Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fallback to native batchnorm implementation if input shape unsupported by cudnn #31976

Closed
wants to merge 323 commits into from
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
323 commits
Select commit Hold shift + click to select a range
2bb9dbe
omit constexpr with nvcc on clang (#32149)
EscapeZero Jan 14, 2020
4a26bb9
Suppress pip logs (#31912)
xta0 Jan 14, 2020
51a3454
Revert D18482934: support torch script call over rpc
suo Jan 14, 2020
9bf0479
Fix the passing-by-ref constructor of OperatorName. (#32170)
Jan 14, 2020
ecc3497
Update Gemfile (#32147)
xta0 Jan 14, 2020
f3b67bf
Fix frontend kwarg defualts error (#32146)
Jan 14, 2020
f6f1e0a
Automatic update of fbcode/onnx to 65020daafa9183c769938b4512ce543fd5…
houseroad Jan 14, 2020
2bd1791
Fix typo in config script to re-enable libtorch build and test in mac…
Jan 14, 2020
02c3493
Fix an invalid peephole transformation if input/output values are wri…
Krovatkin Jan 15, 2020
8dc67a0
Add cummax
anjali411 Jan 15, 2020
7572501
move ProcessGroupGlooTest to gtest (#32133)
rohan-varma Jan 15, 2020
02f09a1
Implement backend-agnostic rpc._wait_all_workers() utility (#32190)
xush6528 Jan 15, 2020
7ad0385
Fix 'template' keyword warning with clang-cl and clang.exe (#32104)
jdonald Jan 15, 2020
879620e
[caffe2] fix how np.clip is used in lengths_reducer_fused_{4,8}_rowwi…
jspark1105 Jan 15, 2020
c70bb0a
Fixes to prim ops (#32179)
Jan 15, 2020
4dce482
dict type unification fix (#32185)
Jan 15, 2020
19bbb4f
Stop building documentation in pytorch_linux_xenial_cuda*_build (#32187)
zou3519 Jan 15, 2020
ef0f96e
[pytorch][PR] update comment in autograd.h for locking (#32222)
albanD Jan 15, 2020
05088da
[pytorch][PR] Fixed error in sample code of documentation (#31682)
vamshichowdary Jan 15, 2020
8c3ee9f
[Python] Deprecate use of scipy.misc.logsumexp and scipy.misc.comb (#…
Jan 15, 2020
62b06b9
Rename TensorTypeId to DispatchKey (#32154)
pbelevich Jan 15, 2020
4314620
[jit] Module clone work with shared ClassType (#31970)
jerryzh168 Jan 15, 2020
0392e83
Fix simple typo: whos -> whose (#31288)
timgates42 Jan 15, 2020
3d01e3d
Notify other threads before running callbacks (#31713)
xush6528 Jan 15, 2020
3363ca2
example_outputs Doc Edit (#31826)
Chetank99 Jan 15, 2020
81048c4
remove simple .data from torch/nn
albanD Jan 15, 2020
74621ca
Add allgather_base as per our discussion re: ProcessGroup interface. …
agolynski Jan 15, 2020
322f34b
Adding DDP Design Note
mrshenli Jan 15, 2020
a5161c7
Update out-of-date comment on Docker image updates. (#32224)
ezyang Jan 15, 2020
01010f5
Add comments to torch::nn::ConvTranspose{1,2,3}d modules explaining h…
Jan 15, 2020
4460a86
Support op registration if name starts with underscore (_) (#32017)
neginraoof Jan 15, 2020
ffc8e25
Sort export w/ negative axes (#31971)
neginraoof Jan 15, 2020
8b4c695
Added cons folding for ONNX mul, div, sqrt ops (#32077)
neginraoof Jan 15, 2020
31b7d08
Add File existence checking (#32208)
xta0 Jan 15, 2020
de5821d
Torchscript print to logcat (#31456)
IvanKobzarev Jan 16, 2020
104b2c6
Tensor prep from image in native (#31426)
IvanKobzarev Jan 16, 2020
90c65b8
Define `repr()` on IValues (#32232)
suo Jan 16, 2020
89c6e18
Updating submodules
Jan 16, 2020
851a7e8
Add CAFFE2_API to video decoding functions (#31187)
h6197627 Jan 16, 2020
f3b62d4
Updating submodules
Jan 16, 2020
d75b6b3
Support shape inference and lowering of SparseLengthsWeightedSumFused…
Jan 16, 2020
7df5dc2
Creating callUnboxedWithDispatchKey method (#32198)
Jan 16, 2020
bab87e4
reimplement __torch_function__ overrides for torch.functional using i…
ngoldbaum Jan 16, 2020
14548c2
out variant for native_batch_norm forward (#29192)
vadimkantorov Jan 16, 2020
f94aab4
Logical condition reduction (#32201)
gaurav1086 Jan 16, 2020
cd99b37
Pin Pillow to latest and use a torchvision that works with it (#32290)
hugovk Jan 16, 2020
b26ee54
For ppc64le, stop presenting the python 2.7 builds (we will no longer…
dncliss Jan 16, 2020
8746f90
Fix weight backward for cudnn conv of large tensor (#31889)
