776c04 [1.6] fix autograd doc subsubsection display issue (#40796)
Master branch PR: https://github.com/pytorch/pytorch/pull/40582
2020-07-01 09:28:25 -07:00
ddea6c552f Ports full dtype inference deprecation to 1.6 (#40799)
* ports full deprecation

* fixtures

* Fixes lint

* Trying to fix phantom lint issue

* nuclear lint option

* Paradoxical linter fix

Co-authored-by: Mike Ruberry <mruberry@devfair044.maas>
2020-07-01 09:27:27 -07:00
091537a764 [JIT][1.6] Shape analysis fixes. (#40716)
* [JIT] Update type of the unsqueeze's output in shape analysis.

* [JIT] Fix shape analysis for aten::masked_select.

The reference says that this op always returns a 1-D tensor, even if
the input and the mask are 0-D.
2020-07-01 08:41:05 -07:00
bf4d905ea1 Fix wrong MSVC version constraint for CUDA 9.2 (#40794) (#40849)
Summary:
Tested with https://github.com/pytorch/pytorch/pull/40782.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40794

Differential Revision: D22318045

Pulled By: malfet

fbshipit-source-id: a737ffd7cb8a6a9efb62b84378318f4c3800ad8f
2020-07-01 08:37:40 -07:00
415e499330 Fix zip serialization for file > 2GiB for Windows (#40852) 2020-07-01 08:36:40 -07:00
eaf7dad5d6 [1.6 cherrypick] Support Pathlike for zipfile serialization (#40793) 2020-06-30 10:38:00 -07:00
75a074abdc 1.6 Port: Dynamic Versioning (#40542)
Co-authored-by: Mike Ruberry <mruberry@devfair044.maas>
2020-06-30 10:18:18 -07:00
dede34eab7 [1.6 cherrypick] Doc fix for complex views
Cherry-pick of https://github.com/pytorch/pytorch/pull/40450

Test Plan: Imported from OSS
2020-06-30 09:37:02 -07:00
0c90b6da5c [1.6 cherrypick] Fix zip serialization for file > 2GiB (#40757)
* [1.6 cherrypick] Fix zip serialization for file > 2GiB

* Update test/test_serialization.py

Co-authored-by: Nikita Shulga <nikita.shulga@gmail.com>
2020-06-30 07:10:02 -07:00
4316199832 Add examples and tests for combining static/class method with async execution (#40619) (#40688)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/40619

Test Plan: Imported from OSS

Differential Revision: D22258407

Pulled By: mrshenli

fbshipit-source-id: 036d85a2affc4505efd2df197fc513dba010e359
2020-06-29 19:34:23 -07:00
f993e5ac88 [1.6] Update TensorPipe submodule (#40634)
Upstream PR: #40614

Summary:
This update pulls in a oneliner fix, which sets the TCP_NODELAY option on the TCP sockets of the UV transport. This leads to exceptional performance gains in terms of latency, with about a 25x improvement in one simple benchmark. This thus resolves a regression that TensorPipe had compared to the ProcessGroup agent and, in fact, ends up beating it by 2x.

The benchmark I ran is this, with the two endpoints pinned to different cores of the same machine:
```
torch.jit.script
def remote_fn(t: int):
    return t

torch.jit.script
def local_fn():
    for _ in range(1_000_000):
        fut = rpc.rpc_async("rhs", remote_fn, (42,))
        fut.wait()
```

And the average round-trip time (one iteration) is:
- TensorPipe with SHM: 97.2 us
- TensorPipe with UV _after the fix_: 205us
- Gloo: 440us
- TensorPipe with UV _before the fix_: 5ms

Test Plan: Ran PyTorch RPC test suite
2020-06-29 19:33:32 -07:00
c5bd737f0c [JIT] script if tracing fix (#40468) (#40572)
Summary:
Currently, torchvision annotates `batched_nms` with `torch.jit.script` so the function gets compiled when it is traced and ONNX will work. Unfortunately, this means we are eagerly compiling batched_nms, which fails if torchvision isn't built with `torchvision.ops.nms`. As a result, torchvision doesn't work on torch hub right now.

`_script_if_tracing` could solve our problem here, but right now it does not correctly interact with recursive compilation. This PR fixes that bug.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40468

Reviewed By: jamesr66a

Differential Revision: D22195771

Pulled By: eellison

fbshipit-source-id: 83022ca0bab6d389a48a478aec03052c9282d2b7

Co-authored-by: Elias Ellison <eellison@fb.com>
2020-06-29 19:30:41 -07:00
fe45c2c986 Allow slicing sequential container (#40538)
- fixes #38034
- works around missing slice functionality in Sequential
  by casting to tuple and slicing that instead
- supports iterating on the resulting slice but not call()
2020-06-29 19:29:19 -07:00
a9996bb482 Fixes caffe2 loading issues on Windows (#39513) (#40487)
Summary:
Addresses https://github.com/pytorch/pytorch/issues/27840#issuecomment-638715422.
Contains a bunch of fixes (https://github.com/pytorch/pytorch/pull/39376 + https://github.com/pytorch/pytorch/pull/39334 + https://github.com/pytorch/pytorch/pull/38302 + https://github.com/pytorch/pytorch/pull/35362)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/39513

Differential Revision: D22190761

Pulled By: malfet

fbshipit-source-id: b2d52f6cb16c233d16071e9c0670dfff7da2710e
(cherry picked from commit e2201e2ed8ed7bf9c6226f8c484192949d94c248)
2020-06-29 19:17:34 -07:00
bdfcbfa18c [release/1.6] .jenkins: Install torch from test channel (#40706)
We're on a test branch so we should install from the test channel

Signed-off-by: Eli Uriegas <eliuriegas@fb.com>
2020-06-29 13:53:14 -07:00
ea1b0dba18 Remove constexpr for NVCC on Windows (#40676) 2020-06-29 13:48:50 -07:00
6d85b2c989 Pin XLA CI to use r1.6 release branch. (#40721) 2020-06-29 13:41:14 -07:00
44f79651a7 Tweak file_diff_from_base for release/1.6 branch (#40712) 2020-06-29 11:41:46 -07:00
8682ac147b Docs merge (#40569)
Co-authored-by: Elias Ellison <eellison@fb.com>
2020-06-26 12:24:08 -07:00
4cc605e80a (1.6) Update docs feature classifications (#40539)
Co-authored-by: Eli Uriegas <1700823+seemethere@users.noreply.github.com>
2020-06-26 12:23:02 -07:00
b0cce716f7 Add beta warning for quant docs (#40540)
Add a beta warning to match stable and master docs: https://github.com/pytorch/pytorch/blob/master/docs/source/quantization.rst
2020-06-26 12:20:06 -07:00
0dc93ac119 [v1.6.0 patch] Install method docstrings from PyRRef to RRef (#40620)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/40461

It turned out `:inheried-members:` (see [doc](https://www.sphinx-doc.org/en/master/usage/extensions/autodoc.html#directive-autoclass)) is not really usable.

Because pybind11 generates a docstring that writes `self` as parent class, `rpc.PyRRef`, type.

As a workaround, I am pulling docstrings on parent-class, `PyRRef` class, into subclass, `RRef`. And do surgery on the docstring generated by pybind11.

{F241283111}

ghstack-source-id: 106472496

P134031188

Differential Revision: D7933834

fbshipit-source-id: c03a8a4c9d98888b64492a8caba1591595bfe247

Co-authored-by: Shihao Xu <shihaoxu@fb.com>
2020-06-26 12:15:28 -07:00
bb848df10b [1.6] Remove table of contents at the top of rpc.rst (#40482)
Master PR: https://github.com/pytorch/pytorch/pull/40205

Remove the table of contents created by the `.. contents:: :local: :depth: 2` since this page isn't one of the large documentation pages (https://github.com/pytorch/pytorch/issues/38010) and is simply a landing page for the Distributed RPC Framework.

Changes made in this original PR: f10fbcc820 (diff-250b9b23fd6f1a5c15aecdb72afb9d7d)
2020-06-26 08:37:49 -07:00
2dc0b84aca Skip test_mem_leak on Windows (#40498)
(cherry picked from commit 3fb6f038256a3a5ce43e857409ce4ffb807d93a5)
2020-06-25 16:45:48 -07:00
168cddf5f1 .circleci: Fix upload to backup directory
Signed-off-by: Eli Uriegas <eliuriegas@fb.com>
2020-06-23 20:57:42 -07:00
bc8760b3db .circleci: Fix pip installation of awscli
Signed-off-by: Eli Uriegas <eliuriegas@fb.com>
2020-06-23 19:05:48 -07:00
4269b9a8fc .circleci: Fix backup uploads
awscli was not loaded on conda builds and the backup upload did not work
since it was a recursive copy instead of just specifically copying what
we want.

Signed-off-by: Eli Uriegas <eliuriegas@fb.com>
2020-06-23 18:18:06 -07:00
6486 changed files with 183119 additions and 755322 deletions

View File

@ -1,63 +0,0 @@
# PyTorch CI Builds Pipeline on Azure DevOps
#
# This pipeline:
# 1) builds PyTorch on select configurations
# 2) runs only TestTorch unit tests.
stages:
- stage: 'Build'
displayName: 'Build PyTorch'
jobs:
- template: job_templates/build-verify-publish-template-unix.yml
parameters:
name: ubuntu_1804_CPU_docker
pool: 'PyTorch-Linux-CPU'
container_endpoint: pytorchms.azurecr.io
build_stage: True
is_ci_build: True
os: ubuntu
cuda: cpu
customMatrixes:
Py_38:
configuration: ubuntu_1804_py_38_cpu
container_image: pytorchms.azurecr.io/ubuntu_1804_py_38_cpu_dev
- template: job_templates/build-verify-publish-template-unix.yml
parameters:
name: ubuntu_1804_GPU_docker
pool: 'PyTorch-Linux-GPU'
container_endpoint: pytorchms.azurecr.io
build_stage: True
is_ci_build: True
os: ubuntu
cuda: gpu
customMatrixes:
Py_39_CUDA_112_cuDNN_810:
configuration: ubuntu_1804_py_39_cuda_112_cudnn_810
container_image: pytorchms.azurecr.io/ubuntu_1804_py_39_cuda_112_cudnn_8_dev
CUDA_VERSION: 112
- template: job_templates/build-verify-publish-template-win.yml
parameters:
name: windows_2019_CPU
pool: 'PyTorch-Win-CPU'
build_stage: True
is_ci_build: True
os: windows
cuda: cpu
customMatrixes:
Py_37:
configuration: windows_2019_py_37_cpu
- template: job_templates/build-verify-publish-template-win.yml
parameters:
name: windows_2019_GPU
pool: 'PyTorch-Win-GPU'
build_stage: True
is_ci_build: True
os: windows
cuda: gpu
customMatrixes:
Py_38_CUDA_102_cuDNN_765:
configuration: windows_2019_py_38_cuda_102_cudnn_765
CUDA_VERSION: 102

View File

@ -1,82 +0,0 @@
# PyTorch Daily Builds Pipeline on Azure DevOps
#
# This pipeline:
# 1) builds PyTorch on all available configurations
# 2) runs all PyTorch unit tests
stages:
- stage: 'BuildTest'
displayName: 'Build and Test PyTorch'
jobs:
- template: job_templates/build-verify-publish-template-unix.yml
parameters:
name: ubuntu_1804_CPU_docker
pool: 'PyTorch-Linux-CPU'
container_endpoint: pytorchms.azurecr.io
build_stage: True
is_daily_build: True
os: ubuntu
cuda: cpu
customMatrixes:
Py_38:
configuration: ubuntu_1804_py_38_cpu
container_image: pytorchms.azurecr.io/ubuntu_1804_py_38_cpu_dev
Py_37:
configuration: ubuntu_1804_py_37_cpu
container_image: pytorchms.azurecr.io/ubuntu_1804_py_37_cpu_dev
- template: job_templates/build-verify-publish-template-unix.yml
parameters:
name: ubuntu_1804_GPU_docker
pool: 'PyTorch-Linux-GPU'
container_endpoint: pytorchms.azurecr.io
build_stage: True
is_daily_build: True
os: ubuntu
cuda: gpu
customMatrixes:
Py_39_CUDA_112_cuDNN_810:
configuration: ubuntu_1804_py_39_cuda_112_cudnn_810
container_image: pytorchms.azurecr.io/ubuntu_1804_py_39_cuda_112_cudnn_8_dev
CUDA_VERSION: 112
Py_38_CUDA_102_cuDNN_810:
configuration: ubuntu_1804_py_38_cuda_102_cudnn_810
container_image: pytorchms.azurecr.io/ubuntu_1804_py_38_cuda_102_cudnn_8_dev
CUDA_VERSION: 102
Py_37_CUDA_101_cuDNN_765:
configuration: ubuntu_1804_py_37_cuda_101_cudnn_765
container_image: pytorchms.azurecr.io/ubuntu_1804_py_37_cuda_101_cudnn_7_dev
CUDA_VERSION: 101
- template: job_templates/build-verify-publish-template-win.yml
parameters:
name: windows_2019_CPU
pool: 'PyTorch-Win-CPU'
build_stage: True
is_daily_build: True
os: windows
cuda: cpu
customMatrixes:
Py_38:
configuration: windows_2019_py_38_cpu
Py_37:
configuration: windows_2019_py_37_cpu
- template: job_templates/build-verify-publish-template-win.yml
parameters:
name: windows_2019_GPU
pool: 'PyTorch-Win-GPU'
build_stage: True
is_daily_build: True
os: windows
cuda: gpu
customMatrixes:
Py_39_CUDA_112_cuDNN_810:
configuration: windows_2019_py_39_cuda_112_cudnn_810
CUDA_VERSION: 112
Py_38_CUDA_102_cuDNN_765:
configuration: windows_2019_py_38_cuda_102_cudnn_765
CUDA_VERSION: 102
Py_37_CUDA_101_cuDNN_764:
configuration: windows_2019_py_37_cuda_101_cudnn_764
CUDA_VERSION: 101

View File

@ -1,134 +0,0 @@
# PyTorch build steps template with Unix images Azure DevOps Instances
#
# This build depends on 3 parameters set as environment variables in the pipeline:
# - AZURE_DEVOPS_CLI_PAT: Secret var for authenticating to Azure DevOps
# - AZURE_DEVOPS_ARTIFACTS_ORGANIZATION: Azure Artifacts Organization name to publish artifacts
# - AZURE_DEVOPS_ARTIFACTS_PROJECT: Azure Artifacts Project name to publish artifacts
parameters:
name: ''
pool: ''
container_endpoint: ''
os: ''
cuda: ''
is_ci_build: False
is_official_build: False
is_daily_build: False
build_stage: False
verify_stage: False
publish_stage: False
customMatrixes: ''
jobs:
- job: ${{parameters.name}}
timeoutInMinutes: 300
strategy:
matrix:
${{ insert }}: ${{parameters.customMatrixes}}
pool:
name: ${{ parameters.pool}}
variables:
DECODE_PERCENTS: false
container:
image: $[variables['container_image']]
endpoint: ${{parameters.container_endpoint}}
steps:
# Build stage
- ${{ if eq(parameters.build_stage, 'True') }}:
# Set up environment variables for specific pipeline build
- template: set-environment-variables.yml
parameters:
os: ${{ parameters.os}}
cuda: ${{ parameters.cuda}}
is_official_build: ${{ parameters.is_official_build}}
# Sync and update PyTorch submodules
- bash: git submodule update --init --recursive
displayName: Update PyTorch submodules
# Build PyTorch and run unit tests - no packaging
- ${{ if or(eq(parameters.is_ci_build, 'True'), eq(parameters.is_daily_build, 'True')) }}:
# Build PyTorch from source in develop mode
- bash: python setup.py develop
displayName: Build PyTorch from source
- ${{ if eq(parameters.is_ci_build, 'True') }}:
# Run TestTorch unit tests to demonstrate successful PyTorch build
- bash: python test/test_torch.py TestTorch
displayName: Run TestTorch unit tests
- ${{ if eq(parameters.is_daily_build, 'True') }}:
# Run all unit tests to demonstrate successful PyTorch build
- bash: python test/run_test.py --continue-through-error --exclude-jit-executor --verbose
displayName: Run all unit tests
# Run ComponentGovernance
- task: ComponentGovernanceComponentDetection@0
inputs:
scanType: 'Register'
verbosity: 'Verbose'
alertWarningLevel: 'High'
# Build PyTorch and produce artifacts for verification stage
- ${{ if eq(parameters.is_official_build, 'True') }}:
# Build PyTorch from source in install mode and exclude test binaries
- bash: python setup.py install
displayName: Build PyTorch from source without test binaries
# Package PyTorch Wheel
- bash: python setup.py bdist_wheel
displayName: Package PyTorch Wheel
# Publish PyTorch Wheel
- task: PublishPipelineArtifact@1
inputs:
targetPath: $(Build.SourcesDirectory)/dist/
artifactName: Build_$(Build.BuildNumber)_$(configuration)
displayName: Publish PyTorch Wheel to Pipeline Artifacts
# Verification stage
- ${{ if eq(parameters.verify_stage, 'True') }}:
# Download PyTorch Wheel
- task: DownloadPipelineArtifact@2
inputs:
artifact: Build_$(Build.BuildNumber)_$(configuration)
path: $(Build.SourcesDirectory)/verify
displayName: Download PyTorch Wheel
# Install PyTorch Wheel on Windows
- bash: python -m pip install $(Build.SourcesDirectory)/verify/torch*linux*.whl
displayName: Install PyTorch Wheel
# Ensure PyTorch installed correctly from produced wheel
- bash: |
cd $(Build.SourcesDirectory)/verify
python -c "import torch; print('Installed Torch version: ' + torch.__version__)"
displayName: Check PyTorch correctly installed from wheel
# Publishing stage
- ${{ if eq(parameters.publish_stage, 'True') }}:
# Download PyTorch Wheel
- task: DownloadPipelineArtifact@2
inputs:
artifact: Build_$(Build.BuildNumber)_$(configuration)
path: $(Build.SourcesDirectory)/publish
displayName: Download PyTorch Wheel
# Publish wheel to Azure Artifacts
# The flag continueOnError=true is needed as the artifact to be published
# may already exist, because the artifact is differentiated based on the
# last commit date.
- bash: |
export TORCH_VERSION=$(head -c 5 ./version.txt)
export LAST_COMMIT=$(git rev-parse --short HEAD)
export LAST_COMMIT_DATE=$(git log -1 --pretty=%ad --date=format:%Y%m%d)
cd $(Build.SourcesDirectory)/publish
export TORCH_WHEEL=$(echo torch*linux*whl)
az extension add -n azure-devops
echo $ADOTOKEN | az devops login
az artifacts universal publish --organization $AZURE_DEVOPS_ARTIFACTS_ORGANIZATION --project $AZURE_DEVOPS_ARTIFACTS_PROJECT --scope project --feed "PyTorch" --name $TORCH_WHEEL --description "PyTorch Official Build Artifact" --version $TORCH_VERSION-$LAST_COMMIT_DATE-$LAST_COMMIT --path .
env:
ADOTOKEN: $(AZURE_DEVOPS_CLI_PAT)
continueOnError: true
displayName: Upload PyTorch Official Build package to Azure Artifacts

View File

@ -1,150 +0,0 @@
# PyTorch build steps template with Windows images Azure DevOps Instances
#
# This build depends on 3 parameters set as environment variables in the pipeline:
# - AZURE_DEVOPS_CLI_PAT: Secret var for authenticating to Azure DevOps
# - AZURE_DEVOPS_ARTIFACTS_ORGANIZATION: Azure Artifacts Organization name to publish artifacts
# - AZURE_DEVOPS_ARTIFACTS_PROJECT: Azure Artifacts Project name to publish artifacts
parameters:
name: ''
pool: ''
os: ''
cuda: ''
is_ci_build: False
is_official_build: False
is_daily_build: False
build_stage: False
verify_stage: False
publish_stage: False
customMatrixes: ''
jobs:
- job: ${{parameters.name}}
timeoutInMinutes: 300
strategy:
matrix:
${{ insert }}: ${{parameters.customMatrixes}}
pool:
name: ${{ parameters.pool}}
variables:
CMAKE_GENERATOR: Ninja
PACKAGE_PDBS: 0
steps:
# Prepare for PyTorch build on Windows
- template: prepare-build-template.yml
parameters:
configuration: $(configuration)
build_stage: ${{ parameters.build_stage}}
# Build Stage
- ${{ if eq(parameters.build_stage, 'True') }}:
# Set up environment variables for specific pipeline build
- template: set-environment-variables.yml
parameters:
os: ${{ parameters.os}}
cuda: ${{ parameters.cuda}}
is_official_build: ${{ parameters.is_official_build}}
# Sync and update PyTorch submodules
- script: git submodule update --init --recursive
displayName: Update PyTorch submodules
# Build PyTorch and run unit tests - no packaging
- ${{ if or(eq(parameters.is_ci_build, 'True'), eq(parameters.is_daily_build, 'True')) }}:
# Build PyTorch from source in develop mode with Ninja
- script: call activate $(configuration) && python setup.py develop
displayName: Build PyTorch from source
- ${{ if eq(parameters.is_ci_build, 'True') }}:
# Run TestTorch unit tests to demonstrate successful PyTorch build
- script: call activate $(configuration) && python test\test_torch.py TestTorch
displayName: Run TestTorch unit tests
- ${{ if eq(parameters.is_daily_build, 'True') }}:
# Run all unit tests to demonstrate successful PyTorch build
- script: call activate $(configuration) && python test/run_test.py --continue-through-error --exclude-jit-executor --verbose
displayName: Run all unit tests
# Run ComponentGovernance
- task: ComponentGovernanceComponentDetection@0
inputs:
scanType: 'Register'
verbosity: 'Verbose'
alertWarningLevel: 'High'
# Build PyTorch and produce artifacts for verification stage
- ${{ if eq(parameters.is_official_build, 'True') }}:
# Build PyTorch from source in install mode with Ninja and exclude test binaries
- script: call activate $(configuration) && python setup.py install
displayName: Build PyTorch from source without test binaries
# Package PyTorch Wheel
- script: call activate $(configuration) && python setup.py bdist_wheel
displayName: Package PyTorch Wheel
# Publish PyTorch Wheel
- task: PublishPipelineArtifact@1
inputs:
targetPath: $(Build.SourcesDirectory)\dist\
artifactName: Build_$(Build.BuildNumber)_$(configuration)
displayName: Publish PyTorch Wheel to Pipeline Artifacts
# Verification Stage
- ${{ if eq(parameters.verify_stage, 'True') }}:
# Download PyTorch Wheel
- task: DownloadPipelineArtifact@2
inputs:
artifact: Build_$(Build.BuildNumber)_$(configuration)
path: $(Build.SourcesDirectory)\verify
displayName: Download PyTorch Wheel
# Install PyTorch Wheel on Windows
- script: |
call activate $(configuration)
cd $(Build.SourcesDirectory)\verify
dir torch*win*.whl /b > whl.txt
set /p whl= < whl.txt
python -m pip install %whl%
displayName: Install PyTorch Wheel
# Ensure PyTorch installed correctly from produced wheel
- script: |
call activate $(configuration)
cd $(Build.SourcesDirectory)\verify
python -c "import torch; print('Installed Torch version: ' + torch.__version__)"
displayName: Check PyTorch correctly installed from wheel
# Publishing stage
- ${{ if eq(parameters.publish_stage, 'True') }}:
# Download PyTorch Wheel
- task: DownloadPipelineArtifact@2
inputs:
artifact: Build_$(Build.BuildNumber)_$(configuration)
path: $(Build.SourcesDirectory)\publish
displayName: Download PyTorch Wheel
# Set up Azure Artifacts for Windows
# The pip install --upgrade command is a bug fix for Azure CLI on Windows
# More info: https://github.com/Azure/azure-cli/issues/16858
- script: |
pip install --upgrade pip --target \opt\az\lib\python3.6\site-packages\
az extension add -n azure-devops
displayName: Set up Azure Artifacts download on Windows
# Publish wheel to Azure Artifacts
# The flag continueOnError=true is needed as the artifact to be published
# may already exist, because the artifact is differentiated based on the
# last commit date.
- script: |
set /p TORCH_VERSION= < version.txt
cd $(Build.SourcesDirectory)\publish
git rev-parse --short HEAD > last_commit.txt && set /p LAST_COMMIT= < last_commit.txt
git log -1 --pretty=%ad --date=format:%Y%m%d > last_commit_date.txt && set /p LAST_COMMIT_DATE= < last_commit_date.txt
dir torch*win*.whl /b > whl.txt && set /p TORCH_WHEEL= < whl.txt
echo %ADOTOKEN% | az devops login
az artifacts universal publish --organization %AZURE_DEVOPS_ARTIFACTS_ORGANIZATION% --project %AZURE_DEVOPS_ARTIFACTS_PROJECT% --scope project --feed "PyTorch" --name %TORCH_WHEEL% --description "PyTorch Official Build Artifact" --version %TORCH_VERSION:~0,5%-%LAST_COMMIT_DATE%-%LAST_COMMIT% --path .
env:
ADOTOKEN: $(AZURE_DEVOPS_CLI_PAT)
continueOnError: true
displayName: Upload PyTorch nigthly package to Azure Artifacts

View File

@ -1,17 +0,0 @@
dependencies:
- python=PYTHON_VERSION
- numpy
- ninja
- pyyaml
- mkl
- mkl-include
- setuptools
- cmake
- cffi
- typing_extensions
- future
- six
- requests
- dataclasses
- pip:
- -r ../../requirements.txt

View File

@ -1,62 +0,0 @@
# Build prepare steps for PyTorch on Azure DevOps to build from source.
# These steps share between normal build process and semmle security scan tasks
parameters:
build_stage: False
configuration: ''
steps:
# End Python tasks that may be lingering over from previous runs
# Note: If python.exe isn't currently running, exit code becomes 128,
# which fails the run. Here exit code is set to 0 to avoid failed run.
- script: |
taskkill /f /im python.exe
IF %ERRORLEVEL% EQU 128 exit 0
displayName: End previous Python processes
# Clean up env directory in conda for fresh builds and set up conda environment YAML
- powershell: |
Remove-Item 'C:\Miniconda\envs' -Recurse -ErrorAction Ignore
$env:PYTHON_VERSION = $env:SYSTEM_JOBNAME.Substring(3,1) + '.' + $env:SYSTEM_JOBNAME.Substring(4,1)
(Get-Content .azure_pipelines\job_templates\common-packages.yml) -replace 'PYTHON_VERSION', $env:PYTHON_VERSION | Out-File -encoding ASCII .azure_pipelines\job_templates\common-packages.yml
displayName: Clean up previous environments and Set up conda environment YAML
# Make conda environment and install required packages
- script: |
call conda clean --all -y
call conda env create -n $(configuration) --file .azure_pipelines\job_templates\common-packages.yml
call activate $(configuration)
call conda install -c conda-forge libuv=1.39
displayName: Set up conda environment for building from source
- ${{ if eq(parameters.build_stage, 'True') }}:
# Install MKL
- script: |
rmdir /s /q mkl
del mkl_2020.2.254.7z
curl https://s3.amazonaws.com/ossci-windows/mkl_2020.2.254.7z -k -O
7z x -aoa mkl_2020.2.254.7z -omkl
displayName: Install MKL
# Install sccache and randomtemp
# Related PyTorch GitHub issue: https://github.com/pytorch/pytorch/issues/25393
# Related fix: https://github.com/pytorch/builder/pull/448/
- script: |
mkdir .\tmp_bin
curl -k https://s3.amazonaws.com/ossci-windows/sccache.exe --output .\tmp_bin\sccache.exe
curl -k https://s3.amazonaws.com/ossci-windows/sccache-cl.exe --output .\tmp_bin\sccache-cl.exe
copy .\tmp_bin\sccache.exe .\tmp_bin\nvcc.exe
curl -kL https://github.com/peterjc123/randomtemp-rust/releases/download/v0.3/randomtemp.exe --output .\tmp_bin\randomtemp.exe
displayName: Install sccache and randomtemp
condition: not(eq(variables.CUDA_VERSION, ''))
# CUDA 11.2's CUB directory conflicts with CUDA 10.2 and 10.1
# builds, where CUDA 11.2's CUB is injected into non-CUDA
# 11.2 builds.
- powershell: Remove-Item "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.2\include\cub" -Recurse -ErrorAction Ignore
displayName: Remove conflicting CUB from CUDA installation
condition: not(eq(variables.CUDA_VERSION, ''))
- powershell: Copy-Item -Path "F:\cuda_11_2\cub\" -Destination "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.2\include" -Recurse
displayName: Copy CUDA CUB for CUDA 11.2 build
condition: eq(variables.CUDA_VERSION, '112')

View File

@ -1,51 +0,0 @@
# PyTorch build steps template with Unix images Azure DevOps Instances
#
# This build depends on 5 parameters set as an environment variables in the pipeline:
# - AZURE_DEVOPS_CLI_PAT: Secret var for authenticating to Azure DevOps
# - AZURE_STORAGE_KEY: Secret var for authenticating to Azure Storage
# - _TS_CLONE_P, _TS_P, _TS_SM_P: Secret vars for specific unit tests
parameters:
name: ''
pool: ''
container_endpoint: ''
customMatrixes: ''
jobs:
- job: ${{parameters.name}}
timeoutInMinutes: 600
strategy:
matrix:
${{ insert }}: ${{parameters.customMatrixes}}
pool:
name: ${{ parameters.pool}}
variables:
DECODE_PERCENTS: false
steps:
# Don't checkout repo contents to save time and CPU compute. Environment variables
# related to checkout branch such as $(BUILD_SOURCEBRANCH) are still available.
- checkout: none
# Delete pytorch_tests repo from previous builds if exists
- bash: rm -rf pytorch_tests/
displayName: Delete pytorch_tests repo from previous builds if exists
# Clone PyTorch Tests repository
- bash: |
B64_PAT=$(printf "%s"":$_ADOTOKEN" | base64)
git -c http.extraHeader="Authorization: Basic ${B64_PAT}" clone $(AZURE_DEVOPS_PYTORCH_TESTS_REPO_URL)
cd pytorch_tests
git checkout $(PYTORCH_TESTS_CHECKOUT_BRANCH)
env:
_ADOTOKEN: $(AZURE_DEVOPS_CLI_PAT)
displayName: Clone PyTorch Tests repo
# Run PyTorch Unit Tests
- bash: bash $(Build.SourcesDirectory)/pytorch_tests/scripts/linux/run.sh
env:
_AZURE_STORAGE_KEY: $(AZURE_STORAGE_KEY)
_TS_CLONE_P: $(TS_CLONE_PASSWORD)
_TS_P: $(TS_PAT)
_TS_SM_P: $(TS_SM_PAT)
displayName: Run PyTorch Unit Tests

View File

@ -1,49 +0,0 @@
# PyTorch build steps template with Windows images Azure DevOps Instances
#
# This build depends on 5 parameters set as an environment variables in the pipeline:
# - AZURE_DEVOPS_CLI_PAT: Secret var for authenticating to Azure DevOps
# - AZURE_STORAGE_KEY: Secret var for authenticating to Azure Storage
# - _TS_CLONE_P, _TS_P, _TS_SM_P: Secret vars for specific unit tests
parameters:
name: ''
pool: ''
customMatrixes: ''
jobs:
- job: ${{parameters.name}}
timeoutInMinutes: 600
strategy:
matrix:
${{ insert }}: ${{parameters.customMatrixes}}
pool:
name: ${{ parameters.pool}}
steps:
# Don't checkout repo contents to save time and CPU compute. Environment variables
# related to checkout branch such as $(BUILD_SOURCEBRANCH) are still available.
- checkout: none
# Delete pytorch_tests repo from previous builds if exists
- script: if exist "pytorch_tests/" rmdir "pytorch_tests/" /q /s
displayName: Delete pytorch_tests repo from previous builds if exists
# Clone PyTorch Tests repository
- powershell: |
$env:B64Pat = [Convert]::ToBase64String([System.Text.Encoding]::UTF8.GetBytes(":$env:_ADOTOKEN"))
git -c http.extraHeader="Authorization: Basic $env:B64Pat" clone $env:AZURE_DEVOPS_pytorch_tests_REPO_URL
cd pytorch_tests
git checkout $(PYTORCH_TESTS_CHECKOUT_BRANCH)
env:
_ADOTOKEN: $(AZURE_DEVOPS_CLI_PAT)
displayName: Clone PyTorch Tests repo
# Run PyTorch Unit Tests
- script: call $(Build.SourcesDirectory)\pytorch_tests\scripts\windows\run.bat
env:
_ADOTOKEN: $(AZURE_DEVOPS_CLI_PAT)
_AZURE_STORAGE_KEY: $(AZURE_STORAGE_KEY)
_TS_CLONE_P: $(TS_CLONE_PASSWORD)
_TS_P: $(TS_PAT)
_TS_SM_P: $(TS_SM_PAT)
displayName: Run PyTorch Unit Tests

View File

@ -1,131 +0,0 @@
# Set environment variables for specific configurations
parameters:
is_official_build: False
os: ''
cuda: ''
steps:
# Environment configuration steps for Ubuntu builds
- ${{ if contains(parameters.os, 'ubuntu') }}:
# Set configuration specific build flags
- ${{ if eq(parameters.is_official_build, True) }}:
- bash: |
echo "##vso[task.setvariable variable=INSTALL_TEST;]0"
echo "##vso[task.setvariable variable=PYTORCH_BUILD_NUMBER;]1"
export PYTORCH_VERSION=$(head -c 5 ./version.txt)
echo "##vso[task.setvariable variable=PYTORCH_BUILD_VERSION;]$PYTORCH_VERSION.dev"
displayName: Set configuration-specific build flags
# Set PyTorch CPU/GPU build flags.
- ${{ if contains(parameters.cuda, 'cpu') }}:
- bash: |
echo "##vso[task.setvariable variable=USE_CUDA;]0"
echo "##vso[task.setvariable variable=PYTORCH_BUILD_VERSION;]$(PYTORCH_BUILD_VERSION).cpu"
displayName: Set CUDA-specific build flag for CPU builds
- ${{ if contains(parameters.cuda, 'gpu') }}:
- bash: |
echo "##vso[task.setvariable variable=USE_CUDA;]1"
echo "##vso[task.setvariable variable=PYTORCH_BUILD_VERSION;]$(PYTORCH_BUILD_VERSION).cu$(CUDA_VERSION)"
displayName: Set CUDA-specific build flag for GPU builds
# Set MKL environment variables
- bash: |
echo "##vso[task.setvariable variable=CMAKE_LIBRARY_PATH;]/opt/intel/lib:$CMAKE_LIBRARY_PATH"
echo "##vso[task.setvariable variable=CMAKE_INCLUDE_PATH;]/opt/intel/include:$CMAKE_INCLUDE_PATH"
displayName: Set MKL paths
# View current environment variables
- bash:
printenv
displayName: Show environment variables
# Environment configuration steps for Windows builds
- ${{ if contains(parameters.os, 'windows') }}:
# Set Conda Lib Path
- powershell: Write-Host "##vso[task.setvariable variable=CONDA_LIB_PATH;]C:\Miniconda\envs\$(configuration)\Library\bin"
displayName: Set Conda Lib Path
# Set configuration specific build flags
- ${{ if eq(parameters.is_official_build, True) }}:
- powershell: |
Write-Host "##vso[task.setvariable variable=INSTALL_TEST;]0"
Write-Host "##vso[task.setvariable variable=PYTORCH_BUILD_NUMBER;]1"
Set-Variable -Name PYTORCH_VERSION -Value (Get-Content .\version.txt).Substring(0,5)
Write-Host "##vso[task.setvariable variable=PYTORCH_BUILD_VERSION;]$PYTORCH_VERSION.dev"
displayName: Set configuration-specific build flags
# Set PyTorch CPU/GPU build flags..
- ${{ if contains(parameters.cuda, 'cpu') }}:
- powershell: |
Write-Host "##vso[task.setvariable variable=USE_CUDA;]0"
Write-Host "##vso[task.setvariable variable=PYTORCH_BUILD_VERSION;]$(PYTORCH_BUILD_VERSION).cpu"
displayName: Set CUDA-specific build flag for CPU build
- ${{ if contains(parameters.cuda, 'gpu') }}:
- powershell: |
Write-Host "##vso[task.setvariable variable=USE_CUDA;]1"
Write-Host "##vso[task.setvariable variable=PYTORCH_BUILD_VERSION;]$(PYTORCH_BUILD_VERSION).cu$(CUDA_VERSION)"
displayName: Set CUDA-specific build flag for GPU build
# Set CUDA 11.2, 10.2 or 10.1 specific build flags
- ${{ if eq(parameters.cuda, 'gpu') }}:
- powershell: |
Write-Host "##vso[task.setvariable variable=TORCH_CUDA_ARCH_LIST;]3.7+PTX;5.0;6.0;6.1;7.0;7.5;8.0;8.6"
Write-Host "##vso[task.setvariable variable=CUDA_PATH;]C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.2\"
displayName: Set CUDA 11.2 specific build flags
condition: eq(variables.CUDA_VERSION, '112')
- powershell: |
Write-Host "##vso[task.setvariable variable=TORCH_CUDA_ARCH_LIST;]3.7+PTX;5.0;6.0;6.1;7.0;7.5"
Write-Host "##vso[task.setvariable variable=CUDA_PATH;]C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.2\"
displayName: Set CUDA 10.2 specific build flags
condition: eq(variables.CUDA_VERSION, '102')
- powershell: |
Write-Host "##vso[task.setvariable variable=TORCH_CUDA_ARCH_LIST;]3.7+PTX;5.0;6.0;6.1;7.0;7.5"
Write-Host "##vso[task.setvariable variable=CUDA_PATH;]C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\"
displayName: Set CUDA 10.1 specific build flags
condition: eq(variables.CUDA_VERSION, '101')
- powershell: |
Write-Host "##vso[task.setvariable variable=CUDA_BIN_PATH;]$env:CUDA_PATH\bin\"
Write-Host "##vso[task.setvariable variable=CUDNN_ROOT;]$env:CUDA_PATH"
Write-Host "##vso[task.setvariable variable=CUDNN_INCLUDE_DIR;]$env:CUDA_PATH\include\"
Write-Host "##vso[task.setvariable variable=CUDNN_LIBRARY;]$env:CUDA_PATH\lib\x64\"
Write-Host "##vso[task.prependpath]$env:CUDA_PATH\bin"
Write-Host "##vso[task.setvariable variable=TORCH_NVCC_FLAGS;]-Xfatbin -compress-all --no-host-device-move-forward"
Write-Host "##vso[task.setvariable variable=THRUST_IGNORE_CUB_VERSION_CHECK;]1"
Write-Host "##vso[task.setvariable variable=NVTOOLSEXT_PATH;]C:\Program Files\NVIDIA Corporation\NvToolsExt\"
displayName: Set CUDA environment variables
- powershell: |
copy "$(CUDA_BIN_PATH)\cusparse*64_*.dll*" $(Build.SourcesDirectory)\torch\lib
copy "$(CUDA_BIN_PATH)\cublas*64_*.dll*" $(Build.SourcesDirectory)\torch\lib
copy "$(CUDA_BIN_PATH)\cudart*64_*.dll*" $(Build.SourcesDirectory)\torch\lib
copy "$(CUDA_BIN_PATH)\curand*64_*.dll*" $(Build.SourcesDirectory)\torch\lib
copy "$(CUDA_BIN_PATH)\cufft*64_*.dll*" $(Build.SourcesDirectory)\torch\lib
copy "$(CUDA_BIN_PATH)\cusolver*64_*.dll*" $(Build.SourcesDirectory)\torch\lib
copy "$(CUDA_BIN_PATH)\cudnn*64_*.dll*" $(Build.SourcesDirectory)\torch\lib
copy "$(CUDA_BIN_PATH)\nvrtc*64_*.dll*" $(Build.SourcesDirectory)\torch\lib
copy "C:\Program Files\NVIDIA Corporation\NvToolsExt\bin\x64\nvToolsExt64_1.dll*" $(Build.SourcesDirectory)\torch\lib
copy "$(CONDA_LIB_PATH)\libiomp*5md.dll" $(Build.SourcesDirectory)\torch\lib
copy "$(CONDA_LIB_PATH)\uv.dll" $(Build.SourcesDirectory)\torch\lib
displayName: Copy CUDA/cuDNN/libomp/libuv dlls to torch\lib
# Set MKL, sccache and randomtemp environment variables
- powershell: |
Write-Host "##vso[task.setvariable variable=CMAKE_INCLUDE_PATH;]$(Build.SourcesDirectory)\mkl\include"
Write-Host "##vso[task.setvariable variable=CMAKE_LIBRARY_PATH;]$(Build.SourcesDirectory)\mkl\lib;$env:CMAKE_LIBRARY_PATH"
Write-Host "##vso[task.setvariable variable=ADDITIONAL_PATH;]$(Build.SourcesDirectory)\tmp_bin"
Write-Host "##vso[task.setvariable variable=SCCACHE_IDLE_TIMEOUT;]1500"
Write-Host "##vso[task.setvariable variable=RANDOMTEMP_EXECUTABLE;]$(Build.SourcesDirectory)\tmp_bin\nvcc.exe"
Write-Host "##vso[task.setvariable variable=CUDA_NVCC_EXECUTABLE;]$(Build.SourcesDirectory)\tmp_bin\randomtemp.exe"
Write-Host "##vso[task.setvariable variable=RANDOMTEMP_BASEDIR;]$(Build.SourcesDirectory)\tmp_bin"
displayName: Set MKL, sccache and randomtemp environment variables
# View current environment variables
- script:
set
displayName: Show environment variables

View File

@ -1,14 +0,0 @@
# Main logic to initiate wait for PR artifact to be ready
steps:
- task: InvokeRESTAPI@1
displayName: 'Wait for job success and wheel ready'
timeoutInMinutes: 60
inputs:
connectionType: 'connectedServiceName'
serviceConnection: circleciconn
method: 'POST'
headers: '{"Content-Type":"application/json", "BranchName":"$(TARGET_BRANCH_TO_CHECK_PR)", "JobName":"$(TARGET_CIRCLECI_PR)", "PlanUrl":"$(System.CollectionUri)", "ProjectId":"$(System.TeamProjectId)", "HubName":"$(System.HostType)", "PlanId":"$(System.PlanId)", "JobId":"$(System.JobId)", "TimelineId":"$(System.TimelineId)", "TaskInstanceId":"$(System.TaskInstanceId)", "AuthToken":"$(System.AccessToken)"}'
body: ''
urlSuffix: 'api/JobStatus'
waitForCompletion: true

View File

@ -1,49 +0,0 @@
# Initiate 5 agentless-server waiting jobs to check on the
# status of PR artifact builds, for a maximum wait time of
# 5 * 60 min =300 minutes. These jobs will pass immediately
# once targeted CircleCI build is ready.
jobs:
- job: checkjob1
pool: server
timeoutInMinutes: 60
continueOnError: true
steps:
- template: wheel-wait-job-template.yml
- job: checkjob2
pool: server
timeoutInMinutes: 60
dependsOn: checkjob1
continueOnError: true
steps:
- template: wheel-wait-job-template.yml
- job: checkjob3
pool: server
timeoutInMinutes: 60
dependsOn: checkjob2
continueOnError: true
steps:
- template: wheel-wait-job-template.yml
- job: checkjob4
pool: server
timeoutInMinutes: 60
dependsOn: checkjob3
continueOnError: true
steps:
- template: wheel-wait-job-template.yml
- job: checkjob5
pool: server
timeoutInMinutes: 60
dependsOn: checkjob4
continueOnError: true
steps:
- template: wheel-wait-job-template.yml

View File

@ -1,50 +0,0 @@
# PyTorch Nightly PyTorch Tests Builds Pipeline on Azure DevOps
#
# This pipeline runs custom PyTorch unit-tests on nightly
# PyTorch wheels.
stages:
- stage: 'NightlyCustomTests'
displayName: 'Run custom unit tests on PyTorch wheels'
jobs:
- template: job_templates/pytorch-template-unix.yml
parameters:
name: ubuntu_1804_CPU_docker
pool: $(BUILD_POOL_LIN_1)
customMatrixes:
Nightly_Custom_Tests:
_DOCKER_IMAGE: $(DOCKER_IMAGE_LIN_1)
_PYTHON_VERSION: $(PYTHON_VERSION_LIN_1)
_CUDA_BUILD_VERSION: $(CUDA_BUILD_VERSION_LIN_1)
_RUN_TESTS: $(RUN_TESTS_LIN)
- template: job_templates/pytorch-template-unix.yml
parameters:
name: ubuntu_1804_GPU_docker
pool: $(BUILD_POOL_LIN_2)
customMatrixes:
Nightly_Custom_Tests:
_DOCKER_IMAGE: $(DOCKER_IMAGE_LIN_2)
_PYTHON_VERSION: $(PYTHON_VERSION_LIN_2)
_CUDA_BUILD_VERSION: $(CUDA_BUILD_VERSION_LIN_2)
_RUN_TESTS: $(RUN_TESTS_LIN)
- template: job_templates/pytorch-template-win.yml
parameters:
name: windows_2019_CPU
pool: $(BUILD_POOL_WIN_1)
customMatrixes:
Nightly_Custom_Tests:
_PYTHON_VERSION: $(PYTHON_VERSION_WIN_1)
_CUDA_BUILD_VERSION: $(CUDA_BUILD_VERSION_WIN_1)
_RUN_TESTS: $(RUN_TESTS_WIN)
- template: job_templates/pytorch-template-win.yml
parameters:
name: windows_2019_GPU
pool: $(BUILD_POOL_WIN_2)
customMatrixes:
Nightly_Custom_Tests:
_PYTHON_VERSION: $(PYTHON_VERSION_WIN_2)
_CUDA_BUILD_VERSION: $(CUDA_BUILD_VERSION_WIN_2)
_RUN_TESTS: $(RUN_TESTS_WIN)

View File

@ -1,30 +0,0 @@
# PyTorch PR PyTorch Tests Builds Pipeline on Azure DevOps
#
# This pipeline:
# 1) ensures that CircleCI builds for a given PR
# have finished, and that its artifacts are
# ready for download
# 2) runs custom PyTorch unit-tests on PyTorch
# wheels generated during PR builds.
stages:
- stage: 'EnsureArtifactsReady'
displayName: 'Ensure PyTorch PR Artifacts are ready'
jobs:
- template: job_templates/wheel-wait-template.yml
- stage: 'PRCustomTests'
displayName: 'Run custom unit tests on PyTorch wheels'
jobs:
- template: job_templates/pytorch-template-unix.yml
parameters:
name: ubuntu_1804_GPU_docker
pool: $(BUILD_POOL_PR)
customMatrixes:
PR_Custom_Tests:
_PYTHON_VERSION: $(PYTHON_VERSION_PR)
_CUDA_BUILD_VERSION: $(CUDA_BUILD_VERSION_PR)
_TARGET_CIRCLECI_BUILD: $(TARGET_CIRCLECI_PR)
_TARGET_BRANCH_TO_CHECK: $(TARGET_BRANCH_TO_CHECK_PR)
_DOCKER_IMAGE: $(DOCKER_IMAGE_PR)
_RUN_TESTS: $(RUN_TESTS_PR)

View File

@ -1,224 +0,0 @@
# PyTorch Official Builds Pipeline on Azure DevOps
#
# This pipeline:
# 1) builds PyTorch on all available configurations
# 2) verifies PyTorch artifacts by installing them in a clean environment
# and checking torch.__version_
# 3) publishes official PyTorch artifacts to Azure DevOps Artifacts for consumption
stages:
- stage: 'Build'
displayName: 'Build PyTorch'
jobs:
- template: job_templates/build-verify-publish-template-unix.yml
parameters:
name: ubuntu_1804_CPU_docker
pool: 'PyTorch-Linux-CPU'
container_endpoint: pytorchms.azurecr.io
build_stage: True
is_official_build: True
os: ubuntu
cuda: cpu
customMatrixes:
Py_38:
configuration: ubuntu_1804_py_38_cpu
container_image: pytorchms.azurecr.io/ubuntu_1804_py_38_cpu_dev
Py_37:
configuration: ubuntu_1804_py_37_cpu
container_image: pytorchms.azurecr.io/ubuntu_1804_py_37_cpu_dev
- template: job_templates/build-verify-publish-template-unix.yml
parameters:
name: ubuntu_1804_GPU_docker
pool: 'PyTorch-Linux-GPU'
container_endpoint: pytorchms.azurecr.io
build_stage: True
is_official_build: True
os: ubuntu
cuda: gpu
customMatrixes:
Py_39_CUDA_112_cuDNN_810:
configuration: ubuntu_1804_py_39_cuda_112_cudnn_810
container_image: pytorchms.azurecr.io/ubuntu_1804_py_39_cuda_112_cudnn_8_dev
CUDA_VERSION: 112
Py_38_CUDA_102_cuDNN_810:
configuration: ubuntu_1804_py_38_cuda_102_cudnn_810
container_image: pytorchms.azurecr.io/ubuntu_1804_py_38_cuda_102_cudnn_8_dev
CUDA_VERSION: 102
Py_37_CUDA_101_cuDNN_765:
configuration: ubuntu_1804_py_37_cuda_101_cudnn_765
container_image: pytorchms.azurecr.io/ubuntu_1804_py_37_cuda_101_cudnn_7_dev
CUDA_VERSION: 101
- template: job_templates/build-verify-publish-template-win.yml
parameters:
name: windows_2019_CPU
pool: 'PyTorch-Win-CPU'
build_stage: True
is_official_build: True
os: windows
cuda: cpu
customMatrixes:
Py_38:
configuration: windows_2019_py_38_cpu
Py_37:
configuration: windows_2019_py_37_cpu
- template: job_templates/build-verify-publish-template-win.yml
parameters:
name: windows_2019_GPU
pool: 'PyTorch-Win-GPU'
build_stage: True
is_official_build: True
os: windows
cuda: gpu
customMatrixes:
Py_39_CUDA_112_cuDNN_810:
configuration: windows_2019_py_39_cuda_112_cudnn_810
CUDA_VERSION: 112
Py_38_CUDA_102_cuDNN_765:
configuration: windows_2019_py_38_cuda_102_cudnn_765
CUDA_VERSION: 102
Py_37_CUDA_101_cuDNN_764:
configuration: windows_2019_py_37_cuda_101_cudnn_764
CUDA_VERSION: 101
- stage: 'Verify'
displayName: 'Verify PyTorch wheels'
dependsOn: Build
condition: succeeded()
jobs:
- template: job_templates/build-verify-publish-template-unix.yml
parameters:
name: ubuntu_1804_CPU_docker
pool: 'PyTorch-Linux-CPU'
container_endpoint: pytorchms.azurecr.io
verify_stage: True
is_official_build: True
customMatrixes:
Py_38:
configuration: ubuntu_1804_py_38_cpu
container_image: pytorchms.azurecr.io/ubuntu_1804_py_38_cpu_dev
Py_37:
configuration: ubuntu_1804_py_37_cpu
container_image: pytorchms.azurecr.io/ubuntu_1804_py_37_cpu_dev
- template: job_templates/build-verify-publish-template-unix.yml
parameters:
name: ubuntu_1804_GPU_docker
pool: 'PyTorch-Linux-GPU'
container_endpoint: pytorchms.azurecr.io
verify_stage: True
is_official_build: True
customMatrixes:
Py_39_CUDA_112_cuDNN_810:
configuration: ubuntu_1804_py_39_cuda_112_cudnn_810
container_image: pytorchms.azurecr.io/ubuntu_1804_py_39_cuda_112_cudnn_8_dev
CUDA_VERSION: 112
Py_38_CUDA_102_cuDNN_810:
configuration: ubuntu_1804_py_38_cuda_102_cudnn_810
container_image: pytorchms.azurecr.io/ubuntu_1804_py_38_cuda_102_cudnn_8_dev
CUDA_VERSION: 102
Py_37_CUDA_101_cuDNN_765:
configuration: ubuntu_1804_py_37_cuda_101_cudnn_765
container_image: pytorchms.azurecr.io/ubuntu_1804_py_37_cuda_101_cudnn_7_dev
CUDA_VERSION: 101
- template: job_templates/build-verify-publish-template-win.yml
parameters:
name: windows_2019_CPU
pool: 'PyTorch-Win-CPU'
verify_stage: True
is_official_build: True
customMatrixes:
Py_38:
configuration: windows_2019_py_38_cpu
Py_37:
configuration: windows_2019_py_37_cpu
- template: job_templates/build-verify-publish-template-win.yml
parameters:
name: windows_2019_GPU
pool: 'PyTorch-Win-GPU'
verify_stage: True
is_official_build: True
customMatrixes:
Py_39_CUDA_112_cuDNN_810:
configuration: windows_2019_py_39_cuda_112_cudnn_810
CUDA_VERSION: 112
Py_38_CUDA_102_cuDNN_765:
configuration: windows_2019_py_38_cuda_102_cudnn_765
CUDA_VERSION: 102
Py_37_CUDA_101_cuDNN_764:
configuration: windows_2019_py_37_cuda_101_cudnn_764
CUDA_VERSION: 101
- stage: 'Publish'
displayName: 'Publish PyTorch wheels'
dependsOn: Verify
condition: succeeded()
jobs:
- template: job_templates/build-verify-publish-template-unix.yml
parameters:
name: ubuntu_1804_CPU_docker
pool: 'PyTorch-Linux-CPU'
container_endpoint: pytorchms.azurecr.io
publish_stage: True
is_official_build: True
customMatrixes:
Py_38:
configuration: ubuntu_1804_py_38_cpu
container_image: pytorchms.azurecr.io/ubuntu_1804_py_38_cpu_dev
Py_37:
configuration: ubuntu_1804_py_37_cpu
container_image: pytorchms.azurecr.io/ubuntu_1804_py_37_cpu_dev
- template: job_templates/build-verify-publish-template-unix.yml
parameters:
name: ubuntu_1804_GPU_docker
pool: 'PyTorch-Linux-GPU'
container_endpoint: pytorchms.azurecr.io
publish_stage: True
is_official_build: True
customMatrixes:
Py_39_CUDA_112_cuDNN_810:
configuration: ubuntu_1804_py_39_cuda_112_cudnn_810
container_image: pytorchms.azurecr.io/ubuntu_1804_py_39_cuda_112_cudnn_8_dev
CUDA_VERSION: 112
Py_38_CUDA_102_cuDNN_810:
configuration: ubuntu_1804_py_38_cuda_102_cudnn_810
container_image: pytorchms.azurecr.io/ubuntu_1804_py_38_cuda_102_cudnn_8_dev
CUDA_VERSION: 102
Py_37_CUDA_101_cuDNN_765:
configuration: ubuntu_1804_py_37_cuda_101_cudnn_765
container_image: pytorchms.azurecr.io/ubuntu_1804_py_37_cuda_101_cudnn_7_dev
CUDA_VERSION: 101
- template: job_templates/build-verify-publish-template-win.yml
parameters:
name: windows_2019_CPU
pool: 'PyTorch-Win-CPU'
publish_stage: True
is_official_build: True
customMatrixes:
Py_38:
configuration: windows_2019_py_38_cpu
Py_37:
configuration: windows_2019_py_37_cpu
- template: job_templates/build-verify-publish-template-win.yml
parameters:
name: windows_2019_GPU
pool: 'PyTorch-Win-GPU'
publish_stage: True
is_official_build: True
customMatrixes:
Py_39_CUDA_112_cuDNN_810:
configuration: windows_2019_py_39_cuda_112_cudnn_810
CUDA_VERSION: 112
Py_38_CUDA_102_cuDNN_765:
configuration: windows_2019_py_38_cuda_102_cudnn_765
CUDA_VERSION: 102
Py_37_CUDA_101_cuDNN_764:
configuration: windows_2019_py_37_cuda_101_cudnn_764
CUDA_VERSION: 101