zasdfgbnm Jan 16, 2020
7b73907
Make an assert on a hotpath trigger only in DEBUG mode. (#32117)
ezyang Jan 16, 2020
36d0919
Move error reporting code out-of-line from header. (#32118)
ezyang Jan 16, 2020
b85dbe8
Out-of-line construction of OperatorName. (#32121)
ezyang Jan 16, 2020
34c751c
Eliminate exception throwing code from dispatch call sites (#32168)
ezyang Jan 16, 2020
8c8bd79
Add CI scripts for Custom Build (#32316)
xta0 Jan 16, 2020
53708e2
classic fixed-point liveness
Krovatkin Jan 16, 2020
c276149
Enhancing the test (#32321)
Jan 16, 2020
b79030d
remove unused code after refactoring optimizations into profiling-sen…
Krovatkin Jan 17, 2020
ef5ae48
Register RoIAlignRotated with C10
SuperIRabbit Jan 17, 2020
91bdb87
fix spelling mistake: excpected -> expected
your-average-bear Jan 17, 2020
7a9c920
add lock for ncclCommAbort (#31901)
zhaojuanmao Jan 17, 2020
904ab09
fix testSend and testRecv in ProcessGroupGlooTest (#32134)
rohan-varma Jan 17, 2020
61a2b34
Updating submodules
Jan 17, 2020
4968bc2
cap the maximum depth of bailout chains at 1 (#32073)
Krovatkin Jan 17, 2020
6a5a55d
use gtest asserts in ProcessGroupGlooTest instead of other checks (#3…
rohan-varma Jan 17, 2020
78d8f69
Don't dispatch to integral types in smooth_l1_kernel
zasdfgbnm Jan 17, 2020
5b815d9
Added cummin
anjali411 Jan 17, 2020
8c1268a
Use default scale/zero_point in fake_quantize module instead of None …
jerryzh168 Jan 17, 2020
7732924
Delete unused bernoulli_Tensor from THTensorRandom.h
pbelevich Jan 17, 2020
aa61d1e
Add a new job to support custom build (#32323)
xta0 Jan 17, 2020
4e69352
Add 64bit atomic fetch add (#32354)
Jan 17, 2020
e133d8b
Fix ASAN / potential segfault in quantized Tensor memory allocations.
AshkanAliabadi Jan 17, 2020
df514fd
C++ C2/Glow operator unittest
Jan 17, 2020
e7bc166
fix unchecked cast alias analysis (#32309)
Jan 17, 2020
9b6ec61
exposing CPU/GPU Copy ops (#32248)
wat3rBro Jan 17, 2020
5a58c16
Updating submodules
Jan 17, 2020
bdd5e15
skip testExceptions in ProcessGroupGloo if built with TSAN (#32242)
rohan-varma Jan 17, 2020
7e3c438
Renaming IValue List functions (#32093)
zdevito Jan 17, 2020
c8ca70e
Updating submodules
Jan 17, 2020
f326045
Fix typos, via a Levenshtein-type corrector (#31523)
bwignall Jan 18, 2020
5bc44fb
TensorIterator unrolling and vectorized load - step 0, 1 (#31974)
zasdfgbnm Jan 18, 2020
0ac31a9
run code analysis against mobile interpreter (#32276)
ljk53 Jan 18, 2020
824e649
Specify requires_grad for Parameter replica so it's not always set to…
jithunnair-amd Jan 18, 2020
10c2bd3
Fix cudnn channels_last descriptors problem (#31952)
jjsjann123 Jan 18, 2020
25e62eb
Updating submodules
Jan 18, 2020
8c40a78
Back out "Calling JITed 8 Bit Fused SLS in FBGEMM from C2" (#32381)
Jan 18, 2020
5342968
Remove stray `@script` (#32235)
Jan 18, 2020
61ee8c9
porting scatter_add to ATen (CPU) (#31662)
nikitaved Jan 18, 2020
ceffdbd
Temporary workaround for BC test due to schema parser changes
Jan 18, 2020
c7fdf5b
Remove __torch__ from custom class qualname
Jan 18, 2020
c7078a1
Fix returning instance of custom class from method
Jan 18, 2020
1ecad2b
Test passing custom class instance to bound method
Jan 18, 2020
58234c0
support torch script call over rpc (#32197)
zhaojuanmao Jan 18, 2020
7fbfb7e
Updating submodules
Jan 19, 2020
b543e3c
support empty batch in group normalization (#32401)
ppwwyyxx Jan 20, 2020
ecbf6f9
Removed unused weight update in prepack. Moved zero point update to (…
kimishpatel Jan 20, 2020
418ebc8
Build: Respect USE_CUDNN=0, even if cudnn is found (#32404)
peterbell10 Jan 21, 2020
a2641e6
Make type of `Tensor.type()` more specific (#32353)
Jan 21, 2020
0c03304
.circleci: Only run macos libtorch on master (#32378)
seemethere Jan 21, 2020
cc2d5b1
F.normalize uses clamp_min_ inplace (#32360)
ssnl Jan 21, 2020
1177191
Synchronize with ShipIt.
ezyang Jan 21, 2020
9ce25cc
add an option to record time spent waiting for GIL (#30842)
rohan-varma Jan 21, 2020
c13df8b
Fix cusparse version check (#32405)
peterbell10 Jan 21, 2020
9482683
Remove dead includes in caffe2/test
Jan 21, 2020
e4f43bf