View File

@ -31,7 +31,7 @@ Usage
1. Make changes to these scripts.
2. Run the `regenerate.sh` script in this directory and commit the script changes and the resulting change to `config.yml`.
You'll see a build failure on GitHub if the scripts don't agree with the checked-in version.
You'll see a build failure on TravisCI if the scripts don't agree with the checked-in version.
Motivation
@ -55,7 +55,7 @@ Future direction
See comment [here](https://github.com/pytorch/pytorch/pull/17323#pullrequestreview-206945747):
In contrast with a full recursive tree traversal of configuration dimensions,
> in the future I think we actually want to decrease our matrix somewhat and have only a few mostly-orthogonal builds that taste as many different features as possible on PRs, plus a more complete suite on every PR and maybe an almost full suite nightly/weekly (we don't have this yet). Specifying PR jobs in the future might be easier to read with an explicit list when we come to this.
> in the future future I think we actually want to decrease our matrix somewhat and have only a few mostly-orthogonal builds that taste as many different features as possible on PRs, plus a more complete suite on every PR and maybe an almost full suite nightly/weekly (we don't have this yet). Specifying PR jobs in the future might be easier to read with an explicit list when we come to this.
----------------
----------------
@ -90,7 +90,7 @@ The binaries are built in CircleCI. There are nightly binaries built every night
We have 3 types of binary packages
* pip packages - nightlies are stored on s3 (pip install -f \<a s3 url\>). releases are stored in a pip repo (pip install torch) (ask Soumith about this)
* pip packages - nightlies are stored on s3 (pip install -f <a s3 url>). releases are stored in a pip repo (pip install torch) (ask Soumith about this)
* conda packages - nightlies and releases are both stored in a conda repo. Nighty packages have a '_nightly' suffix
* libtorch packages - these are zips of all the c++ libraries, header files, and sometimes dependencies. These are c++ only
* shared with dependencies (the only supported option for Windows)
@ -104,16 +104,16 @@ All binaries are built in CircleCI workflows except Windows. There are checked-i
Some quick vocab:
* A \**workflow** is a CircleCI concept; it is a DAG of '**jobs**'. ctrl-f 'workflows' on https://github.com/pytorch/pytorch/blob/master/.circleci/config.yml to see the workflows.
* A\**workflow** is a CircleCI concept; it is a DAG of '**jobs**'. ctrl-f 'workflows' on\https://github.com/pytorch/pytorch/blob/master/.circleci/config.yml to see the workflows.
* **jobs** are a sequence of '**steps**'
* **steps** are usually just a bash script or a builtin CircleCI command. *All steps run in new environments, environment variables declared in one script DO NOT persist to following steps*
* **steps** are usually just a bash script or a builtin CircleCI command.* All steps run in new environments, environment variables declared in one script DO NOT persist to following steps*
* CircleCI has a **workspace**, which is essentially a cache between steps of the *same job* in which you can store artifacts between steps.
## How are the workflows structured?
The nightly binaries have 3 workflows. We have one job (actually 3 jobs: build, test, and upload) per binary configuration
1. binary_builds
1. binarybuilds
1. every day midnight EST
2. linux: https://github.com/pytorch/pytorch/blob/master/.circleci/verbatim-sources/linux-binary-build-defaults.yml
3. macos: https://github.com/pytorch/pytorch/blob/master/.circleci/verbatim-sources/macos-binary-build-defaults.yml
@ -144,7 +144,7 @@ The nightly binaries have 3 workflows. We have one job (actually 3 jobs: build,
## How are the jobs structured?
The jobs are in https://github.com/pytorch/pytorch/tree/master/.circleci/verbatim-sources. Jobs are made of multiple steps. There are some shared steps used by all the binaries/smokes. Steps of these jobs are all delegated to scripts in https://github.com/pytorch/pytorch/tree/master/.circleci/scripts .
The jobs are in https://github.com/pytorch/pytorch/tree/master/.circleci/verbatim-sources . Jobs are made of multiple steps. There are some shared steps used by all the binaries/smokes. Steps of these jobs are all delegated to scripts in https://github.com/pytorch/pytorch/tree/master/.circleci/scripts .
* Linux jobs: https://github.com/pytorch/pytorch/blob/master/.circleci/verbatim-sources/linux-binary-build-defaults.yml
* binary_linux_build.sh
@ -178,7 +178,8 @@ CircleCI creates a final yaml file by inlining every <<* segment, so if we were
So, CircleCI has several executor types: macos, machine, and docker are the ones we use. The 'machine' executor gives you two cores on some linux vm. The 'docker' executor gives you considerably more cores (nproc was 32 instead of 2 back when I tried in February). Since the dockers are faster, we try to run everything that we can in dockers. Thus
* linux build jobs use the docker executor. Running them on the docker executor was at least 2x faster than running them on the machine executor
* linux test jobs use the machine executor in order for them to properly interface with GPUs since docker executors cannot execute with attached GPUs
* linux test jobs use the machine executor and spin up their own docker. Why this nonsense? It's cause we run nvidia-docker for our GPU tests; any code that calls into the CUDA runtime needs to be run on nvidia-docker. To run a nvidia-docker you need to install some nvidia packages on the host machine and then call docker with the '—runtime nvidia' argument. CircleCI doesn't support this, so we have to do it ourself.
* This is not just a mere inconvenience. **This blocks all of our linux tests from using more than 2 cores.** But there is nothing that we can do about it, but wait for a fix on circleci's side. Right now, we only run some smoke tests (some simple imports) on the binaries, but this also affects non-binary test jobs.
* linux upload jobs use the machine executor. The upload jobs are so short that it doesn't really matter what they use
* linux smoke test jobs use the machine executor for the same reason as the linux test jobs
@ -204,7 +205,7 @@ TODO: fill in stuff
## Overview
The code that runs the binaries lives in two places, in the normal [github.com/pytorch/pytorch](http://github.com/pytorch/pytorch), but also in [github.com/pytorch/builder](http://github.com/pytorch/builder), which is a repo that defines how all the binaries are built. The relevant code is
The code that runs the binaries lives in two places, in the normal [github.com/pytorch/pytorch](http://github.com/pytorch/pytorch), but also in [github.com/pytorch/builder](http://github.com/pytorch/builder) , which is a repo that defines how all the binaries are built. The relevant code is
```
@ -260,7 +261,7 @@ Linux, MacOS and Windows use the same code flow for the conda builds.
Conda packages are built with conda-build, see https://conda.io/projects/conda-build/en/latest/resources/commands/conda-build.html
Basically, you pass `conda build` a build folder (pytorch-nightly/ above) that contains a build script and a meta.yaml. The meta.yaml specifies in what python environment to build the package in, and what dependencies the resulting package should have, and the build script gets called in the env to build the thing.
tl;dr on conda-build is
tldr; on conda-build is
1. Creates a brand new conda environment, based off of deps in the meta.yaml
1. Note that environment variables do not get passed into this build env unless they are specified in the meta.yaml
@ -270,7 +271,7 @@ tl;dr on conda-build is
4. Runs some simple import tests (if specified in the meta.yaml)
5. Saves the finished package as a tarball
The build.sh we use is essentially a wrapper around `python setup.py build`, but it also manually copies in some of our dependent libraries into the resulting tarball and messes with some rpaths.
The build.sh we use is essentially a wrapper around ```python setup.py build``` , but it also manually copies in some of our dependent libraries into the resulting tarball and messes with some rpaths.
The entrypoint file `builder/conda/build_conda.sh` is complicated because
@ -355,15 +356,15 @@ The Dockerfiles are available in pytorch/builder, but there is no circleci job o
# How to manually rebuild the binaries
tl;dr make a PR that looks like https://github.com/pytorch/pytorch/pull/21159
tldr; make a PR that looks like https://github.com/pytorch/pytorch/pull/21159
Sometimes we want to push a change to master and then rebuild all of today's binaries after that change. As of May 30, 2019 there isn't a way to manually run a workflow in the UI. You can manually re-run a workflow, but it will use the exact same git commits as the first run and will not include any changes. So we have to make a PR and then force circleci to run the binary workflow instead of the normal tests. The above PR is an example of how to do this; essentially you copy-paste the binarybuilds workflow steps into the default workflow steps. If you need to point the builder repo to a different commit then you'd need to change https://github.com/pytorch/pytorch/blob/master/.circleci/scripts/binary_checkout.sh#L42-L45 to checkout what you want.
## How to test changes to the binaries via .circleci
Writing PRs that test the binaries is annoying, since the default circleci jobs that run on PRs are not the jobs that you want to run. Likely, changes to the binaries will touch something under .circleci/ and require that .circleci/config.yml be regenerated (.circleci/config.yml controls all .circleci behavior, and is generated using `.circleci/regenerate.sh` in python 3.7). But you also need to manually hardcode the binary jobs that you want to test into the .circleci/config.yml workflow, so you should actually make at least two commits, one for your changes and one to temporarily hardcode jobs. See https://github.com/pytorch/pytorch/pull/22928 as an example of how to do this.
Writing PRs that test the binaries is annoying, since the default circleci jobs that run on PRs are not the jobs that you want to run. Likely, changes to the binaries will touch something under .circleci/ and require that .circleci/config.yml be regenerated (.circleci/config.yml controls all .circleci behavior, and is generated using ```.circleci/regenerate.sh``` in python 3.7). But you also need to manually hardcode the binary jobs that you want to test into the .circleci/config.yml workflow, so you should actually make at least two commits, one for your changes and one to temporarily hardcode jobs. See https://github.com/pytorch/pytorch/pull/22928 as an example of how to do this.
```sh
```
# Make your changes
touch .circleci/verbatim-sources/nightly-binary-build-defaults.yml
@ -408,7 +409,7 @@ The advantage of this flow is that you can make new changes to the base commit a
You can build Linux binaries locally easily using docker.
```sh
```
# Run the docker
# Use the correct docker image, pytorch/conda-cuda used here as an example
#
@ -418,6 +419,8 @@ You can build Linux binaries locally easily using docker.
# in the docker container then you will see path/to/foo/baz on your local
# machine. You could also clone the pytorch and builder repos in the docker.
#
# If you're building a CUDA binary then use `nvidia-docker run` instead, see below.
#
# If you know how, add ccache as a volume too and speed up everything
docker run \
-v your/pytorch/repo:/pytorch \
@ -441,7 +444,9 @@ export DESIRED_CUDA=cpu
**Building CUDA binaries on docker**
You can build CUDA binaries on CPU only machines, but you can only run CUDA binaries on CUDA machines. This means that you can build a CUDA binary on a docker on your laptop if you so choose (though its gonna take a long time).
To build a CUDA binary you need to use `nvidia-docker run` instead of just `docker run` (or you can manually pass `--runtime=nvidia`). This adds some needed libraries and things to build CUDA stuff.
You can build CUDA binaries on CPU only machines, but you can only run CUDA binaries on CUDA machines. This means that you can build a CUDA binary on a docker on your laptop if you so choose (though its gonna take a loong time).
For Facebook employees, ask about beefy machines that have docker support and use those instead of your laptop; it will be 5x as fast.
@ -451,7 +456,7 @@ Theres no easy way to generate reproducible hermetic MacOS environments. If y
But if you want to try, then Id recommend
```sh
```
# Create a new terminal
# Clear your LD_LIBRARY_PATH and trim as much out of your PATH as you
# know how to do

View File

@ -25,17 +25,15 @@ DEPS_INCLUSION_DIMENSIONS = [
]
def get_processor_arch_name(gpu_version):
return "cpu" if not gpu_version else (
"cu" + gpu_version.strip("cuda") if gpu_version.startswith("cuda") else gpu_version
)
def get_processor_arch_name(cuda_version):
return "cpu" if not cuda_version else "cu" + cuda_version
LINUX_PACKAGE_VARIANTS = OrderedDict(
manywheel=[
"3.6m",
"3.7m",
"3.8m",
"3.9m"
],
conda=dimensions.STANDARD_PYTHON_VERSIONS,
libtorch=[
@ -44,7 +42,7 @@ LINUX_PACKAGE_VARIANTS = OrderedDict(
)
CONFIG_TREE_DATA = OrderedDict(
linux=(dimensions.GPU_VERSIONS, LINUX_PACKAGE_VARIANTS),
linux=(dimensions.CUDA_VERSIONS, LINUX_PACKAGE_VARIANTS),
macos=([None], OrderedDict(
wheel=dimensions.STANDARD_PYTHON_VERSIONS,
conda=dimensions.STANDARD_PYTHON_VERSIONS,
@ -52,26 +50,13 @@ CONFIG_TREE_DATA = OrderedDict(
"3.7",
],
)),
macos_arm64=([None], OrderedDict(
wheel=[
"3.8",
"3.9",
],
conda=[
"3.8",
"3.9",
windows=(dimensions.CUDA_VERSIONS, OrderedDict(
wheel=dimensions.STANDARD_PYTHON_VERSIONS,
conda=dimensions.STANDARD_PYTHON_VERSIONS,
libtorch=[
"3.7",
],
)),
windows=(
[v for v in dimensions.GPU_VERSIONS if v not in dimensions.ROCM_VERSION_LABELS],
OrderedDict(
wheel=dimensions.STANDARD_PYTHON_VERSIONS,
conda=dimensions.STANDARD_PYTHON_VERSIONS,
libtorch=[
"3.7",
],
)
),
)
# GCC config variants:
@ -108,12 +93,12 @@ class TopLevelNode(ConfigNode):
class OSConfigNode(ConfigNode):
def __init__(self, parent, os_name, gpu_versions, py_tree):
def __init__(self, parent, os_name, cuda_versions, py_tree):
super(OSConfigNode, self).__init__(parent, os_name)
self.py_tree = py_tree
self.props["os_name"] = os_name
self.props["gpu_versions"] = gpu_versions
self.props["cuda_versions"] = cuda_versions
def get_children(self):
return [PackageFormatConfigNode(self, k, v) for k, v in self.py_tree.items()]
@ -132,7 +117,7 @@ class PackageFormatConfigNode(ConfigNode):
elif self.find_prop("os_name") == "windows" and self.find_prop("package_format") == "libtorch":
return [WindowsLibtorchConfigNode(self, v) for v in WINDOWS_LIBTORCH_CONFIG_VARIANTS]
else:
return [ArchConfigNode(self, v) for v in self.find_prop("gpu_versions")]
return [ArchConfigNode(self, v) for v in self.find_prop("cuda_versions")]
class LinuxGccConfigNode(ConfigNode):
@ -142,22 +127,14 @@ class LinuxGccConfigNode(ConfigNode):
self.props["gcc_config_variant"] = gcc_config_variant
def get_children(self):
gpu_versions = self.find_prop("gpu_versions")
cuda_versions = self.find_prop("cuda_versions")
# XXX devtoolset7 on CUDA 9.0 is temporarily disabled
# see https://github.com/pytorch/pytorch/issues/20066
if self.find_prop("gcc_config_variant") == 'devtoolset7':
gpu_versions = filter(lambda x: x != "cuda_90", gpu_versions)
cuda_versions = filter(lambda x: x != "90", cuda_versions)
# XXX disabling conda rocm build since docker images are not there
if self.find_prop("package_format") == 'conda':
gpu_versions = filter(lambda x: x not in dimensions.ROCM_VERSION_LABELS, gpu_versions)
# XXX libtorch rocm build is temporarily disabled
if self.find_prop("package_format") == 'libtorch':
gpu_versions = filter(lambda x: x not in dimensions.ROCM_VERSION_LABELS, gpu_versions)
return [ArchConfigNode(self, v) for v in gpu_versions]
return [ArchConfigNode(self, v) for v in cuda_versions]
class WindowsLibtorchConfigNode(ConfigNode):
@ -167,14 +144,14 @@ class WindowsLibtorchConfigNode(ConfigNode):
self.props["libtorch_config_variant"] = libtorch_config_variant
def get_children(self):
return [ArchConfigNode(self, v) for v in self.find_prop("gpu_versions")]
return [ArchConfigNode(self, v) for v in self.find_prop("cuda_versions")]
class ArchConfigNode(ConfigNode):
def __init__(self, parent, gpu):
super(ArchConfigNode, self).__init__(parent, get_processor_arch_name(gpu))
def __init__(self, parent, cu):
super(ArchConfigNode, self).__init__(parent, get_processor_arch_name(cu))
self.props["gpu"] = gpu
self.props["cu"] = cu
def get_children(self):
return [PyVersionConfigNode(self, v) for v in self.find_prop("python_versions")]

View File

@ -6,10 +6,10 @@ import cimodel.lib.conf_tree as conf_tree
import cimodel.lib.miniutils as miniutils
class Conf(object):
def __init__(self, os, gpu_version, pydistro, parms, smoke, libtorch_variant, gcc_config_variant, libtorch_config_variant):
def __init__(self, os, cuda_version, pydistro, parms, smoke, libtorch_variant, gcc_config_variant, libtorch_config_variant):
self.os = os
self.gpu_version = gpu_version
self.cuda_version = cuda_version
self.pydistro = pydistro
self.parms = parms
self.smoke = smoke
@ -18,7 +18,7 @@ class Conf(object):
self.libtorch_config_variant = libtorch_config_variant
def gen_build_env_parms(self):
elems = [self.pydistro] + self.parms + [binary_build_data.get_processor_arch_name(self.gpu_version)]
elems = [self.pydistro] + self.parms + [binary_build_data.get_processor_arch_name(self.cuda_version)]
if self.gcc_config_variant is not None:
elems.append(str(self.gcc_config_variant))
if self.libtorch_config_variant is not None:
@ -27,19 +27,7 @@ class Conf(object):
def gen_docker_image(self):
if self.gcc_config_variant == 'gcc5.4_cxx11-abi':
if self.gpu_version is None:
return miniutils.quote("pytorch/libtorch-cxx11-builder:cpu")
else:
return miniutils.quote(
f"pytorch/libtorch-cxx11-builder:{self.gpu_version}"
)
if self.pydistro == "conda":
if self.gpu_version is None:
return miniutils.quote("pytorch/conda-builder:cpu")
else:
return miniutils.quote(
f"pytorch/conda-builder:{self.gpu_version}"
)
return miniutils.quote("pytorch/pytorch-binary-docker-image-ubuntu16.04:latest")
docker_word_substitution = {
"manywheel": "manylinux",
@ -49,12 +37,9 @@ class Conf(object):
docker_distro_prefix = miniutils.override(self.pydistro, docker_word_substitution)
# The cpu nightlies are built on the pytorch/manylinux-cuda102 docker image
# TODO cuda images should consolidate into tag-base images similar to rocm
alt_docker_suffix = "cuda102" if not self.gpu_version else (
"rocm:" + self.gpu_version.strip("rocm") if self.gpu_version.startswith("rocm") else self.gpu_version)
docker_distro_suffix = alt_docker_suffix if self.pydistro != "conda" else (
"cuda" if alt_docker_suffix.startswith("cuda") else "rocm")
return miniutils.quote("pytorch/" + docker_distro_prefix + "-" + docker_distro_suffix)
alt_docker_suffix = self.cuda_version or "102"
docker_distro_suffix = "" if self.pydistro == "conda" else alt_docker_suffix
return miniutils.quote("pytorch/" + docker_distro_prefix + "-cuda" + docker_distro_suffix)
def get_name_prefix(self):
return "smoke" if self.smoke else "binary"
@ -84,10 +69,14 @@ class Conf(object):
"update_s3_htmls",
]
job_def["filters"] = branch_filters.gen_filter_dict(
branches_list=["postnightly"],
branches_list=["nightly"],
tags_list=[branch_filters.RC_PATTERN],
)
else:
filter_branch = r"/.*/"
if phase in ["upload"]:
filter_branch = "nightly"
else:
filter_branch = r"/.*/"
job_def["filters"] = branch_filters.gen_filter_dict(
branches_list=[filter_branch],
tags_list=[branch_filters.RC_PATTERN],
@ -100,61 +89,28 @@ class Conf(object):
if not (self.smoke and self.os == "macos") and self.os != "windows":
job_def["docker_image"] = self.gen_docker_image()
# fix this. only works on cuda not rocm
if self.os != "windows" and self.gpu_version:
if self.os != "windows" and self.cuda_version:
job_def["use_cuda_docker_runtime"] = miniutils.quote("1")
else:
if self.os == "linux" and phase != "upload":
job_def["docker_image"] = self.gen_docker_image()
if phase == "test":
if self.gpu_version:
if self.cuda_version:
if self.os == "windows":
job_def["executor"] = "windows-with-nvidia-gpu"
else:
job_def["resource_class"] = "gpu.medium"
if phase == "upload":
job_def["context"] = "org-member"
job_def["requires"] = [
self.gen_build_name(upload_phase_dependency, nightly)
]
os_name = miniutils.override(self.os, {"macos": "mac"})
job_name = "_".join([self.get_name_prefix(), os_name, phase])
return {job_name : job_def}
def gen_upload_job(self, phase, requires_dependency):
"""Generate binary_upload job for configuration
Output looks similar to:
- binary_upload:
name: binary_linux_manywheel_3_7m_cu92_devtoolset7_nightly_upload
context: org-member
requires: binary_linux_manywheel_3_7m_cu92_devtoolset7_nightly_test
filters:
branches:
only:
- nightly
tags:
only: /v[0-9]+(\\.[0-9]+)*-rc[0-9]+/
package_type: manywheel
upload_subfolder: cu92
"""
return {
"binary_upload": OrderedDict({
"name": self.gen_build_name(phase, nightly=True),
"context": "org-member",
"requires": [self.gen_build_name(
requires_dependency,
nightly=True
)],
"filters": branch_filters.gen_filter_dict(
branches_list=["nightly"],
tags_list=[branch_filters.RC_PATTERN],
),
"package_type": self.pydistro,
"upload_subfolder": binary_build_data.get_processor_arch_name(
self.gpu_version,
),
})
}
def get_root(smoke, name):
return binary_build_data.TopLevelNode(
@ -173,10 +129,10 @@ def gen_build_env_list(smoke):
for c in config_list:
conf = Conf(
c.find_prop("os_name"),
c.find_prop("gpu"),
c.find_prop("cu"),
c.find_prop("package_format"),
[c.find_prop("pyver")],
c.find_prop("smoke") and not (c.find_prop("os_name") == "macos_arm64"), # don't test arm64
c.find_prop("smoke"),
c.find_prop("libtorch_variant"),
c.find_prop("gcc_config_variant"),
c.find_prop("libtorch_config_variant"),
@ -193,19 +149,32 @@ def get_nightly_uploads():
mylist = []
for conf in configs:
phase_dependency = "test" if predicate_exclude_macos(conf) else "build"
mylist.append(conf.gen_upload_job("upload", phase_dependency))
mylist.append(conf.gen_workflow_job("upload", phase_dependency, nightly=True))
return mylist
def get_post_upload_jobs():
"""Generate jobs to update HTML indices and report binary sizes"""
configs = gen_build_env_list(False)
common_job_def = {
"context": "org-member",
"filters": branch_filters.gen_filter_dict(
branches_list=["nightly"],
tags_list=[branch_filters.RC_PATTERN],
),
"requires": [],
}
for conf in configs:
upload_job_name = conf.gen_build_name(
build_or_test="upload",
nightly=True
)
common_job_def["requires"].append(upload_job_name)
return [
{
"update_s3_htmls": {
"name": "update_s3_htmls",
"context": "org-member",
"filters": branch_filters.gen_filter_dict(
branches_list=["postnightly"],
),
**common_job_def,
},
},
]
@ -228,9 +197,7 @@ def get_jobs(toplevel_key, smoke):
configs = gen_build_env_list(smoke)
phase = "build" if toplevel_key == "binarybuilds" else "test"
for build_config in configs:
# don't test for macos_arm64 as it's cross compiled
if phase != "test" or build_config.os != "macos_arm64":
jobs_list.append(build_config.gen_workflow_job(phase, nightly=True))
jobs_list.append(build_config.gen_workflow_job(phase, nightly=True))
return jobs_list

View File

@ -0,0 +1,91 @@
from cimodel.lib.conf_tree import ConfigNode, XImportant
from cimodel.lib.conf_tree import Ver
CONFIG_TREE_DATA = [
(Ver("ubuntu", "16.04"), [
([Ver("clang", "7")], [XImportant("onnx_main_py3.6"),
XImportant("onnx_ort1_py3.6"),
XImportant("onnx_ort2_py3.6")]),
]),
]
class TreeConfigNode(ConfigNode):
def __init__(self, parent, node_name, subtree):
super(TreeConfigNode, self).__init__(parent, self.modify_label(node_name))
self.subtree = subtree
self.init2(node_name)
# noinspection PyMethodMayBeStatic
def modify_label(self, label):
return str(label)
def init2(self, node_name):
pass
def get_children(self):
return [self.child_constructor()(self, k, v) for (k, v) in self.subtree]
def is_build_only(self):
if str(self.find_prop("language_version")) == "onnx_main_py3.6" or \
str(self.find_prop("language_version")) == "onnx_ort1_py3.6" or \
str(self.find_prop("language_version")) == "onnx_ort2_py3.6":
return False
return set(str(c) for c in self.find_prop("compiler_version")).intersection({
"clang3.8",
"clang3.9",
"clang7",
"android",
}) or self.find_prop("distro_version").name == "macos"
def is_test_only(self):
if str(self.find_prop("language_version")) == "onnx_ort1_py3.6" or \
str(self.find_prop("language_version")) == "onnx_ort2_py3.6":
return True
return False
class TopLevelNode(TreeConfigNode):
def __init__(self, node_name, subtree):
super(TopLevelNode, self).__init__(None, node_name, subtree)
# noinspection PyMethodMayBeStatic
def child_constructor(self):
return DistroConfigNode
class DistroConfigNode(TreeConfigNode):
def init2(self, node_name):
self.props["distro_version"] = node_name
# noinspection PyMethodMayBeStatic
def child_constructor(self):
return CompilerConfigNode
class CompilerConfigNode(TreeConfigNode):
def init2(self, node_name):
self.props["compiler_version"] = node_name
# noinspection PyMethodMayBeStatic
def child_constructor(self):
return LanguageConfigNode
class LanguageConfigNode(TreeConfigNode):
def init2(self, node_name):
self.props["language_version"] = node_name
self.props["build_only"] = self.is_build_only()
self.props["test_only"] = self.is_test_only()
def child_constructor(self):
return ImportantConfigNode
class ImportantConfigNode(TreeConfigNode):
def init2(self, node_name):
self.props["important"] = True
def get_children(self):
return []

View File

@ -0,0 +1,174 @@
from collections import OrderedDict
import cimodel.data.dimensions as dimensions
import cimodel.lib.conf_tree as conf_tree
from cimodel.lib.conf_tree import Ver
import cimodel.lib.miniutils as miniutils
from cimodel.data.caffe2_build_data import CONFIG_TREE_DATA, TopLevelNode
from cimodel.data.simple.util.branch_filters import gen_filter_dict
from dataclasses import dataclass
DOCKER_IMAGE_PATH_BASE = "308535385114.dkr.ecr.us-east-1.amazonaws.com/caffe2/"
DOCKER_IMAGE_VERSION = "376"
@dataclass
class Conf:
language: str
distro: Ver
# There could be multiple compiler versions configured (e.g. nvcc
# for gpu files and host compiler (gcc/clang) for cpu files)
compilers: [Ver]
build_only: bool
test_only: bool
is_important: bool
@property
def compiler_names(self):
return [c.name for c in self.compilers]
# TODO: Eventually we can probably just remove the cudnn7 everywhere.
def get_cudnn_insertion(self):
omit = self.language == "onnx_main_py3.6" \
or self.language == "onnx_ort1_py3.6" \
or self.language == "onnx_ort2_py3.6" \
or set(self.compiler_names).intersection({"android", "mkl", "clang"}) \
or str(self.distro) in ["ubuntu14.04", "macos10.13"]
return [] if omit else ["cudnn7"]
def get_build_name_root_parts(self):
return [
"caffe2",
self.language,
] + self.get_build_name_middle_parts()
def get_build_name_middle_parts(self):
return [str(c) for c in self.compilers] + self.get_cudnn_insertion() + [str(self.distro)]
def construct_phase_name(self, phase):
root_parts = self.get_build_name_root_parts()
build_name_substitutions = {
"onnx_ort1_py3.6": "onnx_main_py3.6",
"onnx_ort2_py3.6": "onnx_main_py3.6",
}
if phase == "build":
root_parts = [miniutils.override(r, build_name_substitutions) for r in root_parts]
return "_".join(root_parts + [phase]).replace(".", "_")
def get_platform(self):
platform = self.distro.name
if self.distro.name != "macos":
platform = "linux"
return platform
def gen_docker_image(self):
lang_substitutions = {
"onnx_main_py3.6": "py3.6",
"onnx_ort1_py3.6": "py3.6",
"onnx_ort2_py3.6": "py3.6",
"cmake": "py3",
}
lang = miniutils.override(self.language, lang_substitutions)
parts = [lang] + self.get_build_name_middle_parts()
return miniutils.quote(DOCKER_IMAGE_PATH_BASE + "-".join(parts) + ":" + str(DOCKER_IMAGE_VERSION))
def gen_workflow_params(self, phase):
parameters = OrderedDict()
lang_substitutions = {
"onnx_py3": "onnx-py3",
"onnx_main_py3.6": "onnx-main-py3.6",
"onnx_ort1_py3.6": "onnx-ort1-py3.6",
"onnx_ort2_py3.6": "onnx-ort2-py3.6",
}
lang = miniutils.override(self.language, lang_substitutions)
parts = [
"caffe2",
lang,
] + self.get_build_name_middle_parts() + [phase]
build_env_name = "-".join(parts)
parameters["build_environment"] = miniutils.quote(build_env_name)
if "ios" in self.compiler_names:
parameters["build_ios"] = miniutils.quote("1")
if phase == "test":
# TODO cuda should not be considered a compiler
if "cuda" in self.compiler_names:
parameters["use_cuda_docker_runtime"] = miniutils.quote("1")
if self.distro.name != "macos":
parameters["docker_image"] = self.gen_docker_image()
if self.build_only:
parameters["build_only"] = miniutils.quote("1")
if phase == "test":
resource_class = "large" if "cuda" not in self.compiler_names else "gpu.medium"
parameters["resource_class"] = resource_class
return parameters
def gen_workflow_job(self, phase):
job_def = OrderedDict()
job_def["name"] = self.construct_phase_name(phase)
if phase == "test":
job_def["requires"] = [self.construct_phase_name("build")]
job_name = "caffe2_" + self.get_platform() + "_test"
else:
job_name = "caffe2_" + self.get_platform() + "_build"
if not self.is_important:
job_def["filters"] = gen_filter_dict()
job_def.update(self.gen_workflow_params(phase))
return {job_name : job_def}
def get_root():
return TopLevelNode("Caffe2 Builds", CONFIG_TREE_DATA)
def instantiate_configs():
config_list = []
root = get_root()
found_configs = conf_tree.dfs(root)
for fc in found_configs:
c = Conf(
language=fc.find_prop("language_version"),
distro=fc.find_prop("distro_version"),
compilers=fc.find_prop("compiler_version"),
build_only=fc.find_prop("build_only"),
test_only=fc.find_prop("test_only"),
is_important=fc.find_prop("important"),
)
config_list.append(c)
return config_list
def get_workflow_jobs():
configs = instantiate_configs()
x = []
for conf_options in configs:
phases = ["build"]
if not conf_options.build_only:
phases = dimensions.PHASES
if conf_options.test_only:
phases = ["test"]
for phase in phases:
x.append(conf_options.gen_workflow_job(phase))
return x

View File

@ -1,23 +1,14 @@
PHASES = ["build", "test"]
CUDA_VERSIONS = [
None, # cpu build
"92",
"101",
"102",
"111",
]
ROCM_VERSIONS = [
"4.0.1",
"4.1",
"4.2",
]
ROCM_VERSION_LABELS = ["rocm" + v for v in ROCM_VERSIONS]
GPU_VERSIONS = [None] + ["cuda" + v for v in CUDA_VERSIONS] + ROCM_VERSION_LABELS
STANDARD_PYTHON_VERSIONS = [
"3.6",
"3.7",
"3.8",
"3.9"
"3.8"
]

View File

@ -3,13 +3,15 @@ from cimodel.lib.conf_tree import ConfigNode, X, XImportant
CONFIG_TREE_DATA = [
("xenial", [
(None, [
X("nightly"),
]),
("gcc", [
("5.4", [ # All this subtree rebases to master and then build
XImportant("3.6"),
("3.6", [
("important", [X(True)]),
("parallel_tbb", [X(True)]),
("parallel_native", [X(True)]),
("pure_torch", [X(True)]),
]),
]),
# TODO: bring back libtorch test
@ -17,39 +19,27 @@ CONFIG_TREE_DATA = [
]),
("clang", [
("5", [
("3.6", [
("asan", [
(True, [
("shard_test", [XImportant(True)]),
]),
]),
]),
]),
("7", [
("3.6", [
("onnx", [XImportant(True)]),
]),
XImportant("3.6"), # This is actually the ASAN build
]),
]),
("cuda", [
("10.2", [
("9.2", [
X("3.6"),
("3.6", [
("shard_test", [X(True)]),
("libtorch", [
(True, [
('build_only', [X(True)]),
]),
]),
("cuda_gcc_override", [X("gcc5.4")])
])
]),
("10.1", [X("3.6")]),
("10.2", [
XImportant("3.6"),
("3.6", [
("libtorch", [XImportant(True)])
]),
]),
("11.1", [
("11.0", [
X("3.8"),
("3.8", [
("shard_test", [XImportant(True)]),
("libtorch", [
(True, [
('build_only', [X(True)]),
]),
]),
("libtorch", [X(True)])
]),
]),
]),
@ -57,41 +47,16 @@ CONFIG_TREE_DATA = [
("bionic", [
("clang", [
("9", [
("3.6", [
("noarch", [XImportant(True)]),
]),
XImportant("3.6"),
]),
("9", [
("3.6", [
("xla", [XImportant(True)]),
("vulkan", [XImportant(True)]),
]),
]),
]),
("cuda", [
("10.2", [
("3.9", [
("shard_test", [XImportant(True)]),
]),
]),
]),
("gcc", [
("9", [
("3.8", [
("coverage", [
(True, [
("shard_test", [XImportant(True)]),
]),
]),
]),
]),
]),
("rocm", [
("3.9", [
("3.6", [
('build_only', [XImportant(True)]),
]),
]),
("9", [XImportant("3.8")]),
]),
]),
]
@ -145,8 +110,6 @@ class PyVerConfigNode(TreeConfigNode):
def init2(self, node_name):
self.props["pyver"] = node_name
self.props["abbreviated_pyver"] = get_major_pyver(node_name)
if node_name == "3.9":
self.props["abbreviated_pyver"] = "py3.9"
# noinspection PyMethodMayBeStatic
def child_constructor(self):
@ -161,36 +124,17 @@ class ExperimentalFeatureConfigNode(TreeConfigNode):
experimental_feature = self.find_prop("experimental_feature")
next_nodes = {
"asan": AsanConfigNode,
"xla": XlaConfigNode,
"mlc": MLCConfigNode,
"vulkan": VulkanConfigNode,
"parallel_tbb": ParallelTBBConfigNode,
"noarch": NoarchConfigNode,
"parallel_native": ParallelNativeConfigNode,
"onnx": ONNXConfigNode,
"libtorch": LibTorchConfigNode,
"important": ImportantConfigNode,
"build_only": BuildOnlyConfigNode,
"shard_test": ShardTestConfigNode,
"cuda_gcc_override": CudaGccOverrideConfigNode,
"coverage": CoverageConfigNode,
"pure_torch": PureTorchConfigNode,
"cuda_gcc_override": CudaGccOverrideConfigNode
}
return next_nodes[experimental_feature]
class PureTorchConfigNode(TreeConfigNode):
def modify_label(self, label):
return "PURE_TORCH=" + str(label)
def init2(self, node_name):
self.props["is_pure_torch"] = node_name
def child_constructor(self):
return ImportantConfigNode
class XlaConfigNode(TreeConfigNode):
def modify_label(self, label):
return "XLA=" + str(label)
@ -201,49 +145,6 @@ class XlaConfigNode(TreeConfigNode):
def child_constructor(self):
return ImportantConfigNode
class MLCConfigNode(TreeConfigNode):
def modify_label(self, label):
return "MLC=" + str(label)
def init2(self, node_name):
self.props["is_mlc"] = node_name
def child_constructor(self):
return ImportantConfigNode
class AsanConfigNode(TreeConfigNode):
def modify_label(self, label):
return "Asan=" + str(label)
def init2(self, node_name):
self.props["is_asan"] = node_name
def child_constructor(self):
return ExperimentalFeatureConfigNode
class ONNXConfigNode(TreeConfigNode):
def modify_label(self, label):
return "Onnx=" + str(label)
def init2(self, node_name):
self.props["is_onnx"] = node_name
def child_constructor(self):
return ImportantConfigNode
class VulkanConfigNode(TreeConfigNode):
def modify_label(self, label):
return "Vulkan=" + str(label)
def init2(self, node_name):
self.props["is_vulkan"] = node_name
def child_constructor(self):
return ImportantConfigNode
class ParallelTBBConfigNode(TreeConfigNode):
def modify_label(self, label):
@ -256,14 +157,6 @@ class ParallelTBBConfigNode(TreeConfigNode):
return ImportantConfigNode
class NoarchConfigNode(TreeConfigNode):
def init2(self, node_name):
self.props["is_noarch"] = node_name
def child_constructor(self):
return ImportantConfigNode
class ParallelNativeConfigNode(TreeConfigNode):
def modify_label(self, label):
return "PARALLELNATIVE=" + str(label)
@ -283,7 +176,7 @@ class LibTorchConfigNode(TreeConfigNode):
self.props["is_libtorch"] = node_name
def child_constructor(self):
return ExperimentalFeatureConfigNode
return ImportantConfigNode
class CudaGccOverrideConfigNode(TreeConfigNode):
@ -291,33 +184,17 @@ class CudaGccOverrideConfigNode(TreeConfigNode):
self.props["cuda_gcc_override"] = node_name
def child_constructor(self):
return ExperimentalFeatureConfigNode
return ImportantConfigNode
class BuildOnlyConfigNode(TreeConfigNode):
def init2(self, node_name):
self.props["build_only"] = node_name
def child_constructor(self):
return ExperimentalFeatureConfigNode
class ShardTestConfigNode(TreeConfigNode):
def init2(self, node_name):
self.props["shard_test"] = node_name
def child_constructor(self):
return ImportantConfigNode
class CoverageConfigNode(TreeConfigNode):
def init2(self, node_name):
self.props["is_coverage"] = node_name
def child_constructor(self):
return ExperimentalFeatureConfigNode
class ImportantConfigNode(TreeConfigNode):
def modify_label(self, label):
return "IMPORTANT=" + str(label)
@ -330,6 +207,7 @@ class ImportantConfigNode(TreeConfigNode):
class XenialCompilerConfigNode(TreeConfigNode):
def modify_label(self, label):
return label or "<unspecified>"
@ -343,6 +221,7 @@ class XenialCompilerConfigNode(TreeConfigNode):
class BionicCompilerConfigNode(TreeConfigNode):
def modify_label(self, label):
return label or "<unspecified>"

View File

@ -1,13 +1,14 @@
from collections import OrderedDict
from dataclasses import dataclass, field
from typing import List, Optional
from cimodel.data.pytorch_build_data import TopLevelNode, CONFIG_TREE_DATA
import cimodel.data.dimensions as dimensions
import cimodel.lib.conf_tree as conf_tree
import cimodel.lib.miniutils as miniutils
from cimodel.data.pytorch_build_data import CONFIG_TREE_DATA, TopLevelNode
from cimodel.data.simple.util.branch_filters import gen_filter_dict, RC_PATTERN
from cimodel.data.simple.util.docker_constants import gen_docker_image
from cimodel.data.simple.util.branch_filters import gen_filter_dict
from cimodel.data.simple.util.docker_constants import gen_docker_image_path
from dataclasses import dataclass, field
from typing import List, Optional
@dataclass
@ -17,25 +18,19 @@ class Conf:
parms_list_ignored_for_docker_image: Optional[List[str]] = None
pyver: Optional[str] = None
cuda_version: Optional[str] = None
rocm_version: Optional[str] = None
# TODO expand this to cover all the USE_* that we want to test for
# tesnrorrt, leveldb, lmdb, redis, opencv, mkldnn, ideep, etc.
# (from https://github.com/pytorch/pytorch/pull/17323#discussion_r259453608)
is_xla: bool = False
is_vulkan: bool = False
is_pure_torch: bool = False
vulkan: bool = False
restrict_phases: Optional[List[str]] = None
gpu_resource: Optional[str] = None
dependent_tests: List = field(default_factory=list)
parent_build: Optional["Conf"] = None
parent_build: Optional['Conf'] = None
is_libtorch: bool = False
is_important: bool = False
parallel_backend: Optional[str] = None
@staticmethod
def is_test_phase(phase):
return "test" in phase
# TODO: Eliminate the special casing for docker paths
# In the short term, we *will* need to support special casing as docker images are merged for caffe2 and pytorch
def get_parms(self, for_docker):
@ -47,12 +42,8 @@ class Conf:
leading.append("pytorch")
if self.is_xla and not for_docker:
leading.append("xla")
if self.is_vulkan and not for_docker:
leading.append("vulkan")
if self.is_libtorch and not for_docker:
leading.append("libtorch")
if self.is_pure_torch and not for_docker:
leading.append("pure_torch")
if self.parallel_backend is not None and not for_docker:
leading.append(self.parallel_backend)
@ -60,34 +51,23 @@ class Conf:
if self.cuda_version:
cudnn = "cudnn8" if self.cuda_version.startswith("11.") else "cudnn7"
cuda_parms.extend(["cuda" + self.cuda_version, cudnn])
if self.rocm_version:
cuda_parms.extend([f"rocm{self.rocm_version}"])
result = leading + ["linux", self.distro] + cuda_parms + self.parms
if not for_docker and self.parms_list_ignored_for_docker_image is not None:
result = result + self.parms_list_ignored_for_docker_image
return result
def gen_docker_image_path(self):
parms_source = self.parent_build or self
base_build_env_name = "-".join(parms_source.get_parms(True))
image_name, _ = gen_docker_image(base_build_env_name)
return miniutils.quote(image_name)
def gen_docker_image_requires(self):
parms_source = self.parent_build or self
base_build_env_name = "-".join(parms_source.get_parms(True))
_, requires = gen_docker_image(base_build_env_name)
return miniutils.quote(requires)
return miniutils.quote(gen_docker_image_path(base_build_env_name))
def get_build_job_name_pieces(self, build_or_test):
return self.get_parms(False) + [build_or_test]
def gen_build_name(self, build_or_test):
return (
("_".join(map(str, self.get_build_job_name_pieces(build_or_test))))
.replace(".", "_")
.replace("-", "_")
)
return ("_".join(map(str, self.get_build_job_name_pieces(build_or_test)))).replace(".", "_").replace("-", "_")
def get_dependents(self):
return self.dependent_tests or []
@ -99,26 +79,20 @@ class Conf:
build_env_name = "-".join(map(str, build_job_name_pieces))
parameters["build_environment"] = miniutils.quote(build_env_name)
parameters["docker_image"] = self.gen_docker_image_path()
if Conf.is_test_phase(phase) and self.gpu_resource:
if phase == "test" and self.gpu_resource:
parameters["use_cuda_docker_runtime"] = miniutils.quote("1")
if Conf.is_test_phase(phase):
if phase == "test":
resource_class = "large"
if self.gpu_resource:
resource_class = "gpu." + self.gpu_resource
if self.rocm_version is not None:
resource_class = "pytorch/amd-gpu"
parameters["resource_class"] = resource_class
if phase == "build" and self.rocm_version is not None:
parameters["resource_class"] = "xlarge"
if hasattr(self, 'filters'):
parameters['filters'] = self.filters
return parameters
def gen_workflow_job(self, phase):
job_def = OrderedDict()
job_def["name"] = self.gen_build_name(phase)
if Conf.is_test_phase(phase):
if phase == "test":
# TODO When merging the caffe2 and pytorch jobs, it might be convenient for a while to make a
# caffe2 test job dependent on a pytorch build job. This way we could quickly dedup the repeated
@ -130,59 +104,36 @@ class Conf:
job_name = "pytorch_linux_test"
else:
job_name = "pytorch_linux_build"
job_def["requires"] = [self.gen_docker_image_requires()]
if not self.is_important:
job_def["filters"] = gen_filter_dict()
job_def.update(self.gen_workflow_params(phase))
return {job_name: job_def}
return {job_name : job_def}
# TODO This is a hack to special case some configs just for the workflow list
class HiddenConf(object):
def __init__(self, name, parent_build=None, filters=None):
def __init__(self, name, parent_build=None):
self.name = name
self.parent_build = parent_build
self.filters = filters
def gen_workflow_job(self, phase):
return {
self.gen_build_name(phase): {
"requires": [self.parent_build.gen_build_name("build")],
"filters": self.filters,
}
}
return {self.gen_build_name(phase): {"requires": [self.parent_build.gen_build_name("build")]}}
def gen_build_name(self, _):
return self.name
class DocPushConf(object):
def __init__(self, name, parent_build=None, branch="master"):
self.name = name
self.parent_build = parent_build
self.branch = branch
def gen_workflow_job(self, phase):
return {
"pytorch_doc_push": {
"name": self.name,
"branch": self.branch,
"requires": [self.parent_build],
"context": "org-member",
"filters": gen_filter_dict(branches_list=["nightly"],
tags_list=RC_PATTERN)
}
}
# TODO Convert these to graph nodes
def gen_dependent_configs(xenial_parent_config):
extra_parms = [
(["multigpu"], "large"),
(["nogpu", "NO_AVX2"], None),
(["nogpu", "NO_AVX"], None),
(["NO_AVX2"], "medium"),
(["NO_AVX", "NO_AVX2"], "medium"),
(["slow"], "medium"),
(["nogpu"], None),
]
configs = []
@ -191,12 +142,12 @@ def gen_dependent_configs(xenial_parent_config):
c = Conf(
xenial_parent_config.distro,
["py3"] + parms,
pyver=xenial_parent_config.pyver,
pyver="3.6",
cuda_version=xenial_parent_config.cuda_version,
restrict_phases=["test"],
gpu_resource=gpu,
parent_build=xenial_parent_config,
is_important=False,
is_important=xenial_parent_config.is_important,
)
configs.append(c)
@ -207,44 +158,9 @@ def gen_dependent_configs(xenial_parent_config):
def gen_docs_configs(xenial_parent_config):
configs = []
configs.append(
HiddenConf(
"pytorch_python_doc_build",
parent_build=xenial_parent_config,
filters=gen_filter_dict(branches_list=r"/.*/",
tags_list=RC_PATTERN),
)
)
configs.append(
DocPushConf(
"pytorch_python_doc_push",
parent_build="pytorch_python_doc_build",
branch="site",
)
)
for x in ["pytorch_python_doc_push", "pytorch_cpp_doc_push", "pytorch_doc_test"]:
configs.append(HiddenConf(x, parent_build=xenial_parent_config))
configs.append(
HiddenConf(
"pytorch_cpp_doc_build",
parent_build=xenial_parent_config,
filters=gen_filter_dict(branches_list=r"/.*/",
tags_list=RC_PATTERN),
)
)
configs.append(
DocPushConf(
"pytorch_cpp_doc_push",
parent_build="pytorch_cpp_doc_build",
branch="master",
)
)
configs.append(
HiddenConf(
"pytorch_doc_test",
parent_build=xenial_parent_config
)
)
return configs
@ -271,14 +187,12 @@ def instantiate_configs():
compiler_name = fc.find_prop("compiler_name")
compiler_version = fc.find_prop("compiler_version")
is_xla = fc.find_prop("is_xla") or False
is_asan = fc.find_prop("is_asan") or False
is_coverage = fc.find_prop("is_coverage") or False
is_noarch = fc.find_prop("is_noarch") or False
is_onnx = fc.find_prop("is_onnx") or False
is_pure_torch = fc.find_prop("is_pure_torch") or False
is_vulkan = fc.find_prop("is_vulkan") or False
parms_list_ignored_for_docker_image = []
vulkan = fc.find_prop("vulkan") or False
if vulkan:
parms_list_ignored_for_docker_image.append("vulkan")
python_version = None
if compiler_name == "cuda" or compiler_name == "android":
python_version = fc.find_prop("pyver")
@ -287,14 +201,9 @@ def instantiate_configs():
parms_list = ["py" + fc.find_prop("pyver")]
cuda_version = None
rocm_version = None
if compiler_name == "cuda":
cuda_version = fc.find_prop("compiler_version")
elif compiler_name == "rocm":
rocm_version = fc.find_prop("compiler_version")
restrict_phases = ["build", "test1", "test2", "caffe2_test"]
elif compiler_name == "android":
android_ndk_version = fc.find_prop("compiler_version")
# TODO: do we need clang to compile host binaries like protoc?
@ -308,23 +217,11 @@ def instantiate_configs():
gcc_version = compiler_name + (fc.find_prop("compiler_version") or "")
parms_list.append(gcc_version)
if is_asan:
parms_list.append("asan")
python_version = fc.find_prop("pyver")
parms_list[0] = fc.find_prop("abbreviated_pyver")
if is_coverage:
parms_list_ignored_for_docker_image.append("coverage")
python_version = fc.find_prop("pyver")
if is_noarch:
parms_list_ignored_for_docker_image.append("noarch")
if is_onnx:
parms_list.append("onnx")
python_version = fc.find_prop("pyver")
parms_list[0] = fc.find_prop("abbreviated_pyver")
restrict_phases = ["build", "ort_test1", "ort_test2"]
# TODO: This is a nasty special case
if gcc_version == 'clang5' and not is_xla:
parms_list.append("asan")
python_version = fc.find_prop("pyver")
parms_list[0] = fc.find_prop("abbreviated_pyver")
if cuda_version:
cuda_gcc_version = fc.find_prop("cuda_gcc_override") or "gcc7"
@ -334,12 +231,7 @@ def instantiate_configs():
is_important = fc.find_prop("is_important") or False
parallel_backend = fc.find_prop("parallel_backend") or None
build_only = fc.find_prop("build_only") or False
shard_test = fc.find_prop("shard_test") or False
# TODO: fix pure_torch python test packaging issue.
if shard_test:
restrict_phases = ["build"] if restrict_phases is None else restrict_phases
restrict_phases.extend(["test1", "test2"])
if build_only or is_pure_torch:
if build_only and restrict_phases is None:
restrict_phases = ["build"]
gpu_resource = None
@ -352,10 +244,8 @@ def instantiate_configs():
parms_list_ignored_for_docker_image,
python_version,
cuda_version,
rocm_version,
is_xla,
is_vulkan,
is_pure_torch,
vulkan,
restrict_phases,
gpu_resource,
is_libtorch=is_libtorch,
@ -365,33 +255,20 @@ def instantiate_configs():
# run docs builds on "pytorch-linux-xenial-py3.6-gcc5.4". Docs builds
# should run on a CPU-only build that runs on all PRs.
# XXX should this be updated to a more modern build? Projects are
# beginning to drop python3.6
if (
distro_name == "xenial"
and fc.find_prop("pyver") == "3.6"
and cuda_version is None
and parallel_backend is None
and not is_vulkan
and not is_pure_torch
and compiler_name == "gcc"
and fc.find_prop("compiler_version") == "5.4"
):
c.filters = gen_filter_dict(branches_list=r"/.*/",
tags_list=RC_PATTERN)
if distro_name == 'xenial' and fc.find_prop("pyver") == '3.6' \
and cuda_version is None \
and parallel_backend is None \
and compiler_name == 'gcc' \
and fc.find_prop('compiler_version') == '5.4':
c.dependent_tests = gen_docs_configs(c)
if cuda_version == "10.2" and python_version == "3.6" and not is_libtorch:
if cuda_version == "10.1" and python_version == "3.6" and not is_libtorch:
c.dependent_tests = gen_dependent_configs(c)
if (
compiler_name == "gcc"
and compiler_version == "5.4"
and not is_libtorch
and not is_vulkan
and not is_pure_torch
and parallel_backend is None
):
if (compiler_name == "gcc"
and compiler_version == "5.4"
and not is_libtorch
and parallel_backend is None):
bc_breaking_check = Conf(
"backward-compatibility-check",
[],
@ -420,7 +297,7 @@ def get_workflow_jobs():
for phase in phases:
# TODO why does this not have a test?
if Conf.is_test_phase(phase) and conf_options.cuda_version == "10":
if phase == "test" and conf_options.cuda_version == "10":
continue
x.append(conf_options.gen_workflow_job(phase))

View File

@ -1,28 +0,0 @@
from collections import OrderedDict
from cimodel.data.simple.util.branch_filters import gen_filter_dict
from cimodel.lib.miniutils import quote
CHANNELS_TO_PRUNE = ["pytorch-nightly", "pytorch-test"]
PACKAGES_TO_PRUNE = "pytorch torchvision torchaudio torchtext ignite torchcsprng"
def gen_workflow_job(channel: str):
return OrderedDict(
{
"anaconda_prune": OrderedDict(
{
"name": f"anaconda-prune-{channel}",
"context": quote("org-member"),
"packages": quote(PACKAGES_TO_PRUNE),
"channel": channel,
"filters": gen_filter_dict(branches_list=["postnightly"]),
}
)
}
)
def get_workflow_jobs():
return [gen_workflow_job(channel) for channel in CHANNELS_TO_PRUNE]

View File

@ -1,8 +1,5 @@
import cimodel.data.simple.util.branch_filters as branch_filters
from cimodel.data.simple.util.docker_constants import (
DOCKER_IMAGE_NDK, DOCKER_REQUIREMENT_NDK
)
import cimodel.lib.miniutils as miniutils
import cimodel.data.simple.util.branch_filters
from cimodel.data.simple.util.docker_constants import DOCKER_IMAGE_NDK
class AndroidJob:
@ -37,11 +34,10 @@ class AndroidJob:
"name": full_job_name,
"build_environment": "\"{}\"".format(build_env_name),
"docker_image": "\"{}\"".format(DOCKER_IMAGE_NDK),
"requires": [DOCKER_REQUIREMENT_NDK]
}
if self.is_master_only:
props_dict["filters"] = branch_filters.gen_filter_dict(branch_filters.NON_PR_BRANCH_LIST)
props_dict["filters"] = cimodel.data.simple.util.branch_filters.gen_filter_dict()
return [{self.template_name: props_dict}]
@ -51,16 +47,12 @@ class AndroidGradleJob:
job_name,
template_name,
dependencies,
is_master_only=True,
is_pr_only=False,
extra_props=tuple()):
is_master_only=True):
self.job_name = job_name
self.template_name = template_name
self.dependencies = dependencies
self.is_master_only = is_master_only
self.is_pr_only = is_pr_only
self.extra_props = dict(extra_props)
def gen_tree(self):
@ -70,11 +62,7 @@ class AndroidGradleJob:
}
if self.is_master_only:
props_dict["filters"] = branch_filters.gen_filter_dict(branch_filters.NON_PR_BRANCH_LIST)
elif self.is_pr_only:
props_dict["filters"] = branch_filters.gen_filter_dict(branch_filters.PR_BRANCH_LIST)
if self.extra_props:
props_dict.update(self.extra_props)
props_dict["filters"] = cimodel.data.simple.util.branch_filters.gen_filter_dict()
return [{self.template_name: props_dict}]
@ -84,27 +72,12 @@ WORKFLOW_DATA = [
AndroidJob(["x86_64"], "pytorch_linux_build"),
AndroidJob(["arm", "v7a"], "pytorch_linux_build"),
AndroidJob(["arm", "v8a"], "pytorch_linux_build"),
AndroidJob(["vulkan", "x86_32"], "pytorch_linux_build", is_master_only=False),
AndroidGradleJob(
"pytorch-linux-xenial-py3-clang5-android-ndk-r19c-gradle-build-x86_32",
"pytorch_android_gradle_build-x86_32",
["pytorch_linux_xenial_py3_clang5_android_ndk_r19c_x86_32_build"],
is_master_only=False,
is_pr_only=True),
AndroidGradleJob(
"pytorch-linux-xenial-py3-clang5-android-ndk-r19c-gradle-custom-build-single",
"pytorch_android_gradle_custom_build_single",
[DOCKER_REQUIREMENT_NDK],
is_master_only=False,
is_pr_only=True),
AndroidGradleJob(
"pytorch-linux-xenial-py3-clang5-android-ndk-r19c-gradle-custom-build-single-full-jit",
"pytorch_android_gradle_custom_build_single",
[DOCKER_REQUIREMENT_NDK],
is_master_only=False,
is_pr_only=True,
extra_props=tuple({
"lite_interpreter": miniutils.quote(str(int(False)))
}.items())),
is_master_only=False),
AndroidGradleJob(
"pytorch-linux-xenial-py3-clang5-android-ndk-r19c-gradle-build",
"pytorch_android_gradle_build",

View File

@ -1,7 +1,4 @@
from cimodel.data.simple.util.docker_constants import (
DOCKER_IMAGE_GCC7,
DOCKER_REQUIREMENT_GCC7
)
from cimodel.data.simple.util.docker_constants import DOCKER_IMAGE_GCC7
def gen_job_name(phase):
@ -41,10 +38,7 @@ class BazelJob:
full_job_name = gen_job_name(self.phase)
build_env_name = "-".join(build_env_parts)
extra_requires = (
[gen_job_name("build")] if self.phase == "test" else
[DOCKER_REQUIREMENT_GCC7]
)
extra_requires = [gen_job_name("build")] if self.phase == "test" else []
props_dict = {
"build_environment": build_env_name,

View File

@ -5,7 +5,7 @@ TODO: Refactor circleci/cimodel/data/binary_build_data.py to generate this file
NB: If you modify this file, you need to also modify
the binary_and_smoke_tests_on_pr variable in
pytorch-ci-hud to adjust the allowed build list
pytorch-ci-hud to adjust the list of whitelisted builds
at https://github.com/ezyang/pytorch-ci-hud/blob/master/src/BuildHistoryDisplay.js
Note:
@ -77,7 +77,7 @@ WORKFLOW_DATA = [
["libtorch", "3.7m", "cpu", "devtoolset7"],
"pytorch/manylinux-cuda102",
"binary_linux_libtorch_3_7m_cpu_devtoolset7_shared-with-deps_build",
is_master_only=True,
is_master_only=False,
has_libtorch_variant=True,
),
SmoketestJob(
@ -109,14 +109,14 @@ WORKFLOW_DATA = [
["libtorch", "3.7", "cpu", "debug"],
None,
"binary_windows_libtorch_3_7_cpu_debug_build",
is_master_only=True,
is_master_only=False,
),
SmoketestJob(
"binary_windows_build",
["libtorch", "3.7", "cpu", "release"],
None,
"binary_windows_libtorch_3_7_cpu_release_build",
is_master_only=True,
is_master_only=False,
),
SmoketestJob(
"binary_windows_build",
@ -131,7 +131,7 @@ WORKFLOW_DATA = [
["libtorch", "3.7", "cpu", "debug"],
None,
"binary_windows_libtorch_3_7_cpu_debug_test",
is_master_only=True,
is_master_only=False,
requires=["binary_windows_libtorch_3_7_cpu_debug_build"],
),
SmoketestJob(
@ -173,7 +173,7 @@ WORKFLOW_DATA = [
["libtorch", "3.7m", "cpu", "devtoolset7"],
"pytorch/manylinux-cuda102",
"binary_linux_libtorch_3_7m_cpu_devtoolset7_shared-with-deps_test",
is_master_only=True,
is_master_only=False,
requires=["binary_linux_libtorch_3_7m_cpu_devtoolset7_shared-with-deps_build"],
has_libtorch_variant=True,
),
@ -182,7 +182,7 @@ WORKFLOW_DATA = [
["libtorch", "3.7m", "cpu", "gcc5.4_cxx11-abi"],
"pytorch/pytorch-binary-docker-image-ubuntu16.04:latest",
"binary_linux_libtorch_3_7m_cpu_gcc5_4_cxx11-abi_shared-with-deps_test",
is_master_only=True,
is_master_only=False,
requires=["binary_linux_libtorch_3_7m_cpu_gcc5_4_cxx11-abi_shared-with-deps_build"],
has_libtorch_variant=True,
),

View File

@ -1,52 +1,44 @@
from collections import OrderedDict
from cimodel.lib.miniutils import quote
from cimodel.data.simple.util.branch_filters import gen_filter_dict, RC_PATTERN
# TODO: make this generated from a matrix rather than just a static list
IMAGE_NAMES = [
"pytorch-linux-bionic-cuda11.0-cudnn8-py3.6-gcc9",
"pytorch-linux-bionic-cuda11.0-cudnn8-py3.8-gcc9",
"pytorch-linux-bionic-cuda10.2-cudnn7-py3.8-gcc9",
"pytorch-linux-bionic-cuda10.2-cudnn7-py3.9-gcc7",
"pytorch-linux-bionic-py3.6-clang9",
"pytorch-linux-bionic-cuda10.2-cudnn7-py3.6-clang9",
"pytorch-linux-bionic-py3.8-gcc9",
"pytorch-linux-xenial-cuda10-cudnn7-py3-gcc7",
"pytorch-linux-xenial-cuda10.1-cudnn7-py3-gcc7",
"pytorch-linux-xenial-cuda10.2-cudnn7-py3-gcc7",
"pytorch-linux-xenial-cuda11.1-cudnn8-py3-gcc7",
"pytorch-linux-xenial-cuda11.3-cudnn8-py3-gcc7",
"pytorch-linux-xenial-cuda11.0-cudnn8-py3-gcc7",
"pytorch-linux-xenial-cuda9.2-cudnn7-py3-gcc5.4",
"pytorch-linux-xenial-cuda9.2-cudnn7-py3-gcc7",
"pytorch-linux-xenial-py3-clang5-android-ndk-r19c",
"pytorch-linux-xenial-py3-clang5-asan",
"pytorch-linux-xenial-py3-clang7-onnx",
"pytorch-linux-xenial-py3.8",
"pytorch-linux-xenial-py3.6-clang7",
"pytorch-linux-xenial-py3.6-gcc5.4", # this one is used in doc builds
"pytorch-linux-xenial-py3.6-gcc4.8",
"pytorch-linux-xenial-py3.6-gcc5.4",
"pytorch-linux-xenial-py3.6-gcc7.2",
"pytorch-linux-xenial-py3.6-gcc7",
"pytorch-linux-bionic-rocm3.9-py3.6",
"pytorch-linux-bionic-rocm4.0.1-py3.6",
"pytorch-linux-bionic-rocm4.1-py3.6",
"pytorch-linux-bionic-rocm4.2-py3.6",
"pytorch-linux-xenial-pynightly",
"pytorch-linux-xenial-rocm3.3-py3.6",
]
def get_workflow_jobs():
"""Generates a list of docker image build definitions"""
ret = []
for image_name in IMAGE_NAMES:
parameters = OrderedDict({
"name": quote(f"docker-{image_name}"),
"image_name": quote(image_name),
})
if image_name == "pytorch-linux-xenial-py3.6-gcc5.4":
# pushing documentation on tags requires CircleCI to also
# build all the dependencies on tags, including this docker image
parameters['filters'] = gen_filter_dict(branches_list=r"/.*/",
tags_list=RC_PATTERN)
ret.append(OrderedDict(
return [
OrderedDict(
{
"docker_build_job": parameters
"docker_build_job": OrderedDict(
{"name": quote(image_name), "image_name": quote(image_name)}
)
}
))
return ret
)
for image_name in IMAGE_NAMES
]

View File

@ -61,16 +61,41 @@ WORKFLOW_DATA = [
MultiPartVersion([3, 6], "py"),
MultiPartVersion([5, 4], "gcc"),
None,
["jit_legacy", "test"],
["ge_config_legacy", "test"],
["pytorch_linux_xenial_py3_6_gcc5_4_build"]),
GeConfigTestJob(
MultiPartVersion([3, 6], "py"),
MultiPartVersion([5, 4], "gcc"),
None,
["ge_config_profiling", "test"],
["pytorch_linux_xenial_py3_6_gcc5_4_build"]),
GeConfigTestJob(
MultiPartVersion([3, 6], "py"),
MultiPartVersion([5, 4], "gcc"),
None,
["ge_config_simple", "test"],
["pytorch_linux_xenial_py3_6_gcc5_4_build"],
),
GeConfigTestJob(
None,
None,
CudaVersion(10, 2),
["cudnn7", "py3", "jit_legacy", "test"],
["cudnn7", "py3", "ge_config_legacy", "test"],
["pytorch_linux_xenial_cuda10_2_cudnn7_py3_gcc7_build"],
use_cuda_docker=True,
),
# TODO Why does the build environment specify cuda10.1, while the
# job name is cuda10_2?
build_env_override="pytorch-linux-xenial-cuda10.1-cudnn7-ge_config_legacy-test"),
GeConfigTestJob(
None,
None,
CudaVersion(10, 2),
["cudnn7", "py3", "ge_config_profiling", "test"],
["pytorch_linux_xenial_cuda10_2_cudnn7_py3_gcc7_build"],
use_cuda_docker=True,
# TODO Why does the build environment specify cuda10.1, while the
# job name is cuda10_2?
build_env_override="pytorch-linux-xenial-cuda10.1-cudnn7-ge_config_profiling-test"),
]

View File

@ -1,16 +1,16 @@
from cimodel.data.simple.util.versions import MultiPartVersion
import cimodel.lib.miniutils as miniutils
XCODE_VERSION = MultiPartVersion([12, 0, 0])
IOS_VERSION = MultiPartVersion([11, 2, 1])
class ArchVariant:
def __init__(self, name, custom_build_name=""):
def __init__(self, name, is_custom=False):
self.name = name
self.custom_build_name = custom_build_name
self.is_custom = is_custom
def render(self):
extra_parts = [self.custom_build_name] if len(self.custom_build_name) > 0 else []
extra_parts = ["custom"] if self.is_custom else []
return "_".join([self.name] + extra_parts)
@ -19,15 +19,15 @@ def get_platform(arch_variant_name):
class IOSJob:
def __init__(self, xcode_version, arch_variant, is_org_member_context=True, extra_props=None):
self.xcode_version = xcode_version
def __init__(self, ios_version, arch_variant, is_org_member_context=True, extra_props=None):
self.ios_version = ios_version
self.arch_variant = arch_variant
self.is_org_member_context = is_org_member_context
self.extra_props = extra_props
def gen_name_parts(self, with_version_dots):
version_parts = self.xcode_version.render_dots_or_parts(with_version_dots)
version_parts = self.ios_version.render_dots_or_parts(with_version_dots)
build_variant_suffix = "_".join([self.arch_variant.render(), "build"])
return [
@ -61,20 +61,9 @@ class IOSJob:
WORKFLOW_DATA = [
IOSJob(XCODE_VERSION, ArchVariant("x86_64"), is_org_member_context=False, extra_props={
"lite_interpreter": miniutils.quote(str(int(True)))}),
IOSJob(XCODE_VERSION, ArchVariant("x86_64", "full_jit"), is_org_member_context=False, extra_props={
"lite_interpreter": miniutils.quote(str(int(False)))}),
IOSJob(XCODE_VERSION, ArchVariant("arm64"), extra_props={
"lite_interpreter": miniutils.quote(str(int(True)))}),
IOSJob(XCODE_VERSION, ArchVariant("arm64", "metal"), extra_props={
"use_metal": miniutils.quote(str(int(True))),
"lite_interpreter": miniutils.quote(str(int(True)))}),
IOSJob(XCODE_VERSION, ArchVariant("arm64", "full_jit"), extra_props={
"lite_interpreter": miniutils.quote(str(int(False)))}),
IOSJob(XCODE_VERSION, ArchVariant("arm64", "custom"), extra_props={
"op_list": "mobilenetv2.yaml",
"lite_interpreter": miniutils.quote(str(int(True)))}),
IOSJob(IOS_VERSION, ArchVariant("x86_64"), is_org_member_context=False),
IOSJob(IOS_VERSION, ArchVariant("arm64")),
IOSJob(IOS_VERSION, ArchVariant("arm64", True), extra_props={"op_list": "mobilenetv2.yaml"}),
]

View File

@ -1,22 +1,14 @@
class MacOsJob:
def __init__(self, os_version, is_build=False, is_test=False, extra_props=tuple()):
# extra_props is tuple type, because mutable data structures for argument defaults
# is not recommended.
def __init__(self, os_version, is_test=False):
self.os_version = os_version
self.is_build = is_build
self.is_test = is_test
self.extra_props = dict(extra_props)
def gen_tree(self):
non_phase_parts = ["pytorch", "macos", self.os_version, "py3"]
extra_name_list = [name for name, exist in self.extra_props.items() if exist]
full_job_name_list = non_phase_parts + extra_name_list + [
'build' if self.is_build else None,
'test' if self.is_test else None,
]
phase_name = "test" if self.is_test else "build"
full_job_name = "_".join(list(filter(None, full_job_name_list)))
full_job_name = "_".join(non_phase_parts + [phase_name])
test_build_dependency = "_".join(non_phase_parts + ["build"])
extra_dependencies = [test_build_dependency] if self.is_test else []