Set rpath for JNI library on Mac (#32247)
dreiss Jan 21, 2020
839fe71
Fix BC test after TorchBind cahnges (#32429)
Jan 21, 2020
ec4be4e
Redundant condition (#32396)
gaurav1086 Jan 21, 2020
f86d6c6
Enhance NCCL watchdog to acitvely abort communicators for timed out o…
pritamdamania Jan 21, 2020
9e853e7
Revert "Temporary workaround for BC test due to schema parser changes…
Jan 21, 2020
bc60052
Updating submodules
Jan 21, 2020
e1d9702
QNNPACK: Add support for dynamic quantization.
Jan 21, 2020
0ed04bf
Updating submodules
Jan 21, 2020
be6ffac
Adagrad optimizer - updated step function, added param_groups, state …
anjali411 Jan 21, 2020
0b606a4
Enhace DispatchStub to be thread safe from a TSAN point of view. (#32…
pritamdamania Jan 21, 2020
7fdc6cb
Fix test_data_parallel name errors and add to run_test.py (#32428)
peterbell10 Jan 21, 2020
4973695
Updating submodules
Jan 22, 2020
64de93d
Move log_normal to Aten(CPU) (#31854)
XiaobingSuper Jan 22, 2020
60b6c99
Updating submodules
Jan 22, 2020
44b270d
`insert_quant_dequant` pass support shared class types (#31408)
jerryzh168 Jan 22, 2020
0d610b4
Remove the support of build options like NO_*, WITH_* (#32447)
xuhdev Jan 22, 2020
248f6d0
Implement backend fallback fallthrough (#32439)
ezyang Jan 22, 2020
8abaa32
fix torch.eq() doc entry (#32399)
stas00 Jan 22, 2020
e37a24b
Always return a new tensor from nn.functional.pad (#32350)
peterbell10 Jan 22, 2020
c342c35
Put sparse all reduce results to input tensors (#32226)
zhaojuanmao Jan 22, 2020
b77c25d
Fix dll load logic for Python 3.8 on Windows (#32215)
peterjc123 Jan 22, 2020
1c017f0
Migrate max and min (binary) from TH to ATen. (#30851)
xuhdev Jan 22, 2020
510a122
add missing align_corners annotation (#32492)
edgarriba Jan 22, 2020
8e68937
Move some of the helper functions for public use (#32202)
jerryzh168 Jan 22, 2020
e184a88
Fix comparisions for ConcreteModuleType (#32256)
zdevito Jan 22, 2020
adf0916
Add str[] float[] constants resubmit
Jan 22, 2020
b01d824
improve mayContainAlias (#31839)
Jan 22, 2020
69492ad
remove tuple logic in constant propagation (#31840)
Jan 22, 2020
38d122e
implement tuple constants (#31841)
Jan 22, 2020
b7c6277
Adding QConfigTypePtrMap (#32203)
jerryzh168 Jan 22, 2020
43eb931
Remove mis-exposed abort API on ProcessGroup
mrshenli Jan 22, 2020
faffd21
Corrected logical boolean expression (#32249)
gaurav1086 Jan 22, 2020
14e0bec
[caffe2] remove unnecessary np.set_printoptions and fix test errors (…
jspark1105 Jan 22, 2020
556c0b0
Updating submodules
Jan 22, 2020
193ac31
[jit] Enable IValue to hold a PyObject (#32491)
zhaojuanmao Jan 22, 2020
4bdfc71
Fix race condition for to() backward that spans devices (#31930)
definitelynotmcarilli Jan 23, 2020
685f090
[Rowwise Pruning][c2 op] Add Quantile Op (#32448)
Jan 23, 2020
e735395
[caffe2] use 2-stage EmbeddingSpMDM interface (#32271)
jspark1105 Jan 23, 2020
f050b16
Move pytorch distributed tests to separate folder for contbuild. (#30…
pritamdamania Jan 23, 2020
21d475e
[gloo] Skip registry warning (#31126)
jiayisuse Jan 23, 2020
02aa3ba
Raise error for code that risk deadlock (#32295)
albanD Jan 23, 2020
ea7bebb
[PyTorch BC] Clean up the whitelist for PyTorch Op BC check (#32523)
houseroad Jan 23, 2020
583bb97
[quant][graphmode] Default to non-inplace in graph mode quantization …
jerryzh168 Jan 23, 2020
9af5a97
Fix nll_loss to support empty tensors on GPU (#31491)
pbelevich Jan 23, 2020
b6b8620
Add unit test on export_opnames with interface. (#31531)
Jan 23, 2020
db02a4e
Support 3D attention mask in MultiheadAttention. (#31996)
Jan 23, 2020
ef94496
[JIT] throw if no self arg on ignored methods (#32503)
Jan 23, 2020
d234626
[quant][graphmode] Support quantizing shared ClassType with different…
jerryzh168 Jan 23, 2020
49cd83d
no more build_pytorch_libs.sh/.bat (#32319)
den-run-ai Jan 23, 2020
ad4fba0
Only run test_conv_large and test_conv_transposed_large_cuda on 32GB …
zasdfgbnm Jan 23, 2020
dbd29e5
[JIT] Passing custom class as arg (#32260)
Jan 23, 2020
7e14c42
[JIT] Test __getstate__ and __setstate__ for custom bound C++ classes
Jan 23, 2020
ae42e23
[JIT] Fix custom class method binding for const methods
Jan 23, 2020
69f9bf8
[JIT] Support returning tuple from custom bound C++ method
Jan 23, 2020
8ed1dd5
[JIT] Add torch.classes.load_library
Jan 23, 2020
6745bfc
Revert "Remove __torch__ from custom class qualname" (#32514)
Jan 23, 2020
4cd6b5c
[quant] Re-enable test_nested that has different qconfig for shared C…
jerryzh168 Jan 23, 2020
d2f6608
porting gather to ATen using TensorIterator with multithreading suppo…
nikitaved Jan 24, 2020
6f146e1
[JIT] Remove capsule type handling of node hashing (#32540)
Jan 24, 2020
ef2d4e6
Updating submodules
Jan 24, 2020
b474c35
[rpc] Remove template on RRef and add Type to RRef creation (#30630)
zhaojuanmao Jan 24, 2020
3ada2e0
[pytorch][embeddingbag] Parallelize the EmbeddingBag operator (#4049)
jianyuh Jan 24, 2020
8fd3eae
[jit] Fix dict type serialization (#32569)
suo Jan 24, 2020
f0c8557
docker: Refactor Dockerfile process for official images (#32515)
seemethere Jan 24, 2020
7d0f0b6
API for testing bailouts (#32518)
Krovatkin Jan 24, 2020
ef5637f
[jit] allow compilation using optional modules (#32539)
suo Jan 24, 2020
fd1a4f1
[pytorch] update code analyzer build.sh to handle srcs with same name…
ljk53 Jan 24, 2020
fe3eb09
[quant] Re-enable fold_convbn in quantize_script (#32302)
jerryzh168 Jan 24, 2020
d2bda53
[quant][graphmode] Call _jit_pass_dedup_module_ueses in quantize_scri…
jerryzh168 Jan 24, 2020
9e59244
fix view listing in autograd codegen (#32044)
albanD Jan 24, 2020
3ab3075
Make autogen functions correct for multiple outputs and views (#31990)
albanD Jan 24, 2020
573a302
[pytorch] Minor: boilerplate to propagate errors in request_callback_…
jjlilley Jan 24, 2020
2bfd33b
[refactor] Adding FoldConvBatchNorm2dHelper (#32374)
jerryzh168 Jan 24, 2020
9e0ce72
[pytorch] change op dependency output to use double-quoted strings (#…
ljk53 Jan 24, 2020
52f8f03
add diag into pt operator microbenchmark (#32597)
hl475 Jan 24, 2020
91f10a1
[quant][graphmode][refactor] Better API for fold_convbn (#32380)
jerryzh168 Jan 24, 2020
6412ca3
duplicate symbols with AT_PARALLEL_OPENMP=0 (#32568)
EscapeZero Jan 25, 2020
666472a
[docs] Change fut.wait() to torch.jit._wait(fut) in jit overview docs…
rohan-varma Jan 25, 2020
e7edc5f
[jit] Cloning constants in ClassType (#32371)
jerryzh168 Jan 25, 2020
1218a16
[pytorch][refactor] Explicitly use auto* for pointers (#32548)
jianyuh Jan 25, 2020
59dbece
Fix iterator for ncclCommWatchdog. (#32571)
pritamdamania Jan 25, 2020
1695418
Add operator support for dynamic quant on mobile (#32479)
supriyar Jan 25, 2020
e0ffe72
[aten] fix shadowing variable warning (#32573)
jspark1105 Jan 25, 2020
6ad9e5c
Support TorchScript call over remote API (RRef) (#32466)
xush6528 Jan 25, 2020
bd20274
[caffe2] use JIT'ed fp32 SLS (#32413)
jspark1105 Jan 25, 2020
0afe195
[pytorch] move type_derived_methods out of anonymous namespace (#32275)
ljk53 Jan 25, 2020
6928338
[pytorch] codegen flags to whitelist op registrations / generate to s…
ljk53 Jan 25, 2020
320d1a1
Fix wrong typing (torch/nn/parameter.pyi) (#32617)
jeongukjae Jan 26, 2020
5fd037c
Fix MagmaInitializesCorrectly_CUDA by using an invertible matrix (#32…
Jan 26, 2020
3bbb36e
Update linspace types (#32218)
Enealor Jan 26, 2020
90a259e
Add warning regarding pickle insecurity on torch.load documentation (…
andfoy Jan 26, 2020
12d5933
Bug fix of norm minimization for dev mode (#31462)
csummersea Jan 26, 2020
8fbe1cc
faster bailout tests (#32266)
Krovatkin Jan 27, 2020
1e5aead
Make cuda search process of cpp extension quiet (#32620)
peterjc123 Jan 27, 2020
5b321a0
[rpc] make handling of FORWARD_AUTOGRAD_REQ in request_callback_impl …
rohan-varma Jan 27, 2020
957a07f
[ROCm] Enable Caffe2 video operators for ROCm
lcskrishna Jan 27, 2020
19bb496
Enable mkldnn on windows (#31355)
XiaobingSuper Jan 27, 2020
602394e
verify input sizes for instance norm and group norm (#29082)
v0dro Jan 27, 2020
ca9dc67
0-dim batch size input for interpolate. (#32400)
v0dro Jan 27, 2020
5ac2593
[ROCm] Adjust elementwise_kernel settings on ROCm (#32609)
iotamudelta Jan 27, 2020
e36cbb8