@ -29,23 +21,7 @@ class MacOsJob:
return [{full_job_name: props_dict}]
WORKFLOW_DATA = [
MacOsJob("10_15", is_build=True),
MacOsJob("10_13", is_build=True),
MacOsJob(
"10_13",
is_build=False,
is_test=True,
),
MacOsJob(
"10_13",
is_build=True,
is_test=True,
extra_props=tuple({
"lite_interpreter": True
}.items()),
)
]
WORKFLOW_DATA = [MacOsJob("10_13"), MacOsJob("10_13", True)]
def get_workflow_jobs():

View File

@ -4,23 +4,12 @@ PyTorch Mobile PR builds (use linux host toolchain + mobile build options)
import cimodel.lib.miniutils as miniutils
import cimodel.data.simple.util.branch_filters
from cimodel.data.simple.util.docker_constants import (
DOCKER_IMAGE_ASAN,
DOCKER_REQUIREMENT_ASAN,
DOCKER_IMAGE_NDK,
DOCKER_REQUIREMENT_NDK
)
from cimodel.data.simple.util.docker_constants import DOCKER_IMAGE_ASAN, DOCKER_IMAGE_NDK
class MobileJob:
def __init__(
self,
docker_image,
docker_requires,
variant_parts,
is_master_only=False):
def __init__(self, docker_image, variant_parts, is_master_only=False):
self.docker_image = docker_image
self.docker_requires = docker_requires
self.variant_parts = variant_parts
self.is_master_only = is_master_only
@ -41,7 +30,6 @@ class MobileJob:
"build_environment": build_env_name,
"build_only": miniutils.quote(str(int(True))),
"docker_image": self.docker_image,
"requires": self.docker_requires,
"name": full_job_name,
}
@ -52,33 +40,15 @@ class MobileJob:
WORKFLOW_DATA = [
MobileJob(
DOCKER_IMAGE_ASAN,
[DOCKER_REQUIREMENT_ASAN],
["build"]
),
MobileJob(DOCKER_IMAGE_ASAN, ["build"]),
MobileJob(DOCKER_IMAGE_ASAN, ["custom", "build", "static"]),
# Use LLVM-DEV toolchain in android-ndk-r19c docker image
MobileJob(
DOCKER_IMAGE_NDK,
[DOCKER_REQUIREMENT_NDK],
["custom", "build", "dynamic"]
),
MobileJob(
DOCKER_IMAGE_NDK,
[DOCKER_REQUIREMENT_NDK],
["custom", "build", "static"]
),
MobileJob(DOCKER_IMAGE_NDK, ["custom", "build", "dynamic"]),
# Use LLVM-DEV toolchain in android-ndk-r19c docker image
# Most of this CI is already covered by "mobile-custom-build-dynamic" job
MobileJob(
DOCKER_IMAGE_NDK,
[DOCKER_REQUIREMENT_NDK],
["code", "analysis"],
True
),
MobileJob(DOCKER_IMAGE_NDK, ["code", "analysis"], True),
]

View File

@ -1,7 +1,4 @@
from cimodel.data.simple.util.docker_constants import (
DOCKER_IMAGE_NDK,
DOCKER_REQUIREMENT_NDK
)
from cimodel.data.simple.util.docker_constants import DOCKER_IMAGE_NDK
class AndroidNightlyJob:
@ -51,13 +48,12 @@ class AndroidNightlyJob:
return [{self.template_name: props_dict}]
BASE_REQUIRES = [DOCKER_REQUIREMENT_NDK]
WORKFLOW_DATA = [
AndroidNightlyJob(["x86_32"], "pytorch_linux_build", requires=BASE_REQUIRES),
AndroidNightlyJob(["x86_64"], "pytorch_linux_build", requires=BASE_REQUIRES),
AndroidNightlyJob(["arm", "v7a"], "pytorch_linux_build", requires=BASE_REQUIRES),
AndroidNightlyJob(["arm", "v8a"], "pytorch_linux_build", requires=BASE_REQUIRES),
AndroidNightlyJob(["x86_32"], "pytorch_linux_build"),
AndroidNightlyJob(["x86_64"], "pytorch_linux_build"),
AndroidNightlyJob(["arm", "v7a"], "pytorch_linux_build"),
AndroidNightlyJob(["arm", "v8a"], "pytorch_linux_build"),
AndroidNightlyJob(["android_gradle"], "pytorch_android_gradle_build",
with_docker=False,
requires=[

View File

@ -18,7 +18,7 @@ class IOSNightlyJob:
common_name_pieces = [
"ios",
] + ios_definitions.XCODE_VERSION.render_dots_or_parts(with_version_dots) + [
] + ios_definitions.IOS_VERSION.render_dots_or_parts(with_version_dots) + [
"nightly",
self.variant,
"build",

View File

@ -4,11 +4,6 @@ NON_PR_BRANCH_LIST = [
r"/release\/.*/",
]
PR_BRANCH_LIST = [
r"/gh\/.*\/head/",
r"/pull\/.*/",
]
RC_PATTERN = r"/v[0-9]+(\.[0-9]+)*-rc[0-9]+/"
def gen_filter_dict(

View File

@ -1,33 +1,30 @@
AWS_DOCKER_HOST = "308535385114.dkr.ecr.us-east-1.amazonaws.com"
def gen_docker_image(container_type):
return (
"/".join([AWS_DOCKER_HOST, "pytorch", container_type]),
f"docker-{container_type}",
)
def gen_docker_image_requires(image_name):
return [f"docker-{image_name}"]
# ARE YOU EDITING THIS NUMBER? MAKE SURE YOU READ THE GUIDANCE AT THE
# TOP OF .circleci/config.yml
DOCKER_IMAGE_TAG = "209062ef-ab58-422a-b295-36c4eed6e906"
DOCKER_IMAGE_BASIC, DOCKER_REQUIREMENT_BASE = gen_docker_image(
"pytorch-linux-xenial-py3.6-gcc5.4"
)
DOCKER_IMAGE_CUDA_10_2, DOCKER_REQUIREMENT_CUDA_10_2 = gen_docker_image(
"pytorch-linux-xenial-cuda10.2-cudnn7-py3-gcc7"
)
DOCKER_IMAGE_GCC7, DOCKER_REQUIREMENT_GCC7 = gen_docker_image(
"pytorch-linux-xenial-py3.6-gcc7"
)
def gen_docker_image_path(container_type):
return "/".join([
AWS_DOCKER_HOST,
"pytorch",
container_type + ":" + DOCKER_IMAGE_TAG,
])
def gen_mobile_docker(specifier):
DOCKER_IMAGE_BASIC = gen_docker_image_path("pytorch-linux-xenial-py3.6-gcc5.4")
DOCKER_IMAGE_CUDA_10_2 = gen_docker_image_path("pytorch-linux-xenial-cuda10.2-cudnn7-py3-gcc7")
DOCKER_IMAGE_GCC7 = gen_docker_image_path("pytorch-linux-xenial-py3.6-gcc7")
def gen_mobile_docker_name(specifier):
container_type = "pytorch-linux-xenial-py3-clang5-" + specifier
return gen_docker_image(container_type)
return gen_docker_image_path(container_type)
DOCKER_IMAGE_ASAN, DOCKER_REQUIREMENT_ASAN = gen_mobile_docker("asan")
DOCKER_IMAGE_ASAN = gen_mobile_docker_name("asan")
DOCKER_IMAGE_NDK, DOCKER_REQUIREMENT_NDK = gen_mobile_docker("android-ndk-r19c")
DOCKER_IMAGE_NDK = gen_mobile_docker_name("android-ndk-r19c")

View File

@ -9,7 +9,7 @@ class MultiPartVersion:
with the prefix string.
"""
if self.parts:
return [self.prefix + str(self.parts[0])] + [str(part) for part in self.parts[1:]]
return [self.prefix + str(self.parts[0])] + list(map(str, self.parts[1:]))
else:
return [self.prefix]
@ -29,6 +29,3 @@ class CudaVersion(MultiPartVersion):
self.minor = minor
super().__init__([self.major, self.minor], "cuda")
def __str__(self):
return f"{self.major}.{self.minor}"

View File

@ -1,5 +1,5 @@
import cimodel.data.simple.util.branch_filters
import cimodel.lib.miniutils as miniutils
from cimodel.data.simple.util.branch_filters import gen_filter_dict, RC_PATTERN, NON_PR_BRANCH_LIST
from cimodel.data.simple.util.versions import CudaVersion
@ -10,19 +10,13 @@ class WindowsJob:
vscode_spec,
cuda_version,
force_on_cpu=False,
multi_gpu=False,
master_only=False,
nightly_only=False,
master_and_nightly=False
master_only_pred=lambda job: job.vscode_spec.year != 2019,
):
self.test_index = test_index
self.vscode_spec = vscode_spec
self.cuda_version = cuda_version
self.force_on_cpu = force_on_cpu
self.multi_gpu = multi_gpu
self.master_only = master_only
self.nightly_only = nightly_only
self.master_and_nightly = master_and_nightly
self.master_only_pred = master_only_pred
def gen_tree(self):
@ -31,10 +25,7 @@ class WindowsJob:
base_phase if self.test_index is None else base_phase + str(self.test_index)
)
key_parts = ["pytorch", "windows", base_phase]
if self.multi_gpu:
key_parts.append('multigpu')
key_name = "_".join(key_parts)
key_name = "_".join(["pytorch", "windows", base_phase])
cpu_forcing_name_parts = ["on", "cpu"] if self.force_on_cpu else []
@ -52,11 +43,8 @@ class WindowsJob:
if base_phase == "test":
prerequisite_jobs.append("_".join(base_name_parts + ["build"]))
if self.cuda_version:
self.cudnn_version = 8 if self.cuda_version.major == 11 else 7
arch_env_elements = (
["cuda" + str(self.cuda_version.major), "cudnn" + str(self.cudnn_version)]
["cuda" + str(self.cuda_version.major), "cudnn7"]
if self.cuda_version
else ["cpu"]
)
@ -70,66 +58,50 @@ class WindowsJob:
is_running_on_cuda = bool(self.cuda_version) and not self.force_on_cpu
if self.multi_gpu:
props_dict = {"requires": prerequisite_jobs}
else:
props_dict = {
"build_environment": build_environment_string,
"python_version": miniutils.quote("3.6"),
"vc_version": miniutils.quote(self.vscode_spec.dotted_version()),
"vc_year": miniutils.quote(str(self.vscode_spec.year)),
"vc_product": self.vscode_spec.get_product(),
"use_cuda": miniutils.quote(str(int(is_running_on_cuda))),
"requires": prerequisite_jobs,
}
props_dict = {
"build_environment": build_environment_string,
"python_version": miniutils.quote("3.6"),
"vc_version": miniutils.quote(self.vscode_spec.dotted_version()),
"vc_year": miniutils.quote(str(self.vscode_spec.year)),
"vc_product": self.vscode_spec.get_product(),
"use_cuda": miniutils.quote(str(int(is_running_on_cuda))),
"requires": prerequisite_jobs,
}
if self.master_only:
if self.master_only_pred(self):
props_dict[
"filters"
] = gen_filter_dict()
elif self.nightly_only:
props_dict[
"filters"
] = gen_filter_dict(branches_list=["nightly"], tags_list=RC_PATTERN)
elif self.master_and_nightly:
props_dict[
"filters"
] = gen_filter_dict(branches_list=NON_PR_BRANCH_LIST + ["nightly"], tags_list=RC_PATTERN)
] = cimodel.data.simple.util.branch_filters.gen_filter_dict()
name_parts = base_name_parts + cpu_forcing_name_parts + [numbered_phase]
if not self.multi_gpu:
if base_phase == "test":
test_name = "-".join(["pytorch", "windows", numbered_phase])
props_dict["test_name"] = test_name
if base_phase == "test":
test_name = "-".join(["pytorch", "windows", numbered_phase])
props_dict["test_name"] = test_name
if is_running_on_cuda:
props_dict["executor"] = "windows-with-nvidia-gpu"
props_dict["cuda_version"] = (
miniutils.quote(str(self.cuda_version))
if self.cuda_version
else "cpu"
)
if is_running_on_cuda:
props_dict["executor"] = "windows-with-nvidia-gpu"
props_dict["cuda_version"] = (
miniutils.quote(str(self.cuda_version.major))
if self.cuda_version
else "cpu"
)
props_dict["name"] = "_".join(name_parts)
return [{key_name: props_dict}]
class VcSpec:
def __init__(self, year, version_elements=None, hide_version=False):
def __init__(self, year, version_elements=None):
self.year = year
self.version_elements = version_elements or []
self.hide_version = hide_version
def get_elements(self):
if self.hide_version:
return [self.prefixed_year()]
return [self.prefixed_year()] + self.version_elements
def get_product(self):
return "BuildTools"
return "Community" if self.year == 2019 else "BuildTools"
def dotted_version(self):
return ".".join(self.version_elements)
@ -138,25 +110,31 @@ class VcSpec:
return "vs" + str(self.year)
def render(self):
return "_".join(self.get_elements())
return "_".join(filter(None, [self.prefixed_year(), self.dotted_version()]))
_VC2019 = VcSpec(2019)
def FalsePred(_):
return False
def TruePred(_):
return True
WORKFLOW_DATA = [
# VS2017 CUDA-10.1
WindowsJob(None, VcSpec(2017, ["14", "11"]), CudaVersion(10, 1), master_only_pred=FalsePred),
WindowsJob(1, VcSpec(2017, ["14", "11"]), CudaVersion(10, 1)),
# VS2017 no-CUDA (builds only)
WindowsJob(None, VcSpec(2017, ["14", "16"]), CudaVersion(10, 1)),
WindowsJob(None, VcSpec(2017, ["14", "16"]), None),
# VS2019 CUDA-10.1
WindowsJob(None, _VC2019, CudaVersion(10, 1), master_only=True),
WindowsJob(1, _VC2019, CudaVersion(10, 1), master_only=True),
WindowsJob(2, _VC2019, CudaVersion(10, 1), master_only=True),
# VS2019 CUDA-11.1
WindowsJob(None, _VC2019, CudaVersion(11, 1)),
WindowsJob(1, _VC2019, CudaVersion(11, 1), master_only=True),
WindowsJob(2, _VC2019, CudaVersion(11, 1), master_only=True),
WindowsJob('_azure_multi_gpu', _VC2019, CudaVersion(11, 1), multi_gpu=True, nightly_only=True),
WindowsJob(None, VcSpec(2019), CudaVersion(10, 1)),
WindowsJob(1, VcSpec(2019), CudaVersion(10, 1)),
WindowsJob(2, VcSpec(2019), CudaVersion(10, 1)),
# VS2019 CPU-only
WindowsJob(None, _VC2019, None),
WindowsJob(1, _VC2019, None),
WindowsJob(2, _VC2019, None),
WindowsJob(1, _VC2019, CudaVersion(10, 1), force_on_cpu=True, master_only=True),
WindowsJob(None, VcSpec(2019), None),
WindowsJob(1, VcSpec(2019), None),
WindowsJob(2, VcSpec(2019), None, master_only_pred=TruePred),
WindowsJob(1, VcSpec(2019), CudaVersion(10, 1), force_on_cpu=True),
WindowsJob(2, VcSpec(2019), CudaVersion(10, 1), force_on_cpu=True, master_only_pred=TruePred),
]

File diff suppressed because it is too large Load Diff

View File

@ -12,20 +12,8 @@ each image as the `BUILD_ENVIRONMENT` environment variable.
See `build.sh` for valid build environments (it's the giant switch).
Docker builds are now defined with `.circleci/cimodel/data/simple/docker_definitions.py`
## Contents
* `build.sh` -- dispatch script to launch all builds
* `common` -- scripts used to execute individual Docker build stages
* `ubuntu-cuda` -- Dockerfile for Ubuntu image with CUDA support for nvidia-docker
## Usage
```bash
# Build a specific image
./build.sh pytorch-linux-bionic-py3.8-gcc9 -t myimage:latest
# Set flags (see build.sh) and build image
sudo bash -c 'BREAKPAD=1 ./build.sh pytorch-linux-bionic-py3.8-gcc9 -t myimage:latest
```

View File

@ -20,8 +20,10 @@ buildscript {
}
dependencies {
classpath 'com.android.tools.build:gradle:4.1.2'
classpath 'com.vanniktech:gradle-maven-publish-plugin:0.14.2'
classpath 'com.android.tools.build:gradle:3.3.2'
classpath "com.jfrog.bintray.gradle:gradle-bintray-plugin:1.8.0"
classpath "com.github.dcendents:android-maven-gradle-plugin:2.1"
classpath "org.jfrog.buildinfo:build-info-extractor-gradle:4.9.8"
}
}

View File

@ -10,37 +10,18 @@ if [ -z "${image}" ]; then
exit 1
fi
function extract_version_from_image_name() {
eval export $2=$(echo "${image}" | perl -n -e"/$1(\d+(\.\d+)?(\.\d+)?)/ && print \$1")
if [ "x${!2}" = x ]; then
echo "variable '$2' not correctly parsed from image='$image'"
exit 1
fi
}
# TODO: Generalize
OS="ubuntu"
DOCKERFILE="${OS}/Dockerfile"
if [[ "$image" == *-cuda* ]]; then
DOCKERFILE="${OS}-cuda/Dockerfile"
elif [[ "$image" == *-rocm* ]]; then
DOCKERFILE="${OS}-rocm/Dockerfile"
fi
function extract_all_from_image_name() {
# parts $image into array, splitting on '-'
keep_IFS="$IFS"
IFS="-"
declare -a parts=($image)
IFS="$keep_IFS"
unset keep_IFS
for part in "${parts[@]}"; do
name=$(echo "${part}" | perl -n -e"/([a-zA-Z]+)\d+(\.\d+)?(\.\d+)?/ && print \$1")
vername="${name^^}_VERSION"
# "py" is the odd one out, needs this special case
if [ "x${name}" = xpy ]; then
vername=ANACONDA_PYTHON_VERSION
fi
# skip non-conforming fields such as "pytorch", "linux" or "xenial" without version string
if [ -n "${name}" ]; then
extract_version_from_image_name "${name}" "${vername}"
fi
done
}
if [[ "$image" == *-xenial* ]]; then
if [[ "$image" == *-trusty* ]]; then
UBUNTU_VERSION=14.04
elif [[ "$image" == *-xenial* ]]; then
UBUNTU_VERSION=16.04
elif [[ "$image" == *-artful* ]]; then
UBUNTU_VERSION=17.10
@ -48,26 +29,6 @@ elif [[ "$image" == *-bionic* ]]; then
UBUNTU_VERSION=18.04
elif [[ "$image" == *-focal* ]]; then
UBUNTU_VERSION=20.04
elif [[ "$image" == *ubuntu* ]]; then
extract_version_from_image_name ubuntu UBUNTU_VERSION
elif [[ "$image" == *centos* ]]; then
extract_version_from_image_name centos CENTOS_VERSION
fi
if [ -n "${UBUNTU_VERSION}" ]; then
OS="ubuntu"
elif [ -n "${CENTOS_VERSION}" ]; then
OS="centos"
else
echo "Unable to derive operating system base..."
exit 1
fi
DOCKERFILE="${OS}/Dockerfile"
if [[ "$image" == *cuda* ]]; then
DOCKERFILE="${OS}-cuda/Dockerfile"
elif [[ "$image" == *rocm* ]]; then
DOCKERFILE="${OS}-rocm/Dockerfile"
fi
TRAVIS_DL_URL_PREFIX="https://s3.amazonaws.com/travis-python-archives/binaries/ubuntu/14.04/x86_64"
@ -77,10 +38,19 @@ TRAVIS_DL_URL_PREFIX="https://s3.amazonaws.com/travis-python-archives/binaries/u
# from scratch
case "$image" in
pytorch-linux-xenial-py3.8)
ANACONDA_PYTHON_VERSION=3.8
# TODO: This is a hack, get rid of this as soon as you get rid of the travis downloads
TRAVIS_DL_URL_PREFIX="https://s3.amazonaws.com/travis-python-archives/binaries/ubuntu/16.04/x86_64"
TRAVIS_PYTHON_VERSION=3.8
GCC_VERSION=7
# Do not install PROTOBUF, DB, and VISION as a test
;;
pytorch-linux-xenial-py3.6-gcc4.8)
ANACONDA_PYTHON_VERSION=3.6
GCC_VERSION=4.8
PROTOBUF=yes
DB=yes
VISION=yes
;;
pytorch-linux-xenial-py3.6-gcc5.4)
ANACONDA_PYTHON_VERSION=3.6
GCC_VERSION=5
@ -88,7 +58,6 @@ case "$image" in
DB=yes
VISION=yes
KATEX=yes
BREAKPAD=yes
;;
pytorch-linux-xenial-py3.6-gcc7.2)
ANACONDA_PYTHON_VERSION=3.6
@ -101,7 +70,31 @@ case "$image" in
PROTOBUF=yes
DB=yes
VISION=yes
BREAKPAD=yes
;;
pytorch-linux-xenial-pynightly)
TRAVIS_PYTHON_VERSION=nightly
GCC_VERSION=7
PROTOBUF=yes
DB=yes
VISION=yes
;;
pytorch-linux-xenial-cuda9.2-cudnn7-py3-gcc5.4)
CUDA_VERSION=9.2
CUDNN_VERSION=7
ANACONDA_PYTHON_VERSION=3.6
GCC_VERSION=5
PROTOBUF=yes
DB=yes
VISION=yes
;;
pytorch-linux-xenial-cuda9.2-cudnn7-py3-gcc7)
CUDA_VERSION=9.2
CUDNN_VERSION=7
ANACONDA_PYTHON_VERSION=3.6
GCC_VERSION=7
PROTOBUF=yes
DB=yes
VISION=yes
;;
pytorch-linux-xenial-cuda10-cudnn7-py3-gcc7)
CUDA_VERSION=10.0
@ -111,7 +104,6 @@ case "$image" in
PROTOBUF=yes
DB=yes
VISION=yes
BREAKPAD=yes
;;
pytorch-linux-xenial-cuda10.1-cudnn7-py3-gcc7)
CUDA_VERSION=10.1
@ -122,7 +114,6 @@ case "$image" in
DB=yes
VISION=yes
KATEX=yes
BREAKPAD=yes
;;
pytorch-linux-xenial-cuda10.2-cudnn7-py3-gcc7)
CUDA_VERSION=10.2
@ -133,10 +124,10 @@ case "$image" in
DB=yes
VISION=yes
KATEX=yes
BREAKPAD=yes
;;
pytorch-linux-xenial-cuda11.1-cudnn8-py3-gcc7)
CUDA_VERSION=11.1
pytorch-linux-xenial-cuda11.0-cudnn8-py3-gcc7)
UBUNTU_VERSION=16.04-rc
CUDA_VERSION=11.0
CUDNN_VERSION=8
ANACONDA_PYTHON_VERSION=3.6
GCC_VERSION=7
@ -144,18 +135,6 @@ case "$image" in
DB=yes
VISION=yes
KATEX=yes
BREAKPAD=yes
;;
pytorch-linux-xenial-cuda11.3-cudnn8-py3-gcc7)
CUDA_VERSION=11.3.0 # Deviating from major.minor to conform to nvidia's Docker image names
CUDNN_VERSION=8
ANACONDA_PYTHON_VERSION=3.6
GCC_VERSION=7
PROTOBUF=yes
DB=yes
VISION=yes
KATEX=yes
BREAKPAD=yes
;;
pytorch-linux-xenial-py3-clang5-asan)
ANACONDA_PYTHON_VERSION=3.6
@ -163,15 +142,6 @@ case "$image" in
PROTOBUF=yes
DB=yes
VISION=yes
BREAKPAD=yes
;;
pytorch-linux-xenial-py3-clang7-onnx)
ANACONDA_PYTHON_VERSION=3.6
CLANG_VERSION=7
PROTOBUF=yes
DB=yes
VISION=yes
BREAKPAD=yes
;;
pytorch-linux-xenial-py3-clang5-android-ndk-r19c)
ANACONDA_PYTHON_VERSION=3.6
@ -180,7 +150,7 @@ case "$image" in
PROTOBUF=yes
ANDROID=yes
ANDROID_NDK_VERSION=r19c
GRADLE_VERSION=6.8.3
GRADLE_VERSION=4.10.3
CMAKE_VERSION=3.7.0
NINJA_VERSION=1.9.0
;;
@ -190,7 +160,6 @@ case "$image" in
PROTOBUF=yes
DB=yes
VISION=yes
BREAKPAD=yes
;;
pytorch-linux-bionic-py3.6-clang9)
ANACONDA_PYTHON_VERSION=3.6
@ -198,9 +167,6 @@ case "$image" in
PROTOBUF=yes
DB=yes
VISION=yes
BREAKPAD=yes
VULKAN_SDK_VERSION=1.2.162.1
SWIFTSHADER=yes
;;
pytorch-linux-bionic-py3.8-gcc9)
ANACONDA_PYTHON_VERSION=3.8
@ -208,8 +174,6 @@ case "$image" in
PROTOBUF=yes
DB=yes
VISION=yes
BREAKPAD=yes
BREAKPAD=yes
;;
pytorch-linux-bionic-cuda10.2-cudnn7-py3.6-clang9)
CUDA_VERSION=10.2
@ -219,7 +183,6 @@ case "$image" in
PROTOBUF=yes
DB=yes
VISION=yes
BREAKPAD=yes
;;
pytorch-linux-bionic-cuda10.2-cudnn7-py3.8-gcc9)
CUDA_VERSION=10.2
@ -229,19 +192,9 @@ case "$image" in
PROTOBUF=yes
DB=yes
VISION=yes
BREAKPAD=yes
;;
pytorch-linux-bionic-cuda10.2-cudnn7-py3.9-gcc7)
CUDA_VERSION=10.2
CUDNN_VERSION=7
ANACONDA_PYTHON_VERSION=3.9
GCC_VERSION=7
PROTOBUF=yes
DB=yes
VISION=yes
BREAKPAD=yes
;;
pytorch-linux-bionic-cuda11.0-cudnn8-py3.6-gcc9)
UBUNTU_VERSION=18.04-rc
CUDA_VERSION=11.0
CUDNN_VERSION=8
ANACONDA_PYTHON_VERSION=3.6
@ -249,69 +202,35 @@ case "$image" in
PROTOBUF=yes
DB=yes
VISION=yes
BREAKPAD=yes
ROCM_VERSION=3.9
KATEX=yes
;;
pytorch-linux-bionic-rocm4.0.1-py3.6)
ANACONDA_PYTHON_VERSION=3.6
pytorch-linux-bionic-cuda11.0-cudnn8-py3.8-gcc9)
UBUNTU_VERSION=18.04-rc
CUDA_VERSION=11.0
CUDNN_VERSION=8
ANACONDA_PYTHON_VERSION=3.8
GCC_VERSION=9
PROTOBUF=yes
DB=yes
VISION=yes
BREAKPAD=yes
ROCM_VERSION=4.0.1
KATEX=yes
;;
pytorch-linux-bionic-rocm4.1-py3.6)
pytorch-linux-xenial-rocm3.3-py3.6)
ANACONDA_PYTHON_VERSION=3.6
GCC_VERSION=9
PROTOBUF=yes
DB=yes
VISION=yes
BREAKPAD=yes
ROCM_VERSION=4.1
ROCM_VERSION=3.3
# newer cmake version required
CMAKE_VERSION=3.6.3
;;
pytorch-linux-bionic-rocm4.2-py3.6)
pytorch-linux-bionic-rocm3.3-py3.6)
ANACONDA_PYTHON_VERSION=3.6
GCC_VERSION=9
PROTOBUF=yes
DB=yes
VISION=yes
BREAKPAD=yes
ROCM_VERSION=4.2
ROCM_VERSION=3.3
;;
*)
# Catch-all for builds that are not hardcoded.
PROTOBUF=yes
DB=yes
VISION=yes
BREAKPAD=yes
echo "image '$image' did not match an existing build configuration"
if [[ "$image" == *py* ]]; then
extract_version_from_image_name py ANACONDA_PYTHON_VERSION
fi
if [[ "$image" == *cuda* ]]; then
extract_version_from_image_name cuda CUDA_VERSION
extract_version_from_image_name cudnn CUDNN_VERSION
fi
if [[ "$image" == *rocm* ]]; then
extract_version_from_image_name rocm ROCM_VERSION
fi
if [[ "$image" == *gcc* ]]; then
extract_version_from_image_name gcc GCC_VERSION
fi
if [[ "$image" == *clang* ]]; then
extract_version_from_image_name clang CLANG_VERSION
fi
if [[ "$image" == *devtoolset* ]]; then
extract_version_from_image_name devtoolset DEVTOOLSET_VERSION
fi
if [[ "$image" == *glibc* ]]; then
extract_version_from_image_name glibc GLIBC_VERSION
fi
if [[ "$image" == *cmake* ]]; then
extract_version_from_image_name cmake CMAKE_VERSION
fi
;;
esac
# Set Jenkins UID and GID if running Jenkins
@ -320,7 +239,7 @@ if [ -n "${JENKINS:-}" ]; then
JENKINS_GID=$(id -g jenkins)
fi
tmp_tag="tmp-$(cat /dev/urandom | tr -dc 'a-z' | head -c 32)"
tmp_tag="tmp-$(cat /dev/urandom | tr -dc 'a-z' | fold -w 32 | head -n 1)"
# Build image
# TODO: build-arg THRIFT is not turned on for any image, remove it once we confirm
@ -340,20 +259,15 @@ docker build \
--build-arg "JENKINS_UID=${JENKINS_UID:-}" \
--build-arg "JENKINS_GID=${JENKINS_GID:-}" \
--build-arg "UBUNTU_VERSION=${UBUNTU_VERSION}" \
--build-arg "CENTOS_VERSION=${CENTOS_VERSION}" \
--build-arg "DEVTOOLSET_VERSION=${DEVTOOLSET_VERSION}" \
--build-arg "GLIBC_VERSION=${GLIBC_VERSION}" \
--build-arg "CLANG_VERSION=${CLANG_VERSION}" \
--build-arg "ANACONDA_PYTHON_VERSION=${ANACONDA_PYTHON_VERSION}" \
--build-arg "TRAVIS_PYTHON_VERSION=${TRAVIS_PYTHON_VERSION}" \
--build-arg "GCC_VERSION=${GCC_VERSION}" \
--build-arg "CUDA_VERSION=${CUDA_VERSION}" \
--build-arg "CUDNN_VERSION=${CUDNN_VERSION}" \
--build-arg "BREAKPAD=${BREAKPAD}" \
--build-arg "ANDROID=${ANDROID}" \
--build-arg "ANDROID_NDK=${ANDROID_NDK_VERSION}" \
--build-arg "GRADLE_VERSION=${GRADLE_VERSION}" \
--build-arg "VULKAN_SDK_VERSION=${VULKAN_SDK_VERSION}" \
--build-arg "SWIFTSHADER=${SWIFTSHADER}" \
--build-arg "CMAKE_VERSION=${CMAKE_VERSION:-}" \
--build-arg "NINJA_VERSION=${NINJA_VERSION:-}" \
--build-arg "KATEX=${KATEX:-}" \
@ -363,14 +277,6 @@ docker build \
"$@" \
.
# NVIDIA dockers for RC releases use tag names like `11.0-cudnn8-devel-ubuntu18.04-rc`,
# for this case we will set UBUNTU_VERSION to `18.04-rc` so that the Dockerfile could
# find the correct image. As a result, here we have to replace the
# "$UBUNTU_VERSION" == "18.04-rc"
# with
# "$UBUNTU_VERSION" == "18.04"
UBUNTU_VERSION=$(echo ${UBUNTU_VERSION} | sed 's/-rc$//')
function drun() {
docker run --rm "$tmp_tag" $*
}
@ -388,6 +294,19 @@ if [[ "$OS" == "ubuntu" ]]; then
fi
fi
if [ -n "$TRAVIS_PYTHON_VERSION" ]; then
if [[ "$TRAVIS_PYTHON_VERSION" != nightly ]]; then
if !(drun python --version 2>&1 | grep -qF "Python $TRAVIS_PYTHON_VERSION"); then
echo "TRAVIS_PYTHON_VERSION=$TRAVIS_PYTHON_VERSION, but:"
drun python --version
exit 1
fi
else
echo "Please manually check nightly is OK:"
drun python --version
fi
fi
if [ -n "$ANACONDA_PYTHON_VERSION" ]; then
if !(drun python --version 2>&1 | grep -qF "Python $ANACONDA_PYTHON_VERSION"); then
echo "ANACONDA_PYTHON_VERSION=$ANACONDA_PYTHON_VERSION, but:"

View File

@ -13,7 +13,7 @@ retry () {
#until we find a way to reliably reuse previous build, this last_tag is not in use
# last_tag="$(( CIRCLE_BUILD_NUM - 1 ))"
tag="${DOCKER_TAG}"
tag="${CIRCLE_WORKFLOW_ID}"
registry="308535385114.dkr.ecr.us-east-1.amazonaws.com"
@ -45,8 +45,9 @@ trap "docker logout ${registry}" EXIT
docker push "${image}:${tag}"
docker save -o "${IMAGE_NAME}:${tag}.tar" "${image}:${tag}"
# TODO: Get rid of duplicate tagging once ${DOCKER_TAG} becomes the default
docker tag "${image}:${tag}" "${image}:${DOCKER_TAG}"
docker push "${image}:${DOCKER_TAG}"
if [ -z "${DOCKER_SKIP_S3_UPLOAD:-}" ]; then
aws s3 cp "${IMAGE_NAME}:${tag}.tar" "s3://ossci-linux-build/pytorch/base/${IMAGE_NAME}:${tag}.tar" --acl public-read
fi
docker save -o "${IMAGE_NAME}:${tag}.tar" "${image}:${tag}"
aws s3 cp "${IMAGE_NAME}:${tag}.tar" "s3://ossci-linux-build/pytorch/base/${IMAGE_NAME}:${tag}.tar" --acl public-read

View File

@ -1,93 +0,0 @@
ARG CENTOS_VERSION
FROM centos:${CENTOS_VERSION}
ARG CENTOS_VERSION
# Install required packages to build Caffe2
# Install common dependencies (so that this step can be cached separately)
ARG EC2
ADD ./common/install_base.sh install_base.sh
RUN bash ./install_base.sh && rm install_base.sh
# Install devtoolset
ARG DEVTOOLSET_VERSION
ADD ./common/install_devtoolset.sh install_devtoolset.sh
RUN bash ./install_devtoolset.sh && rm install_devtoolset.sh
ENV BASH_ENV "/etc/profile"
# (optional) Install non-default glibc version
ARG GLIBC_VERSION
ADD ./common/install_glibc.sh install_glibc.sh
RUN if [ -n "${GLIBC_VERSION}" ]; then bash ./install_glibc.sh; fi
RUN rm install_glibc.sh
# Install user
ADD ./common/install_user.sh install_user.sh
RUN bash ./install_user.sh && rm install_user.sh
# Install conda and other packages (e.g., numpy, coverage, pytest)
ENV PATH /opt/conda/bin:$PATH
ARG ANACONDA_PYTHON_VERSION
ADD ./common/install_conda.sh install_conda.sh
RUN bash ./install_conda.sh && rm install_conda.sh
# (optional) Install protobuf for ONNX
ARG PROTOBUF
ADD ./common/install_protobuf.sh install_protobuf.sh
RUN if [ -n "${PROTOBUF}" ]; then bash ./install_protobuf.sh; fi
RUN rm install_protobuf.sh
ENV INSTALLED_PROTOBUF ${PROTOBUF}
# (optional) Install database packages like LMDB and LevelDB
ARG DB
ADD ./common/install_db.sh install_db.sh
RUN if [ -n "${DB}" ]; then bash ./install_db.sh; fi
RUN rm install_db.sh
ENV INSTALLED_DB ${DB}
# (optional) Install vision packages like OpenCV and ffmpeg
ARG VISION
ADD ./common/install_vision.sh install_vision.sh
RUN if [ -n "${VISION}" ]; then bash ./install_vision.sh; fi
RUN rm install_vision.sh
ENV INSTALLED_VISION ${VISION}
# Install rocm
ARG ROCM_VERSION
ADD ./common/install_rocm.sh install_rocm.sh
RUN bash ./install_rocm.sh
RUN rm install_rocm.sh
ENV PATH /opt/rocm/bin:$PATH
ENV PATH /opt/rocm/hcc/bin:$PATH
ENV PATH /opt/rocm/hip/bin:$PATH
ENV PATH /opt/rocm/opencl/bin:$PATH
ENV PATH /opt/rocm/llvm/bin:$PATH
ENV MAGMA_HOME /opt/rocm/magma
ENV LANG en_US.utf8
ENV LC_ALL en_US.utf8
# (optional) Install non-default CMake version
ARG CMAKE_VERSION
ADD ./common/install_cmake.sh install_cmake.sh
RUN if [ -n "${CMAKE_VERSION}" ]; then bash ./install_cmake.sh; fi
RUN rm install_cmake.sh
# (optional) Install non-default Ninja version
ARG NINJA_VERSION
ADD ./common/install_ninja.sh install_ninja.sh
RUN if [ -n "${NINJA_VERSION}" ]; then bash ./install_ninja.sh; fi
RUN rm install_ninja.sh
# Install ccache/sccache (do this last, so we get priority in PATH)
ADD ./common/install_cache.sh install_cache.sh
ENV PATH /opt/cache/bin:$PATH
RUN bash ./install_cache.sh && rm install_cache.sh
# Include BUILD_ENVIRONMENT environment variable in image
ARG BUILD_ENVIRONMENT
ENV BUILD_ENVIRONMENT ${BUILD_ENVIRONMENT}
USER jenkins
CMD ["bash"]

View File

@ -4,15 +4,13 @@ set -ex
[ -n "${ANDROID_NDK}" ]
_https_amazon_aws=https://ossci-android.s3.amazonaws.com
apt-get update
apt-get install -y --no-install-recommends autotools-dev autoconf unzip
apt-get autoclean && apt-get clean
rm -rf /var/lib/apt/lists/* /tmp/* /var/tmp/*
pushd /tmp
curl -Os --retry 3 $_https_amazon_aws/android-ndk-${ANDROID_NDK}-linux-x86_64.zip
curl -Os --retry 3 https://dl.google.com/android/repository/android-ndk-${ANDROID_NDK}-linux-x86_64.zip
popd
_ndk_dir=/opt/ndk
mkdir -p "$_ndk_dir"
@ -47,22 +45,43 @@ export JAVA_HOME=/usr/lib/jvm/java-8-openjdk-amd64/
# Installing android sdk
# https://github.com/circleci/circleci-images/blob/staging/android/Dockerfile.m4
_tmp_sdk_zip=/tmp/android-sdk-linux.zip
_sdk_version=sdk-tools-linux-3859397.zip
_android_home=/opt/android/sdk
rm -rf $_android_home
sudo mkdir -p $_android_home
curl --silent --show-error --location --fail --retry 3 --output /tmp/android-sdk-linux.zip $_https_amazon_aws/android-sdk-linux-tools3859397-build-tools2803-2902-platforms28-29.zip
sudo unzip -q $_tmp_sdk_zip -d $_android_home
rm $_tmp_sdk_zip
curl --silent --show-error --location --fail --retry 3 --output /tmp/$_sdk_version https://dl.google.com/android/repository/$_sdk_version
sudo unzip -q /tmp/$_sdk_version -d $_android_home
rm /tmp/$_sdk_version
sudo chmod -R 777 $_android_home
export ANDROID_HOME=$_android_home
export ADB_INSTALL_TIMEOUT=120
export PATH="${ANDROID_HOME}/tools:${ANDROID_HOME}/tools/bin:${ANDROID_HOME}/platform-tools:${PATH}"
export PATH="${ANDROID_HOME}/emulator:${ANDROID_HOME}/tools:${ANDROID_HOME}/tools/bin:${ANDROID_HOME}/platform-tools:${PATH}"
echo "PATH:${PATH}"
alias sdkmanager="$ANDROID_HOME/tools/bin/sdkmanager"
sudo mkdir ~/.android && sudo echo '### User Sources for Android SDK Manager' > ~/.android/repositories.cfg
sudo chmod -R 777 ~/.android
yes | sdkmanager --licenses
yes | sdkmanager --update
sdkmanager \
"tools" \
"platform-tools" \
"emulator"
sdkmanager \
"build-tools;28.0.3" \
"build-tools;29.0.2"
sdkmanager \
"platforms;android-28" \
"platforms;android-29"
sdkmanager --list
# Installing Gradle
echo "GRADLE_VERSION:${GRADLE_VERSION}"
@ -70,7 +89,8 @@ _gradle_home=/opt/gradle
sudo rm -rf $gradle_home
sudo mkdir -p $_gradle_home
curl --silent --output /tmp/gradle.zip --retry 3 $_https_amazon_aws/gradle-${GRADLE_VERSION}-bin.zip
wget --no-verbose --output-document=/tmp/gradle.zip \
"https://services.gradle.org/distributions/gradle-${GRADLE_VERSION}-bin.zip"
sudo unzip -q /tmp/gradle.zip -d $_gradle_home
rm /tmp/gradle.zip
@ -99,7 +119,7 @@ echo "ndk.dir=/opt/ndk" >> $GRADLE_LOCAL_PROPERTIES
chown -R jenkins /var/lib/jenkins/gradledeps
chgrp -R jenkins /var/lib/jenkins/gradledeps
sudo -H -u jenkins $GRADLE_HOME/bin/gradle -Pandroid.useAndroidX=true -p /var/lib/jenkins/gradledeps -g /var/lib/jenkins/.gradle --refresh-dependencies --debug --stacktrace assemble
sudo -H -u jenkins $GRADLE_HOME/bin/gradle -p /var/lib/jenkins/gradledeps -g /var/lib/jenkins/.gradle --refresh-dependencies --debug --stacktrace assemble
chown -R jenkins /var/lib/jenkins/.gradle
chgrp -R jenkins /var/lib/jenkins/.gradle

View File

@ -2,113 +2,55 @@
set -ex
install_ubuntu() {
# NVIDIA dockers for RC releases use tag names like `11.0-cudnn8-devel-ubuntu18.04-rc`,
# for this case we will set UBUNTU_VERSION to `18.04-rc` so that the Dockerfile could
# find the correct image. As a result, here we have to check for
# "$UBUNTU_VERSION" == "18.04"*
# instead of
# "$UBUNTU_VERSION" == "18.04"
if [[ "$UBUNTU_VERSION" == "18.04"* ]]; then
cmake3="cmake=3.10*"
else
cmake3="cmake=3.5*"
fi
# NVIDIA dockers for RC releases use tag names like `11.0-cudnn8-devel-ubuntu18.04-rc`,
# for this case we will set UBUNTU_VERSION to `18.04-rc` so that the Dockerfile could
# find the correct image. As a result, here we have to check for
# "$UBUNTU_VERSION" == "18.04"*
# instead of
# "$UBUNTU_VERSION" == "18.04"
if [[ "$UBUNTU_VERSION" == "18.04"* ]]; then
cmake3="cmake=3.10*"
else
cmake3="cmake=3.5*"
fi
# Install common dependencies
apt-get update