Fixes moving after weight norm application (#32563)
Jan 27, 2020
64323ae
Back out "Use simd version for fp16 conversions" (#32640)
Jan 27, 2020
f0917dc
Revert D19562258: [pytorch][PR] Fixes moving after weight norm applic…
ezyang Jan 27, 2020
f6c46df
Adding native qconcat
z-a-f Jan 27, 2020
897b690
Kill THIntegerTensor, THDenseTensor, THDenseIndexTensor. (#32599)
gchanan Jan 27, 2020
57519bd
Revert "Fix iterator for ncclCommWatchdog. (#32571)" (#32649)
ezyang Jan 27, 2020
389b9c1
Updating submodules
Jan 27, 2020
812b1ad
[quantization] FP16 dynamic quantized Linear
Jan 27, 2020
39987de
[vulkan][caffe2] Add logging for descriptor extensions, fp16 storage
IvanKobzarev Jan 28, 2020
1695915
Make _wait_all_workers() support being called for multiple times (#32…
xush6528 Jan 28, 2020
1217c9b
Updating submodules
Jan 28, 2020
5c8535d
Make C++ RpcAgent::currentRPCAgent_ the source of truth of current RP…
xush6528 Jan 28, 2020
db8ce7e
Back out "Make autogen functions correct for multiple outputs and vie…
albanD Jan 28, 2020
666e543
Clean up mvlgamma doc (including a weird way to link to reference) (#…
xuhdev Jan 28, 2020
b9f764b
Use the C++ current RpcAgent pointer to eliminate the unnecessary arg…
xush6528 Jan 28, 2020
d68592a
[JIT] Fix classes as attributes in recursive scripting
Jan 28, 2020
b3848c5
Fix flaky test_nccl_timeout. (#32653)
pritamdamania Jan 28, 2020
8e41615
div_kernel: throw when dividing by integer zero (#32629)
Baranowski Jan 28, 2020
6317043
[jit] fix segfault on missing getstate (#32642)
suo Jan 28, 2020
9a2691f
Fix spelling errors
henryhjung Jan 28, 2020
e24ce0e
Kill some more unused code in function_wrapper.py
gchanan Jan 28, 2020
c64dec1
Python binding to export bytecode format for lite interpreter (#32621)
Jan 28, 2020
0dc38be
consider FAIL_GUARD while counting indices for GUARDs (#32672)
Krovatkin Jan 28, 2020
da39091
.circleci: Add workflows for Python 3.8 (#31948)
seemethere Jan 28, 2020
1719da1
[JIT] Support for registering C++ lambdas as methods on custom C++ class
Jan 28, 2020
06c1926
[JIT] Serialize attributes and types in ClassType serialization
Jan 28, 2020
34ccfba
[JIT] Include custom_class.h in torch/script.h
Jan 28, 2020
465ebd5
[JIT] pickle serialization for custom bound classes
Jan 28, 2020
0ea65d6
[JIT] Fix stateful lambda stuff and simplify code in custom C++ bindi…
Jan 28, 2020
6e7e595
[rpc][easy] remove redundant test in rpc_test.py (#32588)
rohan-varma Jan 28, 2020
9de3208
[rpc][flaky-tests] fix for test_handle_send_exceptions and (#32656)
rohan-varma Jan 28, 2020
ffdcbad
Minor refactoring to improve code reuse (#32675)
Jan 28, 2020
0327e75
Back out "[caffe2] use JIT'ed fp32 SLS" (#32711)
jspark1105 Jan 28, 2020
2060e0a
Split serialization tests to their own file (#32241)
Jan 28, 2020
ee60cd9
Back out "fix view listing in autograd codegen" (#32720)
albanD Jan 29, 2020
e74e1cc
Use direct vector indexing in Object::getSlot() instead of at(). (#31…
Jan 29, 2020
9922808
Added missing period in README.
Jan 29, 2020
43d31ae
Added ONNX model checker to ONNX export (#32298)
Jan 29, 2020
6f7d5bb
Temporarily disable the test_quantized_rnn test (#32742)
jianyuh Jan 29, 2020
1f78bd0
[caffe2] Early error throwing for currupted embeddings
Jan 29, 2020
18aab32
Move exponential_ from TH to Aten (CPU) (#32501)
hl475 Jan 29, 2020
02f055f
Add mapping for FbFCPacked in fakefp16 transform
Jan 29, 2020
25d33a2
[JIT] Use Type Level Granularity in Alias Analysis Wildcards (#32251)
Jan 29, 2020
c729614
[JIT] Improve May Contain Alias Using Contained Elements (#32326)
Jan 29, 2020
29a2414
add cudnn bn size limits for #29744
ptrblck Nov 28, 2019
83f9ff1
extend smoke test for large batch size in cudnn bn
ptrblck Jan 9, 2020
c1ceb81
fix lint
Jan 9, 2020
3692fd2
rebase
ptrblck Jan 29, 2020
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Jump to
The table of contents is too big for display.
Diff view
Diff view
  •  
  •  
  •  
2 changes: 1 addition & 1 deletion .circleci/cimodel/data/caffe2_build_definitions.py
Original file line number Diff line number Diff line change
Expand Up @@ -12,7 +12,7 @@

DOCKER_IMAGE_PATH_BASE = "308535385114.dkr.ecr.us-east-1.amazonaws.com/caffe2/"

DOCKER_IMAGE_VERSION = 345
DOCKER_IMAGE_VERSION = "345"


@dataclass
Expand Down
1 change: 1 addition & 0 deletions .circleci/cimodel/data/dimensions.py
Original file line number Diff line number Diff line change
Expand Up @@ -12,4 +12,5 @@
"3.5",
"3.6",
"3.7",
"3.8"
]
8 changes: 5 additions & 3 deletions .circleci/cimodel/data/pytorch_build_data.py
Original file line number Diff line number Diff line change
Expand Up @@ -39,14 +39,16 @@
# and
# https://github.com/pytorch/pytorch/blob/master/.jenkins/pytorch/build.sh#L153
# (from https://github.com/pytorch/pytorch/pull/17323#discussion_r259453144)
X("3.6"),
]),
("9.2", [X("3.6")]),
("10", [X("3.6")]),
("10.1", [
XImportant("3.6"),
("3.6", [
("libtorch", [XImportant(True)])
]),
]),
("9.2", [X("3.6")]),
("10", [X("3.6")]),
("10.1", [X("3.6")]),
]),
("android", [
("r19c", [
Expand Down
18 changes: 16 additions & 2 deletions .circleci/cimodel/data/pytorch_build_definitions.py
Original file line number Diff line number Diff line change
Expand Up @@ -13,7 +13,7 @@