# TODO: Some of these may not be necessary
ccache_deps="asciidoc docbook-xml docbook-xsl xsltproc"
numpy_deps="gfortran"
apt-get install -y --no-install-recommends \
$ccache_deps \
$numpy_deps \
${cmake3} \
apt-transport-https \
autoconf \
automake \
build-essential \
ca-certificates \
curl \
git \
libatlas-base-dev \
libc6-dbg \
libiomp-dev \
libyaml-dev \
libz-dev \
libjpeg-dev \
libasound2-dev \
libsndfile-dev \
software-properties-common \
sudo \
wget \
vim
# Cleanup package manager
apt-get autoclean && apt-get clean
rm -rf /var/lib/apt/lists/* /tmp/* /var/tmp/*
}
install_centos() {
# Need EPEL for many packages we depend on.
# See http://fedoraproject.org/wiki/EPEL
yum --enablerepo=extras install -y epel-release
ccache_deps="asciidoc docbook-dtds docbook-style-xsl libxslt"
numpy_deps="gcc-gfortran"
# Note: protobuf-c-{compiler,devel} on CentOS are too old to be used
# for Caffe2. That said, we still install them to make sure the build
# system opts to build/use protoc and libprotobuf from third-party.
yum install -y \
$ccache_deps \
$numpy_deps \
autoconf \
automake \
bzip2 \
cmake \
cmake3 \
curl \
gcc \
gcc-c++ \
gflags-devel \
git \
glibc-devel \
glibc-headers \
glog-devel \
hiredis-devel \
libstdc++-devel \
libsndfile-devel \
make \
opencv-devel \
sudo \
wget \
vim
# Cleanup
yum clean all
rm -rf /var/cache/yum
rm -rf /var/lib/yum/yumdb
rm -rf /var/lib/yum/history
}
# Install base packages depending on the base OS
ID=$(grep -oP '(?<=^ID=).+' /etc/os-release | tr -d '"')
case "$ID" in
ubuntu)
install_ubuntu
;;
centos)
install_centos
;;
*)
echo "Unable to determine OS..."
exit 1
;;
esac
# Install common dependencies
apt-get update
# TODO: Some of these may not be necessary
# TODO: libiomp also gets installed by conda, aka there's a conflict
ccache_deps="asciidoc docbook-xml docbook-xsl xsltproc"
numpy_deps="gfortran"
apt-get install -y --no-install-recommends \
$ccache_deps \
$numpy_deps \
${cmake3} \
apt-transport-https \
autoconf \
automake \
build-essential \
ca-certificates \
curl \
git \
libatlas-base-dev \
libc6-dbg \
libiomp-dev \
libyaml-dev \
libz-dev \
libjpeg-dev \
libasound2-dev \
libsndfile-dev \
python \
python-dev \
python-setuptools \
python-wheel \
software-properties-common \
sudo \
wget \
vim
# Install Valgrind separately since the apt-get version is too old.
mkdir valgrind_build && cd valgrind_build
VALGRIND_VERSION=3.16.1
VALGRIND_VERSION=3.15.0
if ! wget http://valgrind.org/downloads/valgrind-${VALGRIND_VERSION}.tar.bz2
then
wget https://sourceware.org/ftp/valgrind/valgrind-${VALGRIND_VERSION}.tar.bz2
@ -121,3 +63,13 @@ sudo make install
cd ../../
rm -rf valgrind_build
alias valgrind="/usr/local/bin/valgrind"
# TODO: THIS IS A HACK!!!
# distributed nccl(2) tests are a bit busted, see https://github.com/pytorch/pytorch/issues/5877
if dpkg -s libnccl-dev; then
apt-get remove -y libnccl-dev libnccl2 --allow-change-held-packages
fi
# Cleanup package manager
apt-get autoclean && apt-get clean
rm -rf /var/lib/apt/lists/* /tmp/* /var/tmp/*

View File

@ -1,19 +0,0 @@
#!/bin/bash
set -ex
git clone https://github.com/malfet/breakpad.git -b pytorch/release-1.9
pushd breakpad
git clone https://chromium.googlesource.com/linux-syscall-support src/third_party/lss
pushd src/third_party/lss
# same as with breakpad, there are no real releases for this repo so use a
# commit as the pin
git checkout e1e7b0ad8ee99a875b272c8e33e308472e897660
popd
./configure
make
make install
popd
rm -rf breakpad

View File

@ -2,51 +2,17 @@
set -ex
install_ubuntu() {
echo "Preparing to build sccache from source"
apt-get update
apt-get install -y cargo pkg-config libssl-dev
echo "Checking out sccache repo"
git clone https://github.com/pytorch/sccache
cd sccache
echo "Building sccache"
cargo build --release
cp target/release/sccache /opt/cache/bin
echo "Cleaning up"
cd ..
rm -rf sccache
apt-get remove -y cargo rustc
apt-get autoclean && apt-get clean
}
install_binary() {
echo "Downloading sccache binary from S3 repo"
curl --retry 3 https://s3.amazonaws.com/ossci-linux/sccache -o /opt/cache/bin/sccache
}
mkdir -p /opt/cache/bin
mkdir -p /opt/cache/lib
sed -e 's|PATH="\(.*\)"|PATH="/opt/cache/bin:\1"|g' -i /etc/environment
export PATH="/opt/cache/bin:$PATH"
# Setup compiler cache
if [ -n "$ROCM_VERSION" ]; then
curl --retry 3 http://repo.radeon.com/misc/.sccache_amd/sccache -o /opt/cache/bin/sccache
else
ID=$(grep -oP '(?<=^ID=).+' /etc/os-release | tr -d '"')
case "$ID" in
ubuntu)
install_ubuntu
;;
*)
install_binary
;;
esac
fi
curl --retry 3 https://s3.amazonaws.com/ossci-linux/sccache -o /opt/cache/bin/sccache
chmod a+x /opt/cache/bin/sccache
function write_sccache_stub() {
printf "#!/bin/sh\nif [ \$(ps -p \$PPID -o comm=) != sccache ]; then\n exec sccache $(which $1) \"\$@\"\nelse\n exec $(which $1) \"\$@\"\nfi" > "/opt/cache/bin/$1"
printf "#!/bin/sh\nexec sccache $(which $1) \$*" > "/opt/cache/bin/$1"
chmod a+x "/opt/cache/bin/$1"
}
@ -54,12 +20,8 @@ write_sccache_stub cc
write_sccache_stub c++
write_sccache_stub gcc
write_sccache_stub g++
# NOTE: See specific ROCM_VERSION case below.
if [ "x$ROCM_VERSION" = x ]; then
write_sccache_stub clang
write_sccache_stub clang++
fi
write_sccache_stub clang
write_sccache_stub clang++
if [ -n "$CUDA_VERSION" ]; then
# TODO: This is a workaround for the fact that PyTorch's FindCUDA
@ -68,50 +30,6 @@ if [ -n "$CUDA_VERSION" ]; then
# where CUDA is installed. Instead, we install an nvcc symlink outside
# of the PATH, and set CUDA_NVCC_EXECUTABLE so that we make use of it.
write_sccache_stub nvcc
mv /opt/cache/bin/nvcc /opt/cache/lib/
fi
if [ -n "$ROCM_VERSION" ]; then
# ROCm compiler is hcc or clang. However, it is commonly invoked via hipcc wrapper.
# hipcc will call either hcc or clang using an absolute path starting with /opt/rocm,
# causing the /opt/cache/bin to be skipped. We must create the sccache wrappers
# directly under /opt/rocm while also preserving the original compiler names.
# Note symlinks will chain as follows: [hcc or clang++] -> clang -> clang-??
# Final link in symlink chain must point back to original directory.
# Original compiler is moved one directory deeper. Wrapper replaces it.
function write_sccache_stub_rocm() {
OLDCOMP=$1
COMPNAME=$(basename $OLDCOMP)
TOPDIR=$(dirname $OLDCOMP)
WRAPPED="$TOPDIR/original/$COMPNAME"
mv "$OLDCOMP" "$WRAPPED"
printf "#!/bin/sh\nexec sccache $WRAPPED \"\$@\"" > "$OLDCOMP"
chmod a+x "$OLDCOMP"
}
if [[ -e "/opt/rocm/hcc/bin/hcc" ]]; then
# ROCm 3.3 or earlier.
mkdir /opt/rocm/hcc/bin/original
write_sccache_stub_rocm /opt/rocm/hcc/bin/hcc
write_sccache_stub_rocm /opt/rocm/hcc/bin/clang
write_sccache_stub_rocm /opt/rocm/hcc/bin/clang++
# Fix last link in symlink chain, clang points to versioned clang in prior dir
pushd /opt/rocm/hcc/bin/original
ln -s ../$(readlink clang)
popd
elif [[ -e "/opt/rocm/llvm/bin/clang" ]]; then
# ROCm 3.5 and beyond.
mkdir /opt/rocm/llvm/bin/original
write_sccache_stub_rocm /opt/rocm/llvm/bin/clang
write_sccache_stub_rocm /opt/rocm/llvm/bin/clang++
# Fix last link in symlink chain, clang points to versioned clang in prior dir
pushd /opt/rocm/llvm/bin/original
ln -s ../$(readlink clang)
popd
else
echo "Cannot find ROCm compiler."
exit 1
fi
printf "#!/bin/sh\nexec sccache $(which nvcc) \"\$@\"" > /opt/cache/lib/nvcc
chmod a+x /opt/cache/lib/nvcc
fi

View File

@ -24,20 +24,13 @@ if [ -n "$ANACONDA_PYTHON_VERSION" ]; then
mkdir /opt/conda
chown jenkins:jenkins /opt/conda
# Work around bug where devtoolset replaces sudo and breaks it.
if [ -n "$DEVTOOLSET_VERSION" ]; then
SUDO=/bin/sudo
else
SUDO=sudo
fi
as_jenkins() {
# NB: unsetting the environment variables works around a conda bug
# https://github.com/conda/conda/issues/6576
# NB: Pass on PATH and LD_LIBRARY_PATH to sudo invocation
# NB: This must be run from a directory that jenkins has access to,
# works around https://github.com/conda/conda-package-handling/pull/34
$SUDO -H -u jenkins env -u SUDO_UID -u SUDO_GID -u SUDO_COMMAND -u SUDO_USER env "PATH=$PATH" "LD_LIBRARY_PATH=$LD_LIBRARY_PATH" $*
sudo -H -u jenkins env -u SUDO_UID -u SUDO_GID -u SUDO_COMMAND -u SUDO_USER env "PATH=$PATH" "LD_LIBRARY_PATH=$LD_LIBRARY_PATH" $*
}
pushd /tmp
@ -56,10 +49,10 @@ if [ -n "$ANACONDA_PYTHON_VERSION" ]; then
pushd /opt/conda
# Track latest conda update
as_jenkins conda update -y -n base conda
as_jenkins conda update -n base conda
# Install correct Python version
as_jenkins conda install -y python="$ANACONDA_PYTHON_VERSION"
as_jenkins conda install python="$ANACONDA_PYTHON_VERSION"
conda_install() {
# Ensure that the install command don't upgrade/downgrade Python
@ -71,69 +64,33 @@ if [ -n "$ANACONDA_PYTHON_VERSION" ]; then
# Install PyTorch conda deps, as per https://github.com/pytorch/pytorch README
# DO NOT install cmake here as it would install a version newer than 3.5, but
# we want to pin to version 3.5.
SCIPY_VERSION=1.1.0
if [ "$ANACONDA_PYTHON_VERSION" = "3.9" ]; then
if [ "$ANACONDA_PYTHON_VERSION" = "3.8" ]; then
# DO NOT install typing if installing python-3.8, since its part of python-3.8 core packages
# Install llvm-8 as it is required to compile llvmlite-0.30.0 from source
conda_install numpy=1.19.2 astunparse pyyaml mkl mkl-include setuptools cffi future six llvmdev=8.0.0 -c conda-forge
SCIPY_VERSION=1.6.0
elif [ "$ANACONDA_PYTHON_VERSION" = "3.8" ]; then
# Install llvm-8 as it is required to compile llvmlite-0.30.0 from source
conda_install numpy=1.18.5 astunparse pyyaml mkl mkl-include setuptools cffi future six llvmdev=8.0.0
elif [ "$ANACONDA_PYTHON_VERSION" = "3.7" ]; then
# DO NOT install dataclasses if installing python-3.7, since its part of python-3.7 core packages
conda_install numpy=1.18.5 astunparse pyyaml mkl mkl-include setuptools cffi future six typing_extensions
conda_install numpy pyyaml mkl mkl-include setuptools cffi future six llvmdev=8.0.0
else
conda_install numpy=1.18.5 astunparse pyyaml mkl mkl-include setuptools cffi future six dataclasses typing_extensions
conda_install numpy pyyaml mkl mkl-include setuptools cffi typing future six
fi
if [[ "$CUDA_VERSION" == 10.0* ]]; then
if [[ "$CUDA_VERSION" == 9.2* ]]; then
conda_install magma-cuda92 -c pytorch
elif [[ "$CUDA_VERSION" == 10.0* ]]; then
conda_install magma-cuda100 -c pytorch
elif [[ "$CUDA_VERSION" == 10.1* ]]; then
conda_install magma-cuda101 -c pytorch
elif [[ "$CUDA_VERSION" == 10.2* ]]; then
conda_install magma-cuda102 -c pytorch
elif [[ "$CUDA_VERSION" == 11.0* ]]; then
conda_install magma-cuda110 -c pytorch
elif [[ "$CUDA_VERSION" == 11.1* ]]; then
conda_install magma-cuda111 -c pytorch
elif [[ "$CUDA_VERSION" == 11.3* ]]; then
conda_install magma-cuda113 -c pytorch
fi
# TODO: This isn't working atm
conda_install nnpack -c killeent
# Install some other packages, including those needed for Python test reporting
# Install some other packages
# TODO: Why is scipy pinned
# Pin MyPy version because new errors are likely to appear with each release
# Pin hypothesis to avoid flakiness: https://github.com/pytorch/pytorch/issues/31136
# Pin coverage so we can use COVERAGE_RCFILE
as_jenkins pip install --progress-bar off pytest \
scipy==$SCIPY_VERSION \
scikit-image \
psutil \
unittest-xml-reporting \
boto3==1.16.34 \
coverage==5.5 \
hypothesis==4.53.2 \
mypy==0.812 \
tb-nightly
# Install numba only on python-3.8 or below
# For numba issue see https://github.com/pytorch/pytorch/issues/51511
if [[ $(python -c "import sys; print(int(sys.version_info < (3, 9)))") == "1" ]]; then
as_jenkins pip install --progress-bar off numba librosa>=0.6.2
else
as_jenkins pip install --progress-bar off numba==0.49.0 librosa>=0.6.2
fi
# Update scikit-learn to a python-3.8 compatible version
if [[ $(python -c "import sys; print(int(sys.version_info >= (3, 8)))") == "1" ]]; then
as_jenkins pip install --progress-bar off -U scikit-learn
else
# Pinned scikit-learn due to https://github.com/scikit-learn/scikit-learn/issues/14485 (affects gcc 5.5 only)
as_jenkins pip install --progress-bar off scikit-learn==0.20.3
fi
# numba & llvmlite is pinned because of https://github.com/numba/numba/issues/4368
# scikit-learn is pinned because of
# https://github.com/scikit-learn/scikit-learn/issues/14485 (affects gcc 5.5
# only)
as_jenkins pip install --progress-bar off pytest scipy==1.1.0 scikit-learn==0.20.3 scikit-image librosa>=0.6.2 psutil numba==0.46.0 llvmlite==0.30.0
popd
fi

View File

@ -51,16 +51,11 @@ install_centos() {
}
# Install base packages depending on the base OS
ID=$(grep -oP '(?<=^ID=).+' /etc/os-release | tr -d '"')
case "$ID" in
ubuntu)
install_ubuntu
;;
centos)
install_centos
;;
*)
echo "Unable to determine OS..."
exit 1
;;
esac
if [ -f /etc/lsb-release ]; then
install_ubuntu
elif [ -f /etc/os-release ]; then
install_centos
else
echo "Unable to determine OS..."
exit 1
fi

View File

@ -15,7 +15,6 @@ if [ -n "$GCC_VERSION" ]; then
update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-"$GCC_VERSION" 50
update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-"$GCC_VERSION" 50
update-alternatives --install /usr/bin/gcov gcov /usr/bin/gcov-"$GCC_VERSION" 50
# Cleanup package manager
apt-get autoclean && apt-get clean

View File

@ -1,8 +0,0 @@
#!/bin/bash
set -ex
git clone --branch v1.15 https://github.com/linux-test-project/lcov.git
pushd lcov
sudo make install # will be installed in /usr/local/bin/lcov
popd

View File

@ -0,0 +1,30 @@
#!/bin/bash
set -ex
llvm_url="https://github.com/llvm/llvm-project/releases/download/llvmorg-9.0.1/llvm-9.0.1.src.tar.xz"
mkdir /opt/llvm
pushd /tmp
wget --no-verbose --output-document=llvm.tar.xz "$llvm_url"
mkdir llvm
tar -xf llvm.tar.xz -C llvm --strip-components 1
rm -f llvm.tar.xz
cd llvm
mkdir build
cd build
cmake -G "Unix Makefiles" \
-DCMAKE_BUILD_TYPE=MinSizeRel \
-DLLVM_ENABLE_ASSERTIONS=ON \
-DCMAKE_INSTALL_PREFIX=/opt/llvm \
-DLLVM_TARGETS_TO_BUILD="host" \
-DLLVM_BUILD_TOOLS=OFF \
-DLLVM_BUILD_UTILS=OFF \
-DLLVM_TEMPORARILY_ALLOW_OLD_TOOLCHAIN=ON \
../
make -j4
sudo make install
popd

View File

@ -1,4 +0,0 @@
#!/bin/bash
sudo apt-get -qq update
sudo apt-get -qq install --allow-downgrades --allow-change-held-packages libnccl-dev=2.5.6-1+cuda10.1 libnccl2=2.5.6-1+cuda10.1

View File

@ -1,4 +0,0 @@
#!/bin/bash
sudo apt-get update
sudo apt-get install -y --allow-downgrades --allow-change-held-packages openmpi-bin libopenmpi-dev

View File

@ -1,14 +0,0 @@
#!/bin/bash
set -ex
OPENSSL=openssl-1.1.1k
wget -q -O "${OPENSSL}.tar.gz" "https://www.openssl.org/source/${OPENSSL}.tar.gz"
tar xf "${OPENSSL}.tar.gz"
cd "${OPENSSL}"
./config --prefix=/opt/openssl -d '-Wl,--enable-new-dtags,-rpath,$(LIBRPATH)'
# NOTE: opensl errors out when built with the -j option
make install_sw
cd ..
rm -rf "${OPENSSL}"

View File

@ -46,16 +46,11 @@ install_centos() {
}
# Install base packages depending on the base OS
ID=$(grep -oP '(?<=^ID=).+' /etc/os-release | tr -d '"')
case "$ID" in
ubuntu)
install_ubuntu
;;
centos)
install_centos
;;
*)
echo "Unable to determine OS..."
exit 1
;;
esac
if [ -f /etc/lsb-release ]; then
install_ubuntu
elif [ -f /etc/os-release ]; then
install_centos
else
echo "Unable to determine OS..."
exit 1
fi

View File

@ -2,80 +2,47 @@
set -ex
install_magma() {
# "install" hipMAGMA into /opt/rocm/magma by copying after build
git clone https://bitbucket.org/icl/magma.git
pushd magma
git checkout 878b1ce02e9cfe4a829be22c8f911e9c0b6bd88f
cp make.inc-examples/make.inc.hip-gcc-mkl make.inc
echo 'LIBDIR += -L$(MKLROOT)/lib' >> make.inc
echo 'LIB += -Wl,--enable-new-dtags -Wl,--rpath,/opt/rocm/lib -Wl,--rpath,$(MKLROOT)/lib -Wl,--rpath,/opt/rocm/magma/lib' >> make.inc
echo 'DEVCCFLAGS += --amdgpu-target=gfx803 --amdgpu-target=gfx900 --amdgpu-target=gfx906 --amdgpu-target=gfx908 --gpu-max-threads-per-block=256' >> make.inc
# hipcc with openmp flag may cause isnan() on __device__ not to be found; depending on context, compiler may attempt to match with host definition
sed -i 's/^FOPENMP/#FOPENMP/g' make.inc
export PATH="${PATH}:/opt/rocm/bin"
make -f make.gen.hipMAGMA -j $(nproc)
LANG=C.UTF-8 make lib/libmagma.so -j $(nproc) MKLROOT=/opt/conda
make testing/testing_dgemm -j $(nproc) MKLROOT=/opt/conda
popd
mv magma /opt/rocm
}
ver() {
printf "%3d%03d%03d%03d" $(echo "$1" | tr '.' ' ');
}
install_ubuntu() {
apt-get update
if [[ $UBUNTU_VERSION == 18.04 ]]; then
# gpg-agent is not available by default on 18.04
apt-get install -y --no-install-recommends gpg-agent
fi
apt-get install -y kmod
apt-get install -y wget
apt-get install -y libopenblas-dev
# Need the libc++1 and libc++abi1 libraries to allow torch._C to load at runtime
apt-get install -y libc++1
apt-get install -y libc++abi1
ROCM_REPO="ubuntu"
if [[ $(ver $ROCM_VERSION) -lt $(ver 4.2) ]]; then
ROCM_REPO="xenial"
fi
DEB_ROCM_REPO=http://repo.radeon.com/rocm/apt/${ROCM_VERSION}
# Add rocm repository
wget -qO - http://repo.radeon.com/rocm/rocm.gpg.key | apt-key add -
echo "deb [arch=amd64] http://repo.radeon.com/rocm/apt/${ROCM_VERSION} ${ROCM_REPO} main" > /etc/apt/sources.list.d/rocm.list
wget -qO - $DEB_ROCM_REPO/rocm.gpg.key | apt-key add -
echo "deb [arch=amd64] $DEB_ROCM_REPO xenial main" > /etc/apt/sources.list.d/rocm.list
apt-get update --allow-insecure-repositories
DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-unauthenticated \
rocm-dev \
rocm-utils \
rocm-libs \
rocfft \
miopen-hip \
rocblas \
hipsparse \
rocrand \
hipcub \
rocthrust \
rccl \
rocprofiler-dev \
roctracer-dev
# precompiled miopen kernels added in ROCm 3.5; search for all unversioned packages
# if search fails it will abort this script; use true to avoid case where search fails
MIOPENKERNELS=$(apt-cache search --names-only miopenkernels | awk '{print $1}' | grep -F -v . || true)
if [[ "x${MIOPENKERNELS}" = x ]]; then
echo "miopenkernels package not available"
else
DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-unauthenticated ${MIOPENKERNELS}
fi
install_magma
# Cleanup
apt-get autoclean && apt-get clean
rm -rf /var/lib/apt/lists/* /tmp/* /var/tmp/*
# Cleanup
apt-get autoclean && apt-get clean
rm -rf /var/lib/apt/lists/* /tmp/* /var/tmp/*
}
install_centos() {
yum update -y
yum install -y kmod
yum install -y wget
yum install -y openblas-devel
@ -84,7 +51,7 @@ install_centos() {
echo "[ROCm]" > /etc/yum.repos.d/rocm.repo
echo "name=ROCm" >> /etc/yum.repos.d/rocm.repo
echo "baseurl=http://repo.radeon.com/rocm/yum/${ROCM_VERSION}" >> /etc/yum.repos.d/rocm.repo
echo "baseurl=http://repo.radeon.com/rocm/yum/rpm/" >> /etc/yum.repos.d/rocm.repo
echo "enabled=1" >> /etc/yum.repos.d/rocm.repo
echo "gpgcheck=0" >> /etc/yum.repos.d/rocm.repo
@ -93,13 +60,17 @@ install_centos() {
yum install -y \
rocm-dev \
rocm-utils \
rocm-libs \
rocfft \
miopen-hip \
rocblas \
hipsparse \
rocrand \
rccl \
hipcub \
rocthrust \
rocprofiler-dev \
roctracer-dev
install_magma
# Cleanup
yum clean all
rm -rf /var/cache/yum
@ -108,16 +79,11 @@ install_centos() {
}
# Install Python packages depending on the base OS
ID=$(grep -oP '(?<=^ID=).+' /etc/os-release | tr -d '"')
case "$ID" in
ubuntu)
install_ubuntu
;;
centos)
install_centos
;;
*)
echo "Unable to determine OS..."
exit 1
;;
esac
if [ -f /etc/lsb-release ]; then
install_ubuntu
elif [ -f /etc/os-release ]; then
install_centos
else
echo "Unable to determine OS..."
exit 1
fi

View File

@ -1,24 +0,0 @@
#!/bin/bash
set -ex
[ -n "${SWIFTSHADER}" ]
retry () {
$* || (sleep 1 && $*) || (sleep 2 && $*) || (sleep 4 && $*) || (sleep 8 && $*)
}
_https_amazon_aws=https://ossci-android.s3.amazonaws.com
# SwiftShader
_swiftshader_dir=/var/lib/jenkins/swiftshader
_swiftshader_file_targz=swiftshader-abe07b943-prebuilt.tar.gz
mkdir -p $_swiftshader_dir
_tmp_swiftshader_targz="/tmp/${_swiftshader_file_targz}"
curl --silent --show-error --location --fail --retry 3 \
--output "${_tmp_swiftshader_targz}" "$_https_amazon_aws/${_swiftshader_file_targz}"
tar -C "${_swiftshader_dir}" -xzf "${_tmp_swiftshader_targz}"
export VK_ICD_FILENAMES="${_swiftshader_dir}/build/Linux/vk_swiftshader_icd.json"

View File

@ -0,0 +1,97 @@
#!/bin/bash
set -ex
as_jenkins() {
# NB: Preserve PATH and LD_LIBRARY_PATH changes
sudo -H -u jenkins env "PATH=$PATH" "LD_LIBRARY_PATH=$LD_LIBRARY_PATH" $*
}
if [ -n "$TRAVIS_PYTHON_VERSION" ]; then
mkdir -p /opt/python
chown jenkins:jenkins /opt/python
# Download Python binary from Travis
pushd tmp
as_jenkins wget --quiet ${TRAVIS_DL_URL_PREFIX}/python-$TRAVIS_PYTHON_VERSION.tar.bz2
# NB: The tarball also comes with /home/travis virtualenv that we
# don't care about. (Maybe we should, but we've worked around the
# "how do I install to python" issue by making this entire directory
# user-writable "lol")
# NB: Relative ordering of opt/python and flags matters
as_jenkins tar xjf python-$TRAVIS_PYTHON_VERSION.tar.bz2 --strip-components=2 --directory /opt/python opt/python
popd
echo "/opt/python/$TRAVIS_PYTHON_VERSION/lib" > /etc/ld.so.conf.d/travis-python.conf
ldconfig
sed -e 's|PATH="\(.*\)"|PATH="/opt/python/'"$TRAVIS_PYTHON_VERSION"'/bin:\1"|g' -i /etc/environment
export PATH="/opt/python/$TRAVIS_PYTHON_VERSION/bin:$PATH"
python --version
pip --version
# Install pip from source.
# The python-pip package on Ubuntu Trusty is old
# and upon install numpy doesn't use the binary
# distribution, and fails to compile it from source.
pushd tmp
as_jenkins curl -L -O https://pypi.python.org/packages/11/b6/abcb525026a4be042b486df43905d6893fb04f05aac21c32c638e939e447/pip-9.0.1.tar.gz
as_jenkins tar zxf pip-9.0.1.tar.gz
pushd pip-9.0.1
as_jenkins python setup.py install
popd
rm -rf pip-9.0.1*
popd
# Install pip packages
as_jenkins pip install --upgrade pip
pip --version
if [[ "$TRAVIS_PYTHON_VERSION" == nightly ]]; then
# These two packages have broken Cythonizations uploaded
# to PyPi, see:
#
# - https://github.com/numpy/numpy/issues/10500
# - https://github.com/yaml/pyyaml/issues/117
#
# Furthermore, the released version of Cython does not
# have these issues fixed.
#
# While we are waiting on fixes for these, we build
# from Git for now. Feel free to delete this conditional
# branch if things start working again (you may need
# to do this if these packages regress on Git HEAD.)
as_jenkins pip install git+https://github.com/cython/cython.git
as_jenkins pip install git+https://github.com/numpy/numpy.git
as_jenkins pip install git+https://github.com/yaml/pyyaml.git
else
as_jenkins pip install numpy pyyaml
fi
as_jenkins pip install \
future \
hypothesis \
protobuf \
pytest \
pillow \
typing
as_jenkins pip install mkl mkl-devel
# SciPy does not support Python 3.7 or Python 2.7.9
if [[ "$TRAVIS_PYTHON_VERSION" != nightly ]] && [[ "$TRAVIS_PYTHON_VERSION" != "2.7.9" ]]; then
as_jenkins pip install scipy==1.1.0 scikit-image librosa>=0.6.2
fi
# Install psutil for dataloader tests
as_jenkins pip install psutil
# Install dill for serialization tests
as_jenkins pip install "dill>=0.3.1"
# Cleanup package manager
apt-get autoclean && apt-get clean
rm -rf /var/lib/apt/lists/* /tmp/* /var/tmp/*
fi

View File

@ -47,16 +47,11 @@ install_centos() {
}
# Install base packages depending on the base OS
ID=$(grep -oP '(?<=^ID=).+' /etc/os-release | tr -d '"')
case "$ID" in
ubuntu)
install_ubuntu
;;
centos)
install_centos
;;
*)
echo "Unable to determine OS..."
exit 1
;;
esac
if [ -f /etc/lsb-release ]; then
install_ubuntu
elif [ -f /etc/os-release ]; then
install_centos
else
echo "Unable to determine OS..."
exit 1
fi

View File

@ -1,24 +0,0 @@
#!/bin/bash
set -ex
[ -n "${VULKAN_SDK_VERSION}" ]
retry () {
$* || (sleep 1 && $*) || (sleep 2 && $*) || (sleep 4 && $*) || (sleep 8 && $*)
}
_vulkansdk_dir=/var/lib/jenkins/vulkansdk
_tmp_vulkansdk_targz=/tmp/vulkansdk.tar.gz
curl \
--silent \
--show-error \
--location \
--fail \
--retry 3 \
--output "${_tmp_vulkansdk_targz}" "https://ossci-android.s3.amazonaws.com/vulkansdk-linux-x86_64-${VULKAN_SDK_VERSION}.tar.gz"
mkdir -p "${_vulkansdk_dir}"
tar -C "${_vulkansdk_dir}" -xzf "${_tmp_vulkansdk_targz}" --strip-components 1
rm -rf "${_tmp_vulkansdk_targz}"

View File

@ -24,7 +24,7 @@ ARG KATEX
ADD ./common/install_katex.sh install_katex.sh
RUN bash ./install_katex.sh && rm install_katex.sh
# Install conda and other packages (e.g., numpy, coverage, pytest)
# Install conda
ENV PATH /opt/conda/bin:$PATH
ARG ANACONDA_PYTHON_VERSION
ADD ./common/install_conda.sh install_conda.sh
@ -40,6 +40,12 @@ ARG CLANG_VERSION
ADD ./common/install_clang.sh install_clang.sh
RUN bash ./install_clang.sh && rm install_clang.sh
# Install non-standard Python versions (via Travis binaries)
ARG TRAVIS_PYTHON_VERSION
ENV PATH /opt/python/$TRAVIS_PYTHON_VERSION/bin:$PATH
ADD ./common/install_travis_python.sh install_travis_python.sh
RUN bash ./install_travis_python.sh && rm install_travis_python.sh
# (optional) Install protobuf for ONNX
ARG PROTOBUF
ADD ./common/install_protobuf.sh install_protobuf.sh
@ -61,10 +67,6 @@ RUN if [ -n "${VISION}" ]; then bash ./install_vision.sh; fi
RUN rm install_vision.sh
ENV INSTALLED_VISION ${VISION}
ADD ./common/install_openssl.sh install_openssl.sh
ENV OPENSSL_ROOT_DIR /opt/openssl
RUN bash ./install_openssl.sh
# Install ccache/sccache (do this last, so we get priority in PATH)
ADD ./common/install_cache.sh install_cache.sh
ENV PATH /opt/cache/bin:$PATH
@ -76,16 +78,6 @@ ADD ./common/install_jni.sh install_jni.sh
ADD ./java/jni.h jni.h
RUN bash ./install_jni.sh && rm install_jni.sh
# Install NCCL for when CUDA is version 10.1
ADD ./common/install_nccl.sh install_nccl.sh
RUN if [ "${CUDA_VERSION}" = 10.1 ]; then bash ./install_nccl.sh; fi
RUN rm install_nccl.sh
# Install Open MPI for CUDA
ADD ./common/install_openmpi.sh install_openmpi.sh
RUN if [ -n "${CUDA_VERSION}" ]; then bash install_openmpi.sh; fi
RUN rm install_openmpi.sh
# Include BUILD_ENVIRONMENT environment variable in image
ARG BUILD_ENVIRONMENT
ENV BUILD_ENVIRONMENT ${BUILD_ENVIRONMENT}
@ -94,8 +86,9 @@ ENV BUILD_ENVIRONMENT ${BUILD_ENVIRONMENT}
ENV TORCH_CUDA_ARCH_LIST Maxwell
ENV TORCH_NVCC_FLAGS "-Xfatbin -compress-all"
# Install LLVM dev version (Defined in the pytorch/builder github repository)
COPY --from=pytorch/llvm:9.0.1 /opt/llvm /opt/llvm
# Install LLVM dev version
ADD ./common/install_llvm.sh install_llvm.sh
RUN bash ./install_llvm.sh
USER jenkins
CMD ["bash"]

View File

@ -21,17 +21,12 @@ RUN bash ./install_clang.sh && rm install_clang.sh
ADD ./common/install_user.sh install_user.sh
RUN bash ./install_user.sh && rm install_user.sh
# Install conda and other packages (e.g., numpy, coverage, pytest)
# Install conda
ENV PATH /opt/conda/bin:$PATH
ARG ANACONDA_PYTHON_VERSION
ADD ./common/install_conda.sh install_conda.sh
RUN bash ./install_conda.sh && rm install_conda.sh
# Install gcc
ARG GCC_VERSION
ADD ./common/install_gcc.sh install_gcc.sh
RUN bash ./install_gcc.sh && rm install_gcc.sh
# (optional) Install protobuf for ONNX
ARG PROTOBUF
ADD ./common/install_protobuf.sh install_protobuf.sh
@ -62,8 +57,7 @@ ENV PATH /opt/rocm/bin:$PATH
ENV PATH /opt/rocm/hcc/bin:$PATH
ENV PATH /opt/rocm/hip/bin:$PATH
ENV PATH /opt/rocm/opencl/bin:$PATH
ENV PATH /opt/rocm/llvm/bin:$PATH
ENV MAGMA_HOME /opt/rocm/magma
ENV HIP_PLATFORM hcc
ENV LANG C.UTF-8
ENV LC_ALL C.UTF-8

View File

@ -33,7 +33,7 @@ ARG KATEX
ADD ./common/install_katex.sh install_katex.sh
RUN bash ./install_katex.sh && rm install_katex.sh
# Install conda and other packages (e.g., numpy, coverage, pytest)
# Install conda
ENV PATH /opt/conda/bin:$PATH
ARG ANACONDA_PYTHON_VERSION
ADD ./common/install_conda.sh install_conda.sh
@ -44,9 +44,12 @@ ARG GCC_VERSION
ADD ./common/install_gcc.sh install_gcc.sh
RUN bash ./install_gcc.sh && rm install_gcc.sh
# Install lcov for C++ code coverage
ADD ./common/install_lcov.sh install_lcov.sh
RUN bash ./install_lcov.sh && rm install_lcov.sh
# Install non-standard Python versions (via Travis binaries)
ARG TRAVIS_PYTHON_VERSION
ARG TRAVIS_DL_URL_PREFIX
ENV PATH /opt/python/$TRAVIS_PYTHON_VERSION/bin:$PATH
ADD ./common/install_travis_python.sh install_travis_python.sh
RUN bash ./install_travis_python.sh && rm install_travis_python.sh
# (optional) Install protobuf for ONNX
ARG PROTOBUF
@ -82,25 +85,6 @@ RUN rm AndroidManifest.xml
RUN rm build.gradle
ENV INSTALLED_ANDROID ${ANDROID}
# (optional) Install breakpad
ARG BREAKPAD
ADD ./common/install_breakpad.sh install_breakpad.sh
RUN if [ -n "${BREAKPAD}" ]; then bash ./install_breakpad.sh; fi
RUN rm install_breakpad.sh
ENV INSTALLED_BREAKPAD ${BREAKPAD}
# (optional) Install Vulkan SDK
ARG VULKAN_SDK_VERSION
ADD ./common/install_vulkan_sdk.sh install_vulkan_sdk.sh
RUN if [ -n "${VULKAN_SDK_VERSION}" ]; then bash ./install_vulkan_sdk.sh; fi
RUN rm install_vulkan_sdk.sh
# (optional) Install swiftshader
ARG SWIFTSHADER
ADD ./common/install_swiftshader.sh install_swiftshader.sh
RUN if [ -n "${SWIFTSHADER}" ]; then bash ./install_swiftshader.sh; fi
RUN rm install_swiftshader.sh
# (optional) Install non-default CMake version
ARG CMAKE_VERSION
ADD ./common/install_cmake.sh install_cmake.sh
@ -113,10 +97,6 @@ ADD ./common/install_ninja.sh install_ninja.sh
RUN if [ -n "${NINJA_VERSION}" ]; then bash ./install_ninja.sh; fi
RUN rm install_ninja.sh
ADD ./common/install_openssl.sh install_openssl.sh
RUN bash ./install_openssl.sh
ENV OPENSSL_ROOT_DIR /opt/openssl
# Install ccache/sccache (do this last, so we get priority in PATH)
ADD ./common/install_cache.sh install_cache.sh
ENV PATH /opt/cache/bin:$PATH
@ -131,8 +111,9 @@ RUN bash ./install_jni.sh && rm install_jni.sh
ARG BUILD_ENVIRONMENT
ENV BUILD_ENVIRONMENT ${BUILD_ENVIRONMENT}
# Install LLVM dev version (Defined in the pytorch/builder github repository)
COPY --from=pytorch/llvm:9.0.1 /opt/llvm /opt/llvm
# Install LLVM dev version
ADD ./common/install_llvm.sh install_llvm.sh
RUN bash ./install_llvm.sh
USER jenkins
CMD ["bash"]

View File

@ -1,10 +1,10 @@
FROM ubuntu:18.04
FROM ubuntu:16.04
RUN apt-get update && apt-get install -y python3-pip git && rm -rf /var/lib/apt/lists/* /var/log/dpkg.log
RUN apt-get update && apt-get install -y python-pip git && rm -rf /var/lib/apt/lists/* /var/log/dpkg.log
ADD requirements.txt /requirements.txt
RUN pip3 install -r /requirements.txt
RUN pip install -r /requirements.txt
ADD gc.py /usr/bin/gc.py

View File

@ -1,4 +1,4 @@
#!/usr/bin/env python3
#!/usr/bin/env python
from collections import namedtuple

View File

@ -1,11 +1,11 @@
#!/usr/bin/env python3
#!/usr/bin/env python
import argparse
import boto3
import datetime
import boto3
import pytz
import re
import sys
import re
def save_to_s3(project, data):
@ -88,9 +88,6 @@ parser = argparse.ArgumentParser(description="Delete old Docker tags from regist
parser.add_argument(
"--dry-run", action="store_true", help="Dry run; print tags that would be deleted"
)
parser.add_argument(
"--debug", action="store_true", help="Debug, print ignored / saved tags"
)
parser.add_argument(
"--keep-stable-days",
type=int,
@ -148,12 +145,9 @@ def chunks(chunkable, n):
""" Yield successive n-sized chunks from l.
"""
for i in range(0, len(chunkable), n):
yield chunkable[i: i + n]
yield chunkable[i : i + n]
SHA_PATTERN = re.compile(r'^[0-9a-f]{40}$')
def looks_like_git_sha(tag):
"""Returns a boolean to check if a tag looks like a git sha
@ -162,7 +156,6 @@ def looks_like_git_sha(tag):
"""
return re.match(SHA_PATTERN, tag) is not None
stable_window_tags = []
for repo in repos(client):
repositoryName = repo["repositoryName"]
@ -171,48 +164,51 @@ for repo in repos(client):
# Keep list of image digests to delete for this repository
digest_to_delete = []
print(repositoryName)
for image in images(client, repo):
tags = image.get("imageTags")
if not isinstance(tags, (list,)) or len(tags) == 0:
continue
tag = tags[0]
created = image["imagePushedAt"]
age = now - created
for tag in tags:
if any([
looks_like_git_sha(tag),
tag.isdigit(),
tag.count("-") == 4, # TODO: Remove, this no longer applies as tags are now built using a SHA1
tag in ignore_tags]):
window = stable_window
if tag in ignore_tags:
stable_window_tags.append((repositoryName, tag, "", age, created))
elif age < window:
stable_window_tags.append((repositoryName, tag, window, age, created))
else:
window = unstable_window
if tag in ignore_tags or age < window:
if args.debug:
print("Ignoring {}:{} (age: {})".format(repositoryName, tag, age))
break
if any([
looks_like_git_sha(tag),
tag.isdigit(),
tag.count("-") == 4, # TODO: Remove, this no longer applies as tags are now built using a SHA1
tag in ignore_tags]):
window = stable_window
if tag in ignore_tags:
stable_window_tags.append((repositoryName, tag, "", age, created))
elif age < window:
stable_window_tags.append((repositoryName, tag, window, age, created))
else:
for tag in tags:
print("{}Deleting {}:{} (age: {})".format("(dry run) " if args.dry_run else "", repositoryName, tag, age))
digest_to_delete.append(image["imageDigest"])
if args.dry_run:
if args.debug:
print("Skipping actual deletion, moving on...")
else:
# Issue batch delete for all images to delete for this repository
# Note that as of 2018-07-25, the maximum number of images you can
# delete in a single batch is 100, so chunk our list into batches of
# 100
for c in chunks(digest_to_delete, 100):
client.batch_delete_image(
registryId="308535385114",
repositoryName=repositoryName,
imageIds=[{"imageDigest": digest} for digest in c],
)
window = unstable_window
save_to_s3(args.filter_prefix, stable_window_tags)
if tag in ignore_tags:
print("Ignoring tag {}:{} (age: {})".format(repositoryName, tag, age))
continue
if age < window:
print("Not deleting manifest for tag {}:{} (age: {})".format(repositoryName, tag, age))
continue
if args.dry_run:
print("(dry run) Deleting manifest for tag {}:{} (age: {})".format(repositoryName, tag, age))
else:
print("Deleting manifest for tag{}:{} (age: {})".format(repositoryName, tag, age))
digest_to_delete.append(image["imageDigest"])
# Issue batch delete for all images to delete for this repository
# Note that as of 2018-07-25, the maximum number of images you can
# delete in a single batch is 100, so chunk our list into batches of
# 100
for c in chunks(digest_to_delete, 100):
client.batch_delete_image(
registryId="308535385114",
repositoryName=repositoryName,
imageIds=[{"imageDigest": digest} for digest in c],
)
save_to_s3(args.filter_prefix, stable_window_tags)

View File

@ -8,9 +8,10 @@ Please see README.md in this directory for details.
import os
import shutil
import sys
from collections import namedtuple
from collections import OrderedDict, namedtuple
import cimodel.data.binary_build_definitions as binary_build_definitions
import cimodel.data.caffe2_build_definitions as caffe2_build_definitions
import cimodel.data.pytorch_build_definitions as pytorch_build_definitions
import cimodel.data.simple.android_definitions
import cimodel.data.simple.bazel_definitions
@ -22,7 +23,6 @@ import cimodel.data.simple.macos_definitions
import cimodel.data.simple.mobile_definitions
import cimodel.data.simple.nightly_android
import cimodel.data.simple.nightly_ios
import cimodel.data.simple.anaconda_prune_defintions
import cimodel.data.windows_build_definitions as windows_build_definitions
import cimodel.lib.miniutils as miniutils
import cimodel.lib.miniyaml as miniyaml
@ -80,56 +80,9 @@ class Header(object):
for line in filter(None, lines):
output_filehandle.write(line + "\n")
def filter_master_only_jobs(items):
def _for_all_items(items, functor) -> None:
if isinstance(items, list):
for item in items:
_for_all_items(item, functor)
if isinstance(items, dict) and len(items) == 1:
item_type, item = next(iter(items.items()))
functor(item_type, item)
def _is_master_item(item):
filters = item.get('filters', None)
branches = filters.get('branches', None) if filters is not None else None
branches_only = branches.get('only', None) if branches is not None else None
return 'master' in branches_only if branches_only is not None else False
master_deps = set()
def _save_requires_if_master(item_type, item):
requires = item.get('requires', None)
item_name = item.get("name", None)
if not isinstance(requires, list):
return
if _is_master_item(item) or item_name in master_deps:
master_deps.update([n.strip('"') for n in requires])
def _do_filtering(items):
if isinstance(items, list):
rc = [_do_filtering(item) for item in items]
return [item for item in rc if len(item if item is not None else []) > 0]
assert isinstance(items, dict) and len(items) == 1
item_type, item = next(iter(items.items()))
item_name = item.get("name", None)
item_name = item_name.strip('"') if item_name is not None else None
if not _is_master_item(item) and item_name not in master_deps:
return None
if 'filters' in item:
item = item.copy()
item.pop('filters')
return {item_type: item}
# Scan of dependencies twice to pick up nested required jobs
# I.e. jobs depending on jobs that master-only job depend on
_for_all_items(items, _save_requires_if_master)
_for_all_items(items, _save_requires_if_master)
return _do_filtering(items)
def gen_build_workflows_tree():
build_workflows_functions = [
cimodel.data.simple.docker_definitions.get_workflow_jobs,
pytorch_build_definitions.get_workflow_jobs,
cimodel.data.simple.macos_definitions.get_workflow_jobs,
cimodel.data.simple.android_definitions.get_workflow_jobs,
@ -137,36 +90,45 @@ def gen_build_workflows_tree():
cimodel.data.simple.mobile_definitions.get_workflow_jobs,
cimodel.data.simple.ge_config_tests.get_workflow_jobs,
cimodel.data.simple.bazel_definitions.get_workflow_jobs,
caffe2_build_definitions.get_workflow_jobs,
cimodel.data.simple.binary_smoketest.get_workflow_jobs,
cimodel.data.simple.nightly_ios.get_workflow_jobs,
cimodel.data.simple.nightly_android.get_workflow_jobs,
cimodel.data.simple.anaconda_prune_defintions.get_workflow_jobs,
windows_build_definitions.get_windows_workflows,
binary_build_definitions.get_post_upload_jobs,
binary_build_definitions.get_binary_smoke_test_jobs,
]
binary_build_functions = [
binary_build_definitions.get_binary_build_jobs,
binary_build_definitions.get_nightly_tests,
binary_build_definitions.get_nightly_uploads,
binary_build_definitions.get_post_upload_jobs,
binary_build_definitions.get_binary_smoke_test_jobs,
]
build_jobs = [f() for f in build_workflows_functions]
master_build_jobs = filter_master_only_jobs(build_jobs)
docker_builder_functions = [
cimodel.data.simple.docker_definitions.get_workflow_jobs
]
return {
"workflows": {
"binary_builds": {
"when": r"<< pipeline.parameters.run_binary_tests >>",
"jobs": [f() for f in binary_build_functions],
},
"build": {
"when": r"<< pipeline.parameters.run_build >>",
"jobs": build_jobs,
},
"master_build": {
"when": r"<< pipeline.parameters.run_master_build >>",
"jobs": master_build_jobs,
},
"docker_build": OrderedDict(
{
"triggers": [
{
"schedule": {
"cron": miniutils.quote("0 15 * * 0"),
"filters": {"branches": {"only": ["master"]}},
}
}
],
"jobs": [f() for f in docker_builder_functions],
}
),
"build": {"jobs": [f() for f in build_workflows_functions]},
}
}
@ -178,10 +140,12 @@ YAML_SOURCES = [
File("nightly-binary-build-defaults.yml"),
Header("Build parameters"),
File("build-parameters/pytorch-build-params.yml"),
File("build-parameters/caffe2-build-params.yml"),
File("build-parameters/binary-build-params.yml"),
File("build-parameters/promote-build-params.yml"),
Header("Job specs"),
File("job-specs/pytorch-job-specs.yml"),
File("job-specs/caffe2-job-specs.yml"),
File("job-specs/binary-job-specs.yml"),
File("job-specs/job-specs-custom.yml"),
File("job-specs/job-specs-promote.yml"),
@ -190,7 +154,6 @@ YAML_SOURCES = [
File("job-specs/docker_jobs.yml"),
Header("Workflows"),
Treegen(gen_build_workflows_tree, 0),
File("workflows/workflows-scheduled-ci.yml"),
File("workflows/workflows-ecr-gc.yml"),
File("workflows/workflows-promote.yml"),
]

View File

@ -1,5 +0,0 @@
cd $PSScriptRoot;
$NewFile = New-TemporaryFile;
python generate_config_yml.py > $NewFile.name
(Get-Content $NewFile.name -Raw).TrimEnd().Replace("`r`n","`n") | Set-Content config.yml -Force
Remove-Item $NewFile.name

View File

@ -1,17 +1,8 @@
#!/bin/bash -e
#!/bin/bash -xe
# Allows this script to be invoked from any directory:
cd "$(dirname "$0")"
UNCOMMIT_CHANGE=$(git status -s | grep " config.yml" | wc -l | xargs)
if [[ $UNCOMMIT_CHANGE != 0 ]]; then
OLD_FILE=$(mktemp)
cp config.yml "$OLD_FILE"
echo "Uncommitted change detected in .circleci/config.yml"
echo "It has been backed up to $OLD_FILE"
fi
cd $(dirname "$0")
NEW_FILE=$(mktemp)
./generate_config_yml.py > "$NEW_FILE"
cp "$NEW_FILE" config.yml
echo "New config generated in .circleci/config.yml"
./generate_config_yml.py > $NEW_FILE
cp $NEW_FILE config.yml

View File

@ -33,11 +33,6 @@ else
export BUILDER_ROOT="$workdir/builder"
fi
# Try to extract PR number from branch if not already set
if [[ -z "${CIRCLE_PR_NUMBER:-}" ]]; then
CIRCLE_PR_NUMBER="$(echo ${CIRCLE_BRANCH} | sed -E -n 's/pull\/([0-9]*).*/\1/p')"
fi
# Clone the Pytorch branch
retry git clone https://github.com/pytorch/pytorch.git "$PYTORCH_ROOT"
pushd "$PYTORCH_ROOT"
@ -63,7 +58,6 @@ popd
# Clone the Builder master repo
retry git clone -q https://github.com/pytorch/builder.git "$BUILDER_ROOT"
pushd "$BUILDER_ROOT"
git checkout release/1.9
echo "Using builder from "
git --no-pager log --max-count 1
popd

View File

@ -15,8 +15,7 @@ export PATH="~/anaconda/bin:${PATH}"
source ~/anaconda/bin/activate
# Install dependencies
conda install numpy ninja pyyaml mkl mkl-include setuptools cmake cffi requests typing_extensions --yes
conda install -c conda-forge valgrind --yes
conda install numpy ninja pyyaml mkl mkl-include setuptools cmake cffi typing requests --yes
export CMAKE_PREFIX_PATH=${CONDA_PREFIX:-"$(dirname $(which conda))/../"}
# sync submodules

View File

@ -13,7 +13,7 @@ base64 --decode cert.txt -o Certificates.p12
rm cert.txt
bundle exec fastlane install_cert
# install the provisioning profile
PROFILE=PyTorch_CI_2021.mobileprovision
PROFILE=TestApp_CI.mobileprovision
PROVISIONING_PROFILES=~/Library/MobileDevice/Provisioning\ Profiles
mkdir -pv "${PROVISIONING_PROFILES}"
cd "${PROVISIONING_PROFILES}"
@ -24,6 +24,6 @@ rm cert.txt
if ! [ -x "$(command -v xcodebuild)" ]; then
echo 'Error: xcodebuild is not installed.'
exit 1
fi
PROFILE=PyTorch_CI_2021
fi
PROFILE=TestApp_CI
ruby ${PROJ_ROOT}/scripts/xcode_build.rb -i ${PROJ_ROOT}/build_ios/install -x ${PROJ_ROOT}/ios/TestApp/TestApp.xcodeproj -p ${IOS_PLATFORM} -c ${PROFILE} -t ${IOS_DEV_TEAM_ID}

View File

@ -34,13 +34,7 @@ touch version.txt
echo $(date +%s) > version.txt
zip -r ${ZIPFILE} install src version.txt LICENSE
# upload to aws
# Install conda then 'conda install' awscli
curl --retry 3 -o ~/conda.sh https://repo.anaconda.com/miniconda/Miniconda3-latest-MacOSX-x86_64.sh
chmod +x ~/conda.sh
/bin/bash ~/conda.sh -b -p ~/anaconda
export PATH="~/anaconda/bin:${PATH}"
source ~/anaconda/bin/activate
conda install -c conda-forge awscli --yes
brew install awscli
set +x
export AWS_ACCESS_KEY_ID=${AWS_S3_ACCESS_KEY_FOR_PYTORCH_BINARY_UPLOAD}
export AWS_SECRET_ACCESS_KEY=${AWS_S3_ACCESS_SECRET_FOR_PYTORCH_BINARY_UPLOAD}

View File

@ -5,22 +5,26 @@ set -eux -o pipefail
source /env
# Defaults here so they can be changed in one place
export MAX_JOBS=${MAX_JOBS:-$(( $(nproc) - 2 ))}
if [[ "${DESIRED_CUDA}" == "cu111" ]]; then
export BUILD_SPLIT_CUDA="ON"
fi
export MAX_JOBS=12
# Parse the parameters
if [[ "$PACKAGE_TYPE" == 'conda' ]]; then
build_script='conda/build_pytorch.sh'
elif [[ "$DESIRED_CUDA" == cpu ]]; then
build_script='manywheel/build_cpu.sh'
elif [[ "$DESIRED_CUDA" == *"rocm"* ]]; then
build_script='manywheel/build_rocm.sh'
else
build_script='manywheel/build.sh'
fi
# We want to call unbuffer, which calls tclsh which finds the expect
# package. The expect was installed by yum into /usr/bin so we want to
# find /usr/bin/tclsh, but this is shadowed by /opt/conda/bin/tclsh in
# the conda docker images, so we prepend it to the path here.
if [[ "$PACKAGE_TYPE" == 'conda' ]]; then
mkdir /just_tclsh_bin
ln -s /usr/bin/tclsh /just_tclsh_bin/tclsh
export PATH=/just_tclsh_bin:$PATH
fi
# Build the package
SKIP_ALL_TESTS=1 "/builder/$build_script"
SKIP_ALL_TESTS=1 unbuffer "/builder/$build_script" | ts

View File

@ -5,13 +5,12 @@ cat >/home/circleci/project/ci_test_script.sh <<EOL
# =================== The following code will be executed inside Docker container ===================
set -eux -o pipefail
python_nodot="\$(echo $DESIRED_PYTHON | tr -d m.u)"
# Set up Python
if [[ "$PACKAGE_TYPE" == conda ]]; then
retry conda create -qyn testenv python="$DESIRED_PYTHON"
source activate testenv >/dev/null
elif [[ "$PACKAGE_TYPE" != libtorch ]]; then
python_nodot="\$(echo $DESIRED_PYTHON | tr -d m.u)"
python_path="/opt/python/cp\$python_nodot-cp\${python_nodot}"
# Prior to Python 3.8 paths were suffixed with an 'm'
if [[ -d "\${python_path}/bin" ]]; then
@ -21,23 +20,6 @@ elif [[ "$PACKAGE_TYPE" != libtorch ]]; then
fi
fi
EXTRA_CONDA_FLAGS=""
NUMPY_PIN=""
if [[ "\$python_nodot" = *39* ]]; then
EXTRA_CONDA_FLAGS="-c=conda-forge"
# There's an issue with conda channel priority where it'll randomly pick 1.19 over 1.20
# we set a lower boundary here just to be safe
NUMPY_PIN=">=1.20"
fi
if [[ "$DESIRED_CUDA" == "cu112" ]]; then
EXTRA_CONDA_FLAGS="-c=conda-forge"
fi
# Move debug wheels out of the the package dir so they don't get installed
mkdir -p /tmp/debug_final_pkgs
mv /final_pkgs/debug-*.zip /tmp/debug_final_pkgs || echo "no debug packages to move"
# Install the package
# These network calls should not have 'retry's because they are installing
# locally and aren't actually network calls
@ -46,37 +28,23 @@ mv /final_pkgs/debug-*.zip /tmp/debug_final_pkgs || echo "no debug packages to m
# conda build scripts themselves. These should really be consolidated
pkg="/final_pkgs/\$(ls /final_pkgs)"
if [[ "$PACKAGE_TYPE" == conda ]]; then
(
# For some reason conda likes to re-activate the conda environment when attempting this install
# which means that a deactivate is run and some variables might not exist when that happens,
# namely CONDA_MKL_INTERFACE_LAYER_BACKUP from libblas so let's just ignore unbound variables when
# it comes to the conda installation commands
set +u
retry conda install \${EXTRA_CONDA_FLAGS} -yq \
"numpy\${NUMPY_PIN}" \
future \
mkl>=2018 \
ninja \
dataclasses \
typing-extensions \
defaults::protobuf \
six
if [[ "$DESIRED_CUDA" == 'cpu' ]]; then
retry conda install -c pytorch -y cpuonly
conda install -y "\$pkg" --offline
if [[ "$DESIRED_CUDA" == 'cpu' ]]; then
retry conda install -y cpuonly -c pytorch
fi
retry conda install -yq future numpy protobuf six
if [[ "$DESIRED_CUDA" != 'cpu' ]]; then
# DESIRED_CUDA is in format cu90 or cu102
if [[ "${#DESIRED_CUDA}" == 4 ]]; then
cu_ver="${DESIRED_CUDA:2:1}.${DESIRED_CUDA:3}"
else
# DESIRED_CUDA is in format cu90 or cu102
if [[ "${#DESIRED_CUDA}" == 4 ]]; then
cu_ver="${DESIRED_CUDA:2:1}.${DESIRED_CUDA:3}"
else
cu_ver="${DESIRED_CUDA:2:2}.${DESIRED_CUDA:4}"
fi
retry conda install \${EXTRA_CONDA_FLAGS} -yq -c nvidia -c pytorch "cudatoolkit=\${cu_ver}"
cu_ver="${DESIRED_CUDA:2:2}.${DESIRED_CUDA:4}"
fi
conda install \${EXTRA_CONDA_FLAGS} -y "\$pkg" --offline
)
retry conda install -yq -c pytorch "cudatoolkit=\${cu_ver}"
fi
elif [[ "$PACKAGE_TYPE" != libtorch ]]; then
pip install "\$pkg"
retry pip install -q future numpy protobuf typing-extensions six
retry pip install -q future numpy protobuf six
fi
if [[ "$PACKAGE_TYPE" == libtorch ]]; then
pkg="\$(ls /final_pkgs/*-latest.zip)"

View File

@ -0,0 +1,48 @@
#!/bin/bash
# Do NOT set -x
source /home/circleci/project/env
set -eu -o pipefail
set +x
declare -x "AWS_ACCESS_KEY_ID=${PYTORCH_BINARY_AWS_ACCESS_KEY_ID}"
declare -x "AWS_SECRET_ACCESS_KEY=${PYTORCH_BINARY_AWS_SECRET_ACCESS_KEY}"
#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!
# DO NOT TURN -x ON BEFORE THIS LINE
#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!
set -eux -o pipefail
export PATH="$MINICONDA_ROOT/bin:$PATH"
# This gets set in binary_populate_env.sh, but lets have a sane default just in case
PIP_UPLOAD_FOLDER=${PIP_UPLOAD_FOLDER:-nightly}
# TODO: Combine CONDA_UPLOAD_CHANNEL and PIP_UPLOAD_FOLDER into one variable
# The only difference is the trailing slash
# Strip trailing slashes if there
CONDA_UPLOAD_CHANNEL=$(echo "${PIP_UPLOAD_FOLDER}" | sed 's:/*$::')
BACKUP_BUCKET="s3://pytorch-backup"
retry pip install -q awscli
# Upload the package to the final location
pushd /home/circleci/project/final_pkgs
if [[ "$PACKAGE_TYPE" == conda ]]; then
retry conda install -yq anaconda-client
retry anaconda -t "${CONDA_PYTORCHBOT_TOKEN}" upload "$(ls)" -u "pytorch-${CONDA_UPLOAD_CHANNEL}" --label main --no-progress --force
# Fetch platform (eg. win-64, linux-64, etc.) from index file
# Because there's no actual conda command to read this
subdir=$(tar -xOf ./*.bz2 info/index.json | grep subdir | cut -d ':' -f2 | sed -e 's/[[:space:]]//' -e 's/"//g' -e 's/,//')
BACKUP_DIR="conda/${subdir}"
elif [[ "$PACKAGE_TYPE" == libtorch ]]; then
s3_dir="s3://pytorch/libtorch/${PIP_UPLOAD_FOLDER}${DESIRED_CUDA}/"
for pkg in $(ls); do
retry aws s3 cp "$pkg" "$s3_dir" --acl public-read
done
BACKUP_DIR="libtorch/${PIP_UPLOAD_FOLDER}${DESIRED_CUDA}/"
else
s3_dir="s3://pytorch/whl/${PIP_UPLOAD_FOLDER}${DESIRED_CUDA}/"
retry aws s3 cp "$(ls)" "$s3_dir" --acl public-read
BACKUP_DIR="whl/${PIP_UPLOAD_FOLDER}${DESIRED_CUDA}/"
fi
if [[ -n "${CIRCLE_TAG:-}" ]]; then
s3_dir="${BACKUP_BUCKET}/${CIRCLE_TAG}/${BACKUP_DIR}"
retry aws s3 cp --recursive . "$s3_dir"
fi

View File

@ -20,9 +20,9 @@ if [[ "$PACKAGE_TYPE" == libtorch ]]; then
unzip "$pkg" -d /tmp
cd /tmp/libtorch
elif [[ "$PACKAGE_TYPE" == conda ]]; then
conda install -y "$pkg"
conda install -y "$pkg" --offline
else
pip install "$pkg" -v
pip install "$pkg" --no-index --no-dependencies -v
fi
# Test

View File

@ -0,0 +1,48 @@
#!/bin/bash
# Do NOT set -x
set -eu -o pipefail
set +x
export AWS_ACCESS_KEY_ID="${PYTORCH_BINARY_AWS_ACCESS_KEY_ID}"
export AWS_SECRET_ACCESS_KEY="${PYTORCH_BINARY_AWS_SECRET_ACCESS_KEY}"
#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!
# DO NOT TURN -x ON BEFORE THIS LINE
#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!
set -eux -o pipefail
source "/Users/distiller/project/env"
export "PATH=$workdir/miniconda/bin:$PATH"
# This gets set in binary_populate_env.sh, but lets have a sane default just in case
PIP_UPLOAD_FOLDER=${PIP_UPLOAD_FOLDER:-nightly}
# TODO: Combine CONDA_UPLOAD_CHANNEL and PIP_UPLOAD_FOLDER into one variable
# The only difference is the trailing slash
# Strip trailing slashes if there
CONDA_UPLOAD_CHANNEL=$(echo "${PIP_UPLOAD_FOLDER}" | sed 's:/*$::')
BACKUP_BUCKET="s3://pytorch-backup"
retry pip install -q awscli
pushd "$workdir/final_pkgs"
if [[ "$PACKAGE_TYPE" == conda ]]; then
retry conda install -yq anaconda-client
retry anaconda -t "${CONDA_PYTORCHBOT_TOKEN}" upload "$(ls)" -u "pytorch-${CONDA_UPLOAD_CHANNEL}" --label main --no-progress --force
# Fetch platform (eg. win-64, linux-64, etc.) from index file
# Because there's no actual conda command to read this
subdir=$(tar -xOf ./*.bz2 info/index.json | grep subdir | cut -d ':' -f2 | sed -e 's/[[:space:]]//' -e 's/"//g' -e 's/,//')
BACKUP_DIR="conda/${subdir}"
elif [[ "$PACKAGE_TYPE" == libtorch ]]; then
s3_dir="s3://pytorch/libtorch/${PIP_UPLOAD_FOLDER}${DESIRED_CUDA}/"
for pkg in $(ls); do
retry aws s3 cp "$pkg" "$s3_dir" --acl public-read
done
BACKUP_DIR="libtorch/${PIP_UPLOAD_FOLDER}${DESIRED_CUDA}/"
else
s3_dir="s3://pytorch/whl/${PIP_UPLOAD_FOLDER}${DESIRED_CUDA}/"
retry aws s3 cp "$(ls)" "$s3_dir" --acl public-read
BACKUP_DIR="whl/${PIP_UPLOAD_FOLDER}${DESIRED_CUDA}/"
fi
if [[ -n "${CIRCLE_TAG:-}" ]]; then
s3_dir="${BACKUP_BUCKET}/${CIRCLE_TAG}/${BACKUP_DIR}"
retry aws s3 cp --recursive . "$s3_dir"
fi

View File

@ -68,24 +68,12 @@ if [[ -z "$DOCKER_IMAGE" ]]; then
fi
fi
USE_GOLD_LINKER="OFF"
# GOLD linker can not be used if CUPTI is statically linked into PyTorch, see https://github.com/pytorch/pytorch/issues/57744
if [[ ${DESIRED_CUDA} == "cpu" ]]; then
USE_GOLD_LINKER="ON"
fi
USE_WHOLE_CUDNN="OFF"
# Link whole cuDNN for CUDA-11.1 to include fp16 fast kernels
if [[ "$(uname)" == "Linux" && "${DESIRED_CUDA}" == "cu111" ]]; then
USE_WHOLE_CUDNN="ON"
fi
# Default to nightly, since that's where this normally uploads to
PIP_UPLOAD_FOLDER='nightly/'
# We put this here so that OVERRIDE_PACKAGE_VERSION below can read from it
export DATE="$(date -u +%Y%m%d)"
#TODO: We should be pulling semver version from the base version.txt
BASE_BUILD_VERSION="1.9.0.dev$DATE"
BASE_BUILD_VERSION="1.6.0.dev$DATE"
# Change BASE_BUILD_VERSION to git tag when on a git tag
# Use 'git -C' to make doubly sure we're in the correct directory for checking
# the git tag
@ -97,7 +85,7 @@ if tagged_version >/dev/null; then
# Turns tag v1.6.0-rc1 -> v1.6.0
BASE_BUILD_VERSION="$(tagged_version | sed -e 's/^v//' -e 's/-.*$//')"
fi
if [[ "$(uname)" == 'Darwin' ]] || [[ "$PACKAGE_TYPE" == conda ]]; then
if [[ "$(uname)" == 'Darwin' ]] || [[ "$DESIRED_CUDA" == "cu102" ]] || [[ "$PACKAGE_TYPE" == conda ]]; then
export PYTORCH_BUILD_VERSION="${BASE_BUILD_VERSION}"
else
export PYTORCH_BUILD_VERSION="${BASE_BUILD_VERSION}+$DESIRED_CUDA"
@ -112,14 +100,8 @@ if [[ "$PACKAGE_TYPE" == libtorch ]]; then
POSSIBLE_JAVA_HOMES+=(/usr/local)
POSSIBLE_JAVA_HOMES+=(/usr/lib/jvm/java-8-openjdk-amd64)
POSSIBLE_JAVA_HOMES+=(/Library/Java/JavaVirtualMachines/*.jdk/Contents/Home)
# Add the Windows-specific JNI path
POSSIBLE_JAVA_HOMES+=("$PWD/.circleci/windows-jni/")
for JH in "${POSSIBLE_JAVA_HOMES[@]}" ; do
if [[ -e "$JH/include/jni.h" ]] ; then
# Skip if we're not on Windows but haven't found a JAVA_HOME
if [[ "$JH" == "$PWD/.circleci/windows-jni/" && "$OSTYPE" != "msys" ]] ; then
break
fi
echo "Found jni.h under $JH"
JAVA_HOME="$JH"
BUILD_JNI=ON
@ -148,7 +130,7 @@ if [[ "${BUILD_FOR_SYSTEM:-}" == "windows" ]]; then
fi
export DATE="$DATE"
export NIGHTLIES_DATE_PREAMBLE=1.9.0.dev
export NIGHTLIES_DATE_PREAMBLE=1.6.0.dev
export PYTORCH_BUILD_VERSION="$PYTORCH_BUILD_VERSION"
export PYTORCH_BUILD_NUMBER="$PYTORCH_BUILD_NUMBER"
export OVERRIDE_PACKAGE_VERSION="$PYTORCH_BUILD_VERSION"
@ -179,11 +161,6 @@ export CIRCLE_TAG="${CIRCLE_TAG:-}"
export CIRCLE_SHA1="$CIRCLE_SHA1"
export CIRCLE_PR_NUMBER="${CIRCLE_PR_NUMBER:-}"
export CIRCLE_BRANCH="$CIRCLE_BRANCH"
export CIRCLE_WORKFLOW_ID="$CIRCLE_WORKFLOW_ID"
export USE_GOLD_LINKER="${USE_GOLD_LINKER}"
export USE_GLOO_WITH_OPENSSL="ON"
export USE_WHOLE_CUDNN="${USE_WHOLE_CUDNN}"
# =================== The above code will be executed inside Docker container ===================
EOL

View File

@ -19,7 +19,7 @@ chmod +x /home/circleci/project/ci_test_script.sh
VOLUME_MOUNTS="-v /home/circleci/project/:/circleci_stuff -v /home/circleci/project/final_pkgs:/final_pkgs -v ${PYTORCH_ROOT}:/pytorch -v ${BUILDER_ROOT}:/builder"
# Run the docker
if [ -n "${USE_CUDA_DOCKER_RUNTIME:-}" ]; then
export id=$(docker run --cap-add=SYS_PTRACE --security-opt seccomp=unconfined --gpus all ${VOLUME_MOUNTS} -t -d "${DOCKER_IMAGE}")
export id=$(docker run --cap-add=SYS_PTRACE --security-opt seccomp=unconfined --runtime=nvidia ${VOLUME_MOUNTS} -t -d "${DOCKER_IMAGE}")
else
export id=$(docker run --cap-add=SYS_PTRACE --security-opt seccomp=unconfined ${VOLUME_MOUNTS} -t -d "${DOCKER_IMAGE}")
fi

View File

@ -1,98 +0,0 @@
#!/usr/bin/env bash
set -euo pipefail
PACKAGE_TYPE=${PACKAGE_TYPE:-conda}
PKG_DIR=${PKG_DIR:-/tmp/workspace/final_pkgs}
# Designates whether to submit as a release candidate or a nightly build
# Value should be `test` when uploading release candidates
# currently set within `designate_upload_channel`
UPLOAD_CHANNEL=${UPLOAD_CHANNEL:-nightly}
# Designates what subfolder to put packages into
UPLOAD_SUBFOLDER=${UPLOAD_SUBFOLDER:-cpu}
UPLOAD_BUCKET="s3://pytorch"
BACKUP_BUCKET="s3://pytorch-backup"
DRY_RUN=${DRY_RUN:-enabled}
# Don't actually do work unless explicit
ANACONDA="true anaconda"
AWS_S3_CP="aws s3 cp --dryrun"
if [[ "${DRY_RUN}" = "disabled" ]]; then
ANACONDA="anaconda"
AWS_S3_CP="aws s3 cp"
fi
do_backup() {
local backup_dir
backup_dir=$1
(
pushd /tmp/workspace
set -x
${AWS_S3_CP} --recursive . "${BACKUP_BUCKET}/${CIRCLE_TAG}/${backup_dir}/"
)
}
conda_upload() {
(
set -x
${ANACONDA} \
upload \
${PKG_DIR}/*.tar.bz2 \
-u "pytorch-${UPLOAD_CHANNEL}" \
--label main \
--no-progress \
--force
)
}
s3_upload() {
local extension
local pkg_type
extension="$1"
pkg_type="$2"
s3_dir="${UPLOAD_BUCKET}/${pkg_type}/${UPLOAD_CHANNEL}/${UPLOAD_SUBFOLDER}/"
(
for pkg in ${PKG_DIR}/*.${extension}; do
(
set -x
${AWS_S3_CP} --no-progress --acl public-read "${pkg}" "${s3_dir}"
)
done
)
}
case "${PACKAGE_TYPE}" in
conda)
conda_upload
# Fetch platform (eg. win-64, linux-64, etc.) from index file
# Because there's no actual conda command to read this
subdir=$(\
tar -xOf ${PKG_DIR}/*.bz2 info/index.json \
| grep subdir \
| cut -d ':' -f2 \
| sed -e 's/[[:space:]]//' -e 's/"//g' -e 's/,//' \
)
BACKUP_DIR="conda/${subdir}"
;;
libtorch)
s3_upload "zip" "libtorch"
BACKUP_DIR="libtorch/${UPLOAD_CHANNEL}/${UPLOAD_SUBFOLDER}"
;;
# wheel can either refer to wheel/manywheel
*wheel)
s3_upload "whl" "whl"
BACKUP_DIR="whl/${UPLOAD_CHANNEL}/${UPLOAD_SUBFOLDER}"
;;
*)
echo "ERROR: unknown package type: ${PACKAGE_TYPE}"
exit 1
;;
esac
# CIRCLE_TAG is defined by upstream circleci,
# this can be changed to recognize tagged versions
if [[ -n "${CIRCLE_TAG:-}" ]]; then
do_backup "${BACKUP_DIR}"
fi

View File

@ -15,10 +15,6 @@ else
export VC_YEAR=2019
fi
if [[ "${DESIRED_CUDA}" == "cu111" ]]; then
export BUILD_SPLIT_CUDA="ON"
fi
set +x
export AWS_ACCESS_KEY_ID=${CIRCLECI_AWS_ACCESS_KEY_FOR_SCCACHE_S3_BUCKET_V4:-}
export AWS_SECRET_ACCESS_KEY=${CIRCLECI_AWS_SECRET_KEY_FOR_SCCACHE_S3_BUCKET_V4:-}
@ -31,10 +27,6 @@ if [[ "$CIRCLECI" == 'true' && -d "C:\\ProgramData\\Microsoft\\VisualStudio\\Pac
mv _Instances "C:\\ProgramData\\Microsoft\\VisualStudio\\Packages"
fi
if [[ "$CIRCLECI" == 'true' && -d "C:\\Microsoft" ]]; then
rm -rf "C:\\Microsoft\\Android*"
fi
echo "Free space on filesystem before build:"
df -h

View File

@ -0,0 +1,47 @@
#!/bin/bash
set -eu -o pipefail
set +x
declare -x "AWS_ACCESS_KEY_ID=${PYTORCH_BINARY_AWS_ACCESS_KEY_ID}"
declare -x "AWS_SECRET_ACCESS_KEY=${PYTORCH_BINARY_AWS_SECRET_ACCESS_KEY}"
#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!
# DO NOT TURN -x ON BEFORE THIS LINE
#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!#!
set -eux -o pipefail
source "/env"
# This gets set in binary_populate_env.sh, but lets have a sane default just in case
PIP_UPLOAD_FOLDER=${PIP_UPLOAD_FOLDER:-nightly/}
# TODO: Combine CONDA_UPLOAD_CHANNEL and PIP_UPLOAD_FOLDER into one variable
# The only difference is the trailing slash
# Strip trailing slashes if there
CONDA_UPLOAD_CHANNEL=$(echo "${PIP_UPLOAD_FOLDER}" | sed 's:/*$::')
BACKUP_BUCKET="s3://pytorch-backup"
retry pip install -q awscli
pushd /root/workspace/final_pkgs
# Upload the package to the final location
if [[ "$PACKAGE_TYPE" == conda ]]; then
retry conda install -yq anaconda-client
retry anaconda -t "${CONDA_PYTORCHBOT_TOKEN}" upload "$(ls)" -u "pytorch-${CONDA_UPLOAD_CHANNEL}" --label main --no-progress --force
# Fetch platform (eg. win-64, linux-64, etc.) from index file
# Because there's no actual conda command to read this
subdir=$(tar -xOf ./*.bz2 info/index.json | grep subdir | cut -d ':' -f2 | sed -e 's/[[:space:]]//' -e 's/"//g' -e 's/,//')
BACKUP_DIR="conda/${subdir}"
elif [[ "$PACKAGE_TYPE" == libtorch ]]; then
s3_dir="s3://pytorch/libtorch/${PIP_UPLOAD_FOLDER}${DESIRED_CUDA}/"
for pkg in $(ls); do
retry aws s3 cp "$pkg" "$s3_dir" --acl public-read
done
BACKUP_DIR="libtorch/${PIP_UPLOAD_FOLDER}${DESIRED_CUDA}/"
else
s3_dir="s3://pytorch/whl/${PIP_UPLOAD_FOLDER}${DESIRED_CUDA}/"
retry aws s3 cp "$(ls)" "$s3_dir" --acl public-read
BACKUP_DIR="whl/${PIP_UPLOAD_FOLDER}${DESIRED_CUDA}/"
fi
if [[ -n "${CIRCLE_TAG:-}" ]]; then
s3_dir="${BACKUP_BUCKET}/${CIRCLE_TAG}/${BACKUP_DIR}"
retry aws s3 cp --recursive . "$s3_dir"
fi

View File

@ -1,44 +1,15 @@
#!/usr/bin/env bash
set -eux -o pipefail
env
echo "BUILD_ENVIRONMENT:$BUILD_ENVIRONMENT"
export ANDROID_NDK_HOME=/opt/ndk
export ANDROID_NDK=/opt/ndk
export ANDROID_HOME=/opt/android/sdk
# Must be in sync with GRADLE_VERSION in docker image for android
# https://github.com/pietern/pytorch-dockerfiles/blob/master/build.sh#L155
export GRADLE_VERSION=6.8.3
export GRADLE_VERSION=4.10.3
export GRADLE_HOME=/opt/gradle/gradle-$GRADLE_VERSION
export GRADLE_PATH=$GRADLE_HOME/bin/gradle
# touch gradle cache files to prevent expiration
while IFS= read -r -d '' file
do
touch "$file" || true
done < <(find /var/lib/jenkins/.gradle -type f -print0)
export GRADLE_LOCAL_PROPERTIES=~/workspace/android/local.properties
rm -f $GRADLE_LOCAL_PROPERTIES
echo "sdk.dir=/opt/android/sdk" >> $GRADLE_LOCAL_PROPERTIES
echo "ndk.dir=/opt/ndk" >> $GRADLE_LOCAL_PROPERTIES
echo "cmake.dir=/usr/local" >> $GRADLE_LOCAL_PROPERTIES
retry () {
$* || (sleep 1 && $*) || (sleep 2 && $*) || (sleep 4 && $*) || (sleep 8 && $*)
}
# Run custom build script
if [[ "${BUILD_ENVIRONMENT}" == *-gradle-custom-build* ]]; then
# Install torch & torchvision - used to download & dump used ops from test model.
retry pip install torch torchvision --progress-bar off
exec "$(dirname "${BASH_SOURCE[0]}")/../../android/build_test_app_custom.sh" armeabi-v7a
fi
# Run default build
BUILD_ANDROID_INCLUDE_DIR_x86=~/workspace/build_android/install/include
BUILD_ANDROID_LIB_DIR_x86=~/workspace/build_android/install/lib
@ -73,6 +44,9 @@ ln -s ${BUILD_ANDROID_INCLUDE_DIR_arm_v8a} ${JNI_INCLUDE_DIR}/arm64-v8a
ln -s ${BUILD_ANDROID_LIB_DIR_arm_v8a} ${JNI_LIBS_DIR}/arm64-v8a
fi
env
echo "BUILD_ENVIRONMENT:$BUILD_ENVIRONMENT"
GRADLE_PARAMS="-p android assembleRelease --debug --stacktrace"
if [[ "${BUILD_ENVIRONMENT}" == *-gradle-build-only-x86_32* ]]; then
GRADLE_PARAMS+=" -PABI_FILTERS=x86"
@ -82,6 +56,20 @@ if [ -n "{GRADLE_OFFLINE:-}" ]; then
GRADLE_PARAMS+=" --offline"
fi
# touch gradle cache files to prevent expiration
while IFS= read -r -d '' file
do
touch "$file" || true
done < <(find /var/lib/jenkins/.gradle -type f -print0)
env
export GRADLE_LOCAL_PROPERTIES=~/workspace/android/local.properties
rm -f $GRADLE_LOCAL_PROPERTIES
echo "sdk.dir=/opt/android/sdk" >> $GRADLE_LOCAL_PROPERTIES