# ARE YOU EDITING THIS NUMBER? MAKE SURE YOU READ THE GUIDANCE AT THE
# TOP OF .circleci/config.yml
DOCKER_IMAGE_VERSION = 405
DOCKER_IMAGE_VERSION = "07597f23-fa81-474c-8bef-5c8a91b50595"


@dataclass
Expand Down Expand Up @@ -160,6 +160,11 @@ def gen_dependent_configs(xenial_parent_config):

configs.append(c)

return configs

def gen_docs_configs(xenial_parent_config):
configs = []

for x in ["pytorch_python_doc_push", "pytorch_cpp_doc_push"]:
configs.append(HiddenConf(x, parent_build=xenial_parent_config))

Expand Down Expand Up @@ -247,7 +252,16 @@ def instantiate_configs():
parallel_backend=parallel_backend,
)

if cuda_version == "9" and python_version == "3.6" and not is_libtorch:
# 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.
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 == "10.1" and python_version == "3.6" and not is_libtorch:
c.dependent_tests = gen_dependent_configs(c)

if (compiler_name == "gcc"
Expand Down
531 changes: 390 additions & 141 deletions .circleci/config.yml

Large diffs are not rendered by default.

28 changes: 11 additions & 17 deletions .circleci/docker/build.sh
Original file line number Diff line number Diff line change
Expand Up @@ -27,6 +27,8 @@ elif [[ "$image" == *-bionic* ]]; then
UBUNTU_VERSION=18.04
fi

TRAVIS_DL_URL_PREFIX="https://s3.amazonaws.com/travis-python-archives/binaries/ubuntu/14.04/x86_64"

# It's annoying to rename jobs every time you want to rewrite a
# configuration, so we hardcode everything here rather than do it
# from scratch
Expand Down Expand Up @@ -54,6 +56,13 @@ case "$image" in
GCC_VERSION=7
# Do not install PROTOBUF, DB, and VISION as a test
;;
pytorch-linux-xenial-py3.8)
# TODO: This is a hack, get