echo "ndk.dir=/opt/ndk" >> $GRADLE_LOCAL_PROPERTIES
echo "cmake.dir=/usr/local" >> $GRADLE_LOCAL_PROPERTIES
$GRADLE_PATH $GRADLE_PARAMS
find . -type f -name "*.a" -exec ls -lh {} \;

View File

@ -30,7 +30,13 @@ if [ "$version" == "master" ]; then
is_master_doc=true
fi
echo "install_path: $install_path version: $version"
# Argument 3: (optional) If present, we will NOT do any pushing. Used for testing.
dry_run=false
if [ "$3" != "" ]; then
dry_run=true
fi
echo "install_path: $install_path version: $version dry_run: $dry_run"
# ======================== Building PyTorch C++ API Docs ========================
@ -47,22 +53,31 @@ sudo apt-get -y install doxygen
# Generate ATen files
pushd "${pt_checkout}"
pip install -r requirements.txt
time python -m tools.codegen.gen \
time python aten/src/ATen/gen.py \
-s aten/src/ATen \
-d build/aten/src/ATen
-d build/aten/src/ATen \
aten/src/ATen/Declarations.cwrap \
aten/src/THCUNN/generic/THCUNN.h \
aten/src/ATen/nn.yaml \
aten/src/ATen/native/native_functions.yaml
# Copy some required files
cp aten/src/ATen/common_with_cwrap.py tools/shared/cwrap_common.py
cp torch/_utils_internal.py tools/shared
# Generate PyTorch files
time python tools/setup_helpers/generate_code.py \
--declarations-path build/aten/src/ATen/Declarations.yaml \
--native-functions-path aten/src/ATen/native/native_functions.yaml \
--nn-path aten/src/
# Build the docs
pushd docs/cpp
pip install -r requirements.txt
pip install breathe==4.13.0 bs4 lxml six
pip install --no-cache-dir -e "git+https://github.com/pytorch/pytorch_sphinx_theme.git#egg=pytorch_sphinx_theme"
pip install exhale>=0.2.1
pip install sphinx==2.4.4
# Uncomment once it is fixed
# pip install -r requirements.txt
time make VERBOSE=1 html -j
popd
@ -88,8 +103,24 @@ git status
git config user.email "soumith+bot@pytorch.org"
git config user.name "pytorchbot"
# If there aren't changes, don't make a commit; push is no-op
git commit -m "Generate C++ docs from pytorch/pytorch@$CIRCLE_SHA1" || true
git commit -m "Automatic sync on $(date)" || true
git status
if [ "$dry_run" = false ]; then
echo "Pushing to https://github.com/pytorch/cppdocs"
set +x
/usr/bin/expect <<DONE
spawn git push -u origin master
expect "Username*"
send "pytorchbot\n"
expect "Password*"
send "$::env(GITHUB_PYTORCHBOT_TOKEN)\n"
expect eof
DONE
set -x
else
echo "Skipping push due to dry_run"
fi
popd
# =================== The above code **should** be executed inside Docker container ===================

View File

@ -1,8 +0,0 @@
set "DRIVER_DOWNLOAD_LINK=https://s3.amazonaws.com/ossci-windows/452.39-data-center-tesla-desktop-win10-64bit-international.exe"
curl --retry 3 -kL %DRIVER_DOWNLOAD_LINK% --output 452.39-data-center-tesla-desktop-win10-64bit-international.exe
if errorlevel 1 exit /b 1
start /wait 452.39-data-center-tesla-desktop-win10-64bit-international.exe -s -noreboot
if errorlevel 1 exit /b 1
del 452.39-data-center-tesla-desktop-win10-64bit-international.exe || ver > NUL

View File

@ -5,7 +5,7 @@ set -eu -o pipefail
export ANDROID_NDK_HOME=/opt/ndk
export ANDROID_HOME=/opt/android/sdk
export GRADLE_VERSION=6.8.3
export GRADLE_VERSION=4.10.3
export GRADLE_HOME=/opt/gradle/gradle-$GRADLE_VERSION
export GRADLE_PATH=$GRADLE_HOME/bin/gradle
@ -35,9 +35,7 @@ else
echo "ndk.dir=/opt/ndk" >> $GRADLE_LOCAL_PROPERTIES
echo "SONATYPE_NEXUS_USERNAME=${SONATYPE_NEXUS_USERNAME}" >> $GRADLE_PROPERTIES
echo "mavenCentralRepositoryUsername=${SONATYPE_NEXUS_USERNAME}" >> $GRADLE_PROPERTIES
echo "SONATYPE_NEXUS_PASSWORD=${SONATYPE_NEXUS_PASSWORD}" >> $GRADLE_PROPERTIES
echo "mavenCentralRepositoryPassword=${SONATYPE_NEXUS_PASSWORD}" >> $GRADLE_PROPERTIES
echo "signing.keyId=${ANDROID_SIGN_KEY}" >> $GRADLE_PROPERTIES
echo "signing.password=${ANDROID_SIGN_PASS}" >> $GRADLE_PROPERTIES

View File

@ -7,8 +7,6 @@ sudo apt-get -y install expect-dev
# This is where the local pytorch install in the docker image is located
pt_checkout="/var/lib/jenkins/workspace"
source "$pt_checkout/.jenkins/pytorch/common_utils.sh"
echo "python_doc_push_script.sh: Invoked with $*"
set -ex
@ -40,30 +38,15 @@ echo "error: python_doc_push_script.sh: branch (arg3) not specified"
exit 1
fi
echo "install_path: $install_path version: $version"
# Argument 4: (optional) If present, we will NOT do any pushing. Used for testing.
dry_run=false
if [ "$4" != "" ]; then
dry_run=true
fi
echo "install_path: $install_path version: $version dry_run: $dry_run"
build_docs () {
set +e
set -o pipefail
make $1 2>&1 | tee /tmp/docs_build.txt
code=$?
if [ $code -ne 0 ]; then
set +x
echo =========================
grep "WARNING:" /tmp/docs_build.txt
echo =========================
echo Docs build failed. If the failure is not clear, scan back in the log
echo for any WARNINGS or for the line "build finished with problems"
echo "(tried to echo the WARNINGS above the ==== line)"
echo =========================
fi
set -ex
return $code
}
git clone https://github.com/pytorch/pytorch.github.io -b $branch --depth 1
git clone https://github.com/pytorch/pytorch.github.io -b $branch
pushd pytorch.github.io
export LC_ALL=C
@ -71,15 +54,26 @@ export PATH=/opt/conda/bin:$PATH
rm -rf pytorch || true
# Install TensorBoard in python 3 so torch.utils.tensorboard classes render
pip install -q https://s3.amazonaws.com/ossci-linux/wheels/tensorboard-1.14.0a0-py3-none-any.whl
# Get all the documentation sources, put them in one place
pushd "$pt_checkout"
git clone https://github.com/pytorch/vision
pushd vision
conda install -q pillow
time python setup.py install
popd
pushd docs
rm -rf source/torchvision
cp -a ../vision/docs/source source/torchvision
# Build the docs
pip -q install -r requirements.txt
pip -q install -r requirements.txt || true
if [ "$is_master_doc" = true ]; then
build_docs html
[ $? -eq 0 ] || exit $?
# TODO: fix gh-38011 then enable this which changes warnings into errors
# export SPHINXOPTS="-WT --keep-going"
make html
make coverage
# Now we have the coverage report, we need to make sure it is empty.
# Count the number of lines in the file and turn that number into a variable
@ -100,9 +94,8 @@ if [ "$is_master_doc" = true ]; then
exit 1
fi
else
# skip coverage, format for stable or tags
build_docs html-stable
[ $? -eq 0 ] || exit $?
# Don't fail the build on coverage problems
make html-stable
fi
# Move them into the docs repo
@ -111,6 +104,14 @@ popd
git rm -rf "$install_path" || true
mv "$pt_checkout/docs/build/html" "$install_path"
# Add the version handler by search and replace.
# XXX: Consider moving this to the docs Makefile or site build
if [ "$is_master_doc" = true ]; then
find "$install_path" -name "*.html" -print0 | xargs -0 perl -pi -w -e "s@master\s+\((\d\.\d\.[A-Fa-f0-9]+\+[A-Fa-f0-9]+)\s+\)@<a href='http://pytorch.org/docs/versions.html'>\1 \&#x25BC</a>@g"
else
find "$install_path" -name "*.html" -print0 | xargs -0 perl -pi -w -e "s@master\s+\((\d\.\d\.[A-Fa-f0-9]+\+[A-Fa-f0-9]+)\s+\)@<a href='http://pytorch.org/docs/versions.html'>$version \&#x25BC</a>@g"
fi
# Prevent Google from indexing $install_path/_modules. This folder contains
# generated source files.
# NB: the following only works on gnu sed. The sed shipped with mac os is different.
@ -122,8 +123,24 @@ git status
git config user.email "soumith+bot@pytorch.org"
git config user.name "pytorchbot"
# If there aren't changes, don't make a commit; push is no-op
git commit -m "Generate Python docs from pytorch/pytorch@$CIRCLE_SHA1" || true
git commit -m "auto-generating sphinx docs" || true
git status
if [ "$dry_run" = false ]; then
echo "Pushing to pytorch.github.io:$branch"
set +x
/usr/bin/expect <<DONE
spawn git push origin $branch
expect "Username*"
send "pytorchbot\n"
expect "Password*"
send "$::env(GITHUB_PYTORCHBOT_TOKEN)\n"
expect eof
DONE
set -x
else
echo "Skipping push due to dry_run"
fi
popd
# =================== The above code **should** be executed inside Docker container ===================

View File

@ -1,6 +1,12 @@
#!/usr/bin/env bash
set -ex -o pipefail
# Set up NVIDIA docker repo
curl -s -L --retry 3 https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
echo "deb https://nvidia.github.io/libnvidia-container/ubuntu16.04/amd64 /" | sudo tee -a /etc/apt/sources.list.d/nvidia-docker.list
echo "deb https://nvidia.github.io/nvidia-container-runtime/ubuntu16.04/amd64 /" | sudo tee -a /etc/apt/sources.list.d/nvidia-docker.list
echo "deb https://nvidia.github.io/nvidia-docker/ubuntu16.04/amd64 /" | sudo tee -a /etc/apt/sources.list.d/nvidia-docker.list
# Remove unnecessary sources
sudo rm -f /etc/apt/sources.list.d/google-chrome.list
sudo rm -f /etc/apt/heroku.list
@ -8,7 +14,7 @@ sudo rm -f /etc/apt/openjdk-r-ubuntu-ppa-xenial.list
sudo rm -f /etc/apt/partner.list
retry () {
$* || $* || $* || $* || $*
$* || $* || $* || $* || $*
}
# Method adapted from here: https://askubuntu.com/questions/875213/apt-get-to-retry-downloading
@ -16,84 +22,70 @@ retry () {
# This is better than retrying the whole apt-get command
echo "APT::Acquire::Retries \"3\";" | sudo tee /etc/apt/apt.conf.d/80-retries
retry sudo apt-get update -qq
sudo apt-get -y update
sudo apt-get -y remove linux-image-generic linux-headers-generic linux-generic docker-ce
# WARNING: Docker version is hardcoded here; you must update the
# version number below for docker-ce and nvidia-docker2 to get newer
# versions of Docker. We hardcode these numbers because we kept
# getting broken CI when Docker would update their docker version,
# and nvidia-docker2 would be out of date for a day until they
# released a newer version of their package.
#
# How to figure out what the correct versions of these packages are?
# My preferred method is to start a Docker instance of the correct
# Ubuntu version (e.g., docker run -it ubuntu:16.04) and then ask
# apt what the packages you need are. Note that the CircleCI image
# comes with Docker.
#
# Using 'retry' here as belt-and-suspenders even though we are
# presumably retrying at the single-package level via the
# apt.conf.d/80-retries technique.
retry sudo apt-get -y install \
linux-headers-$(uname -r) \
linux-image-generic \
moreutils \
docker-ce=5:18.09.4~3-0~ubuntu-xenial \
nvidia-container-runtime=2.0.0+docker18.09.4-1 \
nvidia-docker2=2.0.3+docker18.09.4-1 \
expect-dev
echo "== DOCKER VERSION =="
docker version
sudo pkill -SIGHUP dockerd
if ! command -v aws >/dev/null; then
retry sudo pip3 -q install awscli==1.19.64
fi
retry sudo pip -q install awscli==1.16.35
if [ -n "${USE_CUDA_DOCKER_RUNTIME:-}" ]; then
DRIVER_FN="NVIDIA-Linux-x86_64-460.39.run"
DRIVER_FN="NVIDIA-Linux-x86_64-440.59.run"
wget "https://s3.amazonaws.com/ossci-linux/nvidia_driver/$DRIVER_FN"
sudo /bin/bash "$DRIVER_FN" -s --no-drm || (sudo cat /var/log/nvidia-installer.log && false)
nvidia-smi
# Taken directly from https://github.com/NVIDIA/nvidia-docker
# Add the package repositories
distribution=$(. /etc/os-release;echo "$ID$VERSION_ID")
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
curl -s -L "https://nvidia.github.io/nvidia-docker/${distribution}/nvidia-docker.list" | sudo tee /etc/apt/sources.list.d/nvidia-docker.list
sudo apt-get update -qq
# Necessary to get the `--gpus` flag to function within docker
sudo apt-get install -y nvidia-container-toolkit
sudo systemctl restart docker
else
# Explicitly remove nvidia docker apt repositories if not building for cuda
sudo rm -rf /etc/apt/sources.list.d/nvidia-docker.list
fi
add_to_env_file() {
local name=$1
local value=$2
case "$value" in
*\ *)
# BASH_ENV should be set by CircleCI
echo "${name}='${value}'" >> "${BASH_ENV:-/tmp/env}"
;;
*)
echo "${name}=${value}" >> "${BASH_ENV:-/tmp/env}"
;;
esac
}
add_to_env_file IN_CI 1
add_to_env_file COMMIT_SOURCE "${CIRCLE_BRANCH:-}"
add_to_env_file BUILD_ENVIRONMENT "${BUILD_ENVIRONMENT}"
add_to_env_file CIRCLE_PULL_REQUEST "${CIRCLE_PULL_REQUEST}"
if [[ "${BUILD_ENVIRONMENT}" == *-build ]]; then
add_to_env_file SCCACHE_BUCKET ossci-compiler-cache-circleci-v2
SCCACHE_MAX_JOBS=$(( $(nproc) - 1 ))
MEMORY_LIMIT_MAX_JOBS=8 # the "large" resource class on CircleCI has 32 CPU cores, if we use all of them we'll OOM
MAX_JOBS=$(( ${SCCACHE_MAX_JOBS} > ${MEMORY_LIMIT_MAX_JOBS} ? ${MEMORY_LIMIT_MAX_JOBS} : ${SCCACHE_MAX_JOBS} ))
add_to_env_file MAX_JOBS "${MAX_JOBS}"
echo "declare -x IN_CIRCLECI=1" > /home/circleci/project/env
echo "declare -x COMMIT_SOURCE=${CIRCLE_BRANCH:-}" >> /home/circleci/project/env
echo "declare -x SCCACHE_BUCKET=ossci-compiler-cache-circleci-v2" >> /home/circleci/project/env
if [ -n "${USE_CUDA_DOCKER_RUNTIME:-}" ]; then
add_to_env_file TORCH_CUDA_ARCH_LIST 5.2
echo "declare -x TORCH_CUDA_ARCH_LIST=5.2" >> /home/circleci/project/env
fi
export SCCACHE_MAX_JOBS=`expr $(nproc) - 1`
export MEMORY_LIMIT_MAX_JOBS=8 # the "large" resource class on CircleCI has 32 CPU cores, if we use all of them we'll OOM
export MAX_JOBS=$(( ${SCCACHE_MAX_JOBS} > ${MEMORY_LIMIT_MAX_JOBS} ? ${MEMORY_LIMIT_MAX_JOBS} : ${SCCACHE_MAX_JOBS} ))
echo "declare -x MAX_JOBS=${MAX_JOBS}" >> /home/circleci/project/env
if [[ "${BUILD_ENVIRONMENT}" == *xla* ]]; then
# This IAM user allows write access to S3 bucket for sccache & bazels3cache
set +x
add_to_env_file XLA_CLANG_CACHE_S3_BUCKET_NAME "${XLA_CLANG_CACHE_S3_BUCKET_NAME:-}"
add_to_env_file AWS_ACCESS_KEY_ID "${CIRCLECI_AWS_ACCESS_KEY_FOR_SCCACHE_AND_XLA_BAZEL_S3_BUCKET_V2:-}"
add_to_env_file AWS_SECRET_ACCESS_KEY "${CIRCLECI_AWS_SECRET_KEY_FOR_SCCACHE_AND_XLA_BAZEL_S3_BUCKET_V2:-}"
echo "declare -x XLA_CLANG_CACHE_S3_BUCKET_NAME=${XLA_CLANG_CACHE_S3_BUCKET_NAME:-}" >> /home/circleci/project/env
echo "declare -x AWS_ACCESS_KEY_ID=${CIRCLECI_AWS_ACCESS_KEY_FOR_SCCACHE_AND_XLA_BAZEL_S3_BUCKET_V2:-}" >> /home/circleci/project/env
echo "declare -x AWS_SECRET_ACCESS_KEY=${CIRCLECI_AWS_SECRET_KEY_FOR_SCCACHE_AND_XLA_BAZEL_S3_BUCKET_V2:-}" >> /home/circleci/project/env
set -x
else
# This IAM user allows write access to S3 bucket for sccache
set +x
add_to_env_file XLA_CLANG_CACHE_S3_BUCKET_NAME "${XLA_CLANG_CACHE_S3_BUCKET_NAME:-}"
add_to_env_file AWS_ACCESS_KEY_ID "${CIRCLECI_AWS_ACCESS_KEY_FOR_SCCACHE_S3_BUCKET_V4:-}"
add_to_env_file AWS_SECRET_ACCESS_KEY "${CIRCLECI_AWS_SECRET_KEY_FOR_SCCACHE_S3_BUCKET_V4:-}"
echo "declare -x XLA_CLANG_CACHE_S3_BUCKET_NAME=${XLA_CLANG_CACHE_S3_BUCKET_NAME:-}" >> /home/circleci/project/env
echo "declare -x AWS_ACCESS_KEY_ID=${CIRCLECI_AWS_ACCESS_KEY_FOR_SCCACHE_S3_BUCKET_V4:-}" >> /home/circleci/project/env
echo "declare -x AWS_SECRET_ACCESS_KEY=${CIRCLECI_AWS_SECRET_KEY_FOR_SCCACHE_S3_BUCKET_V4:-}" >> /home/circleci/project/env
set -x
fi
fi
@ -102,7 +94,5 @@ fi
set +x
export AWS_ACCESS_KEY_ID=${CIRCLECI_AWS_ACCESS_KEY_FOR_ECR_READ_WRITE_V4:-}
export AWS_SECRET_ACCESS_KEY=${CIRCLECI_AWS_SECRET_KEY_FOR_ECR_READ_WRITE_V4:-}
export AWS_ACCOUNT_ID=$(aws sts get-caller-identity|grep Account|cut -f4 -d\")
export AWS_REGION=us-east-1
aws ecr get-login-password --region $AWS_REGION|docker login --username AWS --password-stdin $AWS_ACCOUNT_ID.dkr.ecr.$AWS_REGION.amazonaws.com
eval $(aws ecr get-login --region us-east-1 --no-include-email)
set -x

View File

@ -33,7 +33,7 @@ systemctl list-units --all | cat
sudo pkill apt-get || true
# For even better luck, purge unattended-upgrades
sudo apt-get purge -y unattended-upgrades || true
sudo apt-get purge -y unattended-upgrades
cat /etc/apt/sources.list

View File

@ -1,140 +0,0 @@
# Documentation: https://docs.microsoft.com/en-us/rest/api/azure/devops/build/?view=azure-devops-rest-6.0
import re
import json
import os
import sys
import requests
import time
AZURE_PIPELINE_BASE_URL = "https://aiinfra.visualstudio.com/PyTorch/"
AZURE_DEVOPS_PAT_BASE64 = os.environ.get("AZURE_DEVOPS_PAT_BASE64_SECRET", "")
PIPELINE_ID = "911"
PROJECT_ID = "0628bce4-2d33-499e-bac5-530e12db160f"
TARGET_BRANCH = os.environ.get("CIRCLE_BRANCH", "master")
TARGET_COMMIT = os.environ.get("CIRCLE_SHA1", "")
build_base_url = AZURE_PIPELINE_BASE_URL + "_apis/build/builds?api-version=6.0"
s = requests.Session()
s.headers.update({"Authorization": "Basic " + AZURE_DEVOPS_PAT_BASE64})
def submit_build(pipeline_id, project_id, source_branch, source_version):
print("Submitting build for branch: " + source_branch)
print("Commit SHA1: ", source_version)
run_build_raw = s.post(build_base_url, json={
"definition": {"id": pipeline_id},
"project": {"id": project_id},
"sourceBranch": source_branch,
"sourceVersion": source_version
})
try:
run_build_json = run_build_raw.json()
except json.decoder.JSONDecodeError as e:
print(e)
print("Failed to parse the response. Check if the Azure DevOps PAT is incorrect or expired.")
sys.exit(-1)
build_id = run_build_json['id']
print("Submitted bulid: " + str(build_id))
print("Bulid URL: " + run_build_json['url'])
return build_id
def get_build(_id):
get_build_url = AZURE_PIPELINE_BASE_URL + f"/_apis/build/builds/{_id}?api-version=6.0"
get_build_raw = s.get(get_build_url)
return get_build_raw.json()
def get_build_logs(_id):
get_build_logs_url = AZURE_PIPELINE_BASE_URL + f"/_apis/build/builds/{_id}/logs?api-version=6.0"
get_build_logs_raw = s.get(get_build_logs_url)
return get_build_logs_raw.json()
def get_log_content(url):
resp = s.get(url)
return resp.text
def wait_for_build(_id):
build_detail = get_build(_id)
build_status = build_detail['status']
while build_status == 'notStarted':
print('Waiting for run to start: ' + str(_id))
sys.stdout.flush()
try:
build_detail = get_build(_id)
build_status = build_detail['status']
except Exception as e:
print("Error getting build")
print(e)
time.sleep(30)
print("Bulid started: ", str(_id))
handled_logs = set()
while build_status == 'inProgress':
try:
print("Waiting for log: " + str(_id))
logs = get_build_logs(_id)
except Exception as e:
print("Error fetching logs")
print(e)
time.sleep(30)
continue
for log in logs['value']:
log_id = log['id']
if log_id in handled_logs:
continue
handled_logs.add(log_id)
print('Fetching log: \n' + log['url'])
try:
log_content = get_log_content(log['url'])
print(log_content)
except Exception as e:
print("Error getting log content")
print(e)
sys.stdout.flush()
build_detail = get_build(_id)
build_status = build_detail['status']
time.sleep(30)
build_result = build_detail['result']
print("Bulid status: " + build_status)
print("Bulid result: " + build_result)
return build_status, build_result
if __name__ == '__main__':
# Convert the branch name for Azure DevOps
match = re.search(r'pull/(\d+)', TARGET_BRANCH)
if match is not None:
pr_num = match.group(1)
SOURCE_BRANCH = f'refs/pull/{pr_num}/head'
else:
SOURCE_BRANCH = f'refs/heads/{TARGET_BRANCH}'
MAX_RETRY = 2
retry = MAX_RETRY
while retry > 0:
build_id = submit_build(PIPELINE_ID, PROJECT_ID, SOURCE_BRANCH, TARGET_COMMIT)
build_status, build_result = wait_for_build(build_id)
if build_result != 'succeeded':
retry = retry - 1
if retry > 0:
print("Retrying... remaining attempt: " + str(retry))
# Wait a bit before retrying
time.sleep((MAX_RETRY - retry) * 120)
continue
else:
print("No more chance to retry. Giving up.")
sys.exit(-1)
else:
break

View File

@ -17,7 +17,7 @@ def get_size(file_dir):
# we should only expect one file, if no, something is wrong
file_name = glob.glob(os.path.join(file_dir, "*"))[0]
return os.stat(file_name).st_size
except Exception:
except:
logging.exception(f"error getting file from: {file_dir}")
return 0
@ -41,13 +41,11 @@ def build_message(size):
"build_num": os.environ.get("CIRCLE_BUILD_NUM"),
"sha1": os.environ.get("CIRCLE_SHA1"),
"branch": os.environ.get("CIRCLE_BRANCH"),
"workflow_id": os.environ.get("CIRCLE_WORKFLOW_ID"),
},
"int": {
"time": int(time.time()),
"size": size,
"commit_time": int(os.environ.get("COMMIT_TIME", "0")),
"run_duration": int(time.time() - os.path.getmtime(os.path.realpath(__file__))),
},
}
@ -116,12 +114,10 @@ def report_android_sizes(file_dir):
"build_num": os.environ.get("CIRCLE_BUILD_NUM"),
"sha1": os.environ.get("CIRCLE_SHA1"),
"branch": os.environ.get("CIRCLE_BRANCH"),
"workflow_id": os.environ.get("CIRCLE_WORKFLOW_ID"),
},
"int": {
"time": int(time.time()),
"commit_time": int(os.environ.get("COMMIT_TIME", "0")),
"run_duration": int(time.time() - os.path.getmtime(os.path.realpath(__file__))),
"size": comp_size,
"raw_size": uncomp_size,
},
@ -145,5 +141,5 @@ if __name__ == "__main__":
if size != 0:
try:
send_message([build_message(size)])
except Exception:
except:
logging.exception("can't send message")

View File

@ -1,10 +1,7 @@
# https://developercommunity.visualstudio.com/t/install-specific-version-of-vs-component/1142479
# https://docs.microsoft.com/en-us/visualstudio/releases/2019/history#release-dates-and-build-numbers
# 16.8.5 BuildTools
$VS_DOWNLOAD_LINK = "https://download.visualstudio.microsoft.com/download/pr/20130c62-1bc8-43d6-b4f0-c20bb7c79113/145a319d79a83376915d8f855605e152ef5f6fa2b2f1d2dca411fb03722eea72/vs_BuildTools.exe"
$VS_DOWNLOAD_LINK = "https://aka.ms/vs/15/release/vs_buildtools.exe"
$COLLECT_DOWNLOAD_LINK = "https://aka.ms/vscollect.exe"
$VS_INSTALL_ARGS = @("--nocache","--quiet","--wait", "--add Microsoft.VisualStudio.Workload.VCTools",
"--add Microsoft.VisualStudio.Component.VC.Tools.14.11",
"--add Microsoft.Component.MSBuild",
"--add Microsoft.VisualStudio.Component.Roslyn.Compiler",
"--add Microsoft.VisualStudio.Component.TextTemplating",
@ -16,25 +13,10 @@ $VS_INSTALL_ARGS = @("--nocache","--quiet","--wait", "--add Microsoft.VisualStud
curl.exe --retry 3 -kL $VS_DOWNLOAD_LINK --output vs_installer.exe
if ($LASTEXITCODE -ne 0) {
echo "Download of the VS 2019 Version 16.8.5 installer failed"
echo "Download of the VS 2017 installer failed"
exit 1
}
if (Test-Path "${env:ProgramFiles(x86)}\Microsoft Visual Studio\Installer\vswhere.exe") {
$existingPath = & "${env:ProgramFiles(x86)}\Microsoft Visual Studio\Installer\vswhere.exe" -products "Microsoft.VisualStudio.Product.BuildTools" -version "[16, 17)" -property installationPath
if ($existingPath -ne $null) {
echo "Found existing BuildTools installation in $existingPath"
$VS_UNINSTALL_ARGS = @("uninstall", "--installPath", "`"$existingPath`"", "--quiet","--wait")
$process = Start-Process "${PWD}\vs_installer.exe" -ArgumentList $VS_UNINSTALL_ARGS -NoNewWindow -Wait -PassThru
$exitCode = $process.ExitCode
if (($exitCode -ne 0) -and ($exitCode -ne 3010)) {
echo "Original BuildTools uninstall failed with code $exitCode"
exit 1
}
echo "Original BuildTools uninstalled"
}
}
$process = Start-Process "${PWD}\vs_installer.exe" -ArgumentList $VS_INSTALL_ARGS -NoNewWindow -Wait -PassThru
Remove-Item -Path vs_installer.exe -Force
$exitCode = $process.ExitCode

View File

@ -1,5 +0,0 @@
$CMATH_DOWNLOAD_LINK = "https://raw.githubusercontent.com/microsoft/STL/12c684bba78f9b032050526abdebf14f58ca26a3/stl/inc/cmath"
$VC14_28_INSTALL_PATH="C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Tools\MSVC\14.28.29910\include"
curl.exe --retry 3 -kL $CMATH_DOWNLOAD_LINK --output "$home\cmath"
Move-Item -Path "$home\cmath" -Destination "$VC14_28_INSTALL_PATH" -Force

View File

@ -1,63 +1,30 @@
#!/bin/bash
set -eux -o pipefail
cuda_major_version=${CUDA_VERSION%.*}
if [[ "$cuda_major_version" == "10" ]]; then
cuda_installer_name="cuda_10.1.243_426.00_win10"
msbuild_project_dir="CUDAVisualStudioIntegration/extras/visual_studio_integration/MSBuildExtensions"
cuda_install_packages="nvcc_10.1 cuobjdump_10.1 nvprune_10.1 cupti_10.1 cublas_10.1 cublas_dev_10.1 cudart_10.1 cufft_10.1 cufft_dev_10.1 curand_10.1 curand_dev_10.1 cusolver_10.1 cusolver_dev_10.1 cusparse_10.1 cusparse_dev_10.1 nvgraph_10.1 nvgraph_dev_10.1 npp_10.1 npp_dev_10.1 nvrtc_10.1 nvrtc_dev_10.1 nvml_dev_10.1"
elif [[ "$cuda_major_version" == "11" ]]; then
if [[ "${CUDA_VERSION}" == "11.1" ]]; then
cuda_installer_name="cuda_11.1.0_456.43_win10"
msbuild_project_dir="visual_studio_integration/CUDAVisualStudioIntegration/extras/visual_studio_integration/MSBuildExtensions"
cuda_install_packages="nvcc_11.1 cuobjdump_11.1 nvprune_11.1 nvprof_11.1 cupti_11.1 cublas_11.1 cublas_dev_11.1 cudart_11.1 cufft_11.1 cufft_dev_11.1 curand_11.1 curand_dev_11.1 cusolver_11.1 cusolver_dev_11.1 cusparse_11.1 cusparse_dev_11.1 npp_11.1 npp_dev_11.1 nvrtc_11.1 nvrtc_dev_11.1 nvml_dev_11.1"
elif [[ "${CUDA_VERSION}" == "11.3" ]]; then
cuda_installer_name="cuda_11.3.0_465.89_win10"
msbuild_project_dir="visual_studio_integration/CUDAVisualStudioIntegration/extras/visual_studio_integration/MSBuildExtensions"
cuda_install_packages="thrust_11.3 nvcc_11.3 cuobjdump_11.3 nvprune_11.3 nvprof_11.3 cupti_11.3 cublas_11.3 cublas_dev_11.3 cudart_11.3 cufft_11.3 cufft_dev_11.3 curand_11.3 curand_dev_11.3 cusolver_11.3 cusolver_dev_11.3 cusparse_11.3 cusparse_dev_11.3 npp_11.3 npp_dev_11.3 nvrtc_11.3 nvrtc_dev_11.3 nvml_dev_11.3"
else
echo "This should not happen! ABORT."
exit 1
fi
else
echo "CUDA_VERSION $CUDA_VERSION is not supported yet"
exit 1
fi
if [[ "$cuda_major_version" == "11" && "${JOB_EXECUTOR}" == "windows-with-nvidia-gpu" ]]; then
cuda_install_packages="${cuda_install_packages} Display.Driver"
fi
cuda_installer_link="https://ossci-windows.s3.amazonaws.com/${cuda_installer_name}.exe"
curl --retry 3 -kLO $cuda_installer_link
7z x ${cuda_installer_name}.exe -o${cuda_installer_name}
cd ${cuda_installer_name}
curl --retry 3 -kLO https://ossci-windows.s3.amazonaws.com/cuda_10.1.243_426.00_win10.exe
7z x cuda_10.1.243_426.00_win10.exe -ocuda_10.1.243_426.00_win10
cd cuda_10.1.243_426.00_win10
mkdir cuda_install_logs
set +e
./setup.exe -s ${cuda_install_packages} -loglevel:6 -log:"$(pwd -W)/cuda_install_logs"
./setup.exe -s nvcc_10.1 cuobjdump_10.1 nvprune_10.1 cupti_10.1 cublas_10.1 cublas_dev_10.1 cudart_10.1 cufft_10.1 cufft_dev_10.1 curand_10.1 curand_dev_10.1 cusolver_10.1 cusolver_dev_10.1 cusparse_10.1 cusparse_dev_10.1 nvgraph_10.1 nvgraph_dev_10.1 npp_10.1 npp_dev_10.1 nvrtc_10.1 nvrtc_dev_10.1 nvml_dev_10.1 -loglevel:6 -log:"$(pwd -W)/cuda_install_logs"
set -e
if [[ "${VC_YEAR}" == "2017" ]]; then
cp -r ${msbuild_project_dir}/* "C:/Program Files (x86)/Microsoft Visual Studio/2017/${VC_PRODUCT}/Common7/IDE/VC/VCTargets/BuildCustomizations/"
cp -r CUDAVisualStudioIntegration/extras/visual_studio_integration/MSBuildExtensions/* "C:/Program Files (x86)/Microsoft Visual Studio/2017/${VC_PRODUCT}/Common7/IDE/VC/VCTargets/BuildCustomizations/"
else
cp -r ${msbuild_project_dir}/* "C:/Program Files (x86)/Microsoft Visual Studio/2019/${VC_PRODUCT}/MSBuild/Microsoft/VC/v160/BuildCustomizations/"
cp -r CUDAVisualStudioIntegration/extras/visual_studio_integration/MSBuildExtensions/* "C:/Program Files (x86)/Microsoft Visual Studio/2019/${VC_PRODUCT}/MSBuild/Microsoft/VC/v160/BuildCustomizations/"
fi
if ! ls "/c/Program Files/NVIDIA Corporation/NvToolsExt/bin/x64/nvToolsExt64_1.dll"
then
curl --retry 3 -kLO https://ossci-windows.s3.amazonaws.com/NvToolsExt.7z
7z x NvToolsExt.7z -oNvToolsExt
mkdir -p "C:/Program Files/NVIDIA Corporation/NvToolsExt"
cp -r NvToolsExt/* "C:/Program Files/NVIDIA Corporation/NvToolsExt/"
export NVTOOLSEXT_PATH="C:\\Program Files\\NVIDIA Corporation\\NvToolsExt\\"
fi
curl --retry 3 -kLO https://ossci-windows.s3.amazonaws.com/NvToolsExt.7z
7z x NvToolsExt.7z -oNvToolsExt
mkdir -p "C:/Program Files/NVIDIA Corporation/NvToolsExt"
cp -r NvToolsExt/* "C:/Program Files/NVIDIA Corporation/NvToolsExt/"
export NVTOOLSEXT_PATH="C:\\Program Files\\NVIDIA Corporation\\NvToolsExt\\"
if ! ls "/c/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v${CUDA_VERSION}/bin/nvcc.exe"
if ! ls "/c/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.1/bin/nvcc.exe"
then
echo "CUDA installation failed"
mkdir -p /c/w/build-results
@ -66,5 +33,5 @@ then
fi
cd ..
rm -rf ./${cuda_installer_name}
rm -f ./${cuda_installer_name}.exe
rm -rf ./cuda_10.1.243_426.00_win10
rm -f ./cuda_10.1.243_426.00_win10.exe

View File

@ -1,28 +0,0 @@
#!/bin/bash
set -eux -o pipefail
cuda_major_version=${CUDA_VERSION%.*}
if [[ "$cuda_major_version" == "10" ]]; then
cudnn_installer_name="cudnn-${CUDA_VERSION}-windows10-x64-v7.6.4.38"
elif [[ "$cuda_major_version" == "11" ]]; then
if [[ "${CUDA_VERSION}" == "11.1" ]]; then
cudnn_installer_name="cudnn-${CUDA_VERSION}-windows-x64-v8.0.5.39"
elif [[ "${CUDA_VERSION}" == "11.3" ]]; then
cudnn_installer_name="cudnn-${CUDA_VERSION}-windows-x64-v8.2.0.53"
else
echo "This should not happen! ABORT."
exit 1
fi
else
echo "CUDNN for CUDA_VERSION $CUDA_VERSION is not supported yet"
exit 1
fi
cudnn_installer_link="https://ossci-windows.s3.amazonaws.com/${cudnn_installer_name}.zip"
curl --retry 3 -O $cudnn_installer_link
7z x ${cudnn_installer_name}.zip -ocudnn
cp -r cudnn/cuda/* "C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v${CUDA_VERSION}/"
rm -rf cudnn
rm -f ${cudnn_installer_name}.zip

View File

@ -0,0 +1,45 @@
#!/usr/bin/env python3
import cimodel.data.caffe2_build_definitions as caffe2_build_definitions
import cimodel.data.simple.util.docker_constants as pytorch_docker_constants
from yaml import load
try:
from yaml import CLoader as Loader
except ImportError:
from yaml import Loader
def load_config(filename=".circleci/config.yml"):
with open(filename, "r") as fh:
return load("".join(fh.readlines()), Loader)
def load_tags_for_projects(workflow_config):
return {
v["ecr_gc_job"]["project"]: v["ecr_gc_job"]["tags_to_keep"]
for v in workflow_config["workflows"]["ecr_gc"]["jobs"]
if isinstance(v, dict) and "ecr_gc_job" in v
}
def check_version(job, tags, expected_version):
valid_versions = tags[job].split(",")
if expected_version not in valid_versions:
raise RuntimeError(
"We configured {} to use Docker version {}; but this "
"version is not configured in job ecr_gc_job_for_{}. Non-deployed versions will be "
"garbage collected two weeks after they are created. DO NOT LAND "
"THIS TO MASTER without also updating ossci-job-dsl with this version."
"\n\nDeployed versions: {}".format(job, expected_version, job, tags[job])
)
def validate_docker_version():
tags = load_tags_for_projects(load_config())
check_version("pytorch", tags, pytorch_docker_constants.DOCKER_IMAGE_TAG)
check_version("caffe2", tags, caffe2_build_definitions.DOCKER_IMAGE_VERSION)
if __name__ == "__main__":
validate_docker_version()

View File

@ -59,7 +59,7 @@ binary_windows_params: &binary_windows_params
default: ""
executor:
type: string
default: "windows-xlarge-cpu-with-nvidia-cuda"
default: "windows-cpu-with-nvidia-cuda"
environment:
BUILD_ENVIRONMENT: << parameters.build_environment >>
BUILD_FOR_SYSTEM: windows

View File

@ -0,0 +1,27 @@
caffe2_params: &caffe2_params
parameters:
build_environment:
type: string
default: ""
build_ios:
type: string
default: ""
docker_image:
type: string
default: ""
use_cuda_docker_runtime:
type: string
default: ""
build_only:
type: string
default: ""
resource_class:
type: string
default: "large"
environment:
BUILD_ENVIRONMENT: << parameters.build_environment >>
BUILD_IOS: << parameters.build_ios >>
USE_CUDA_DOCKER_RUNTIME: << parameters.use_cuda_docker_runtime >>
DOCKER_IMAGE: << parameters.docker_image >>
BUILD_ONLY: << parameters.build_only >>
resource_class: << parameters.resource_class >>

Some files were not shown because too many files have changed in this diff Show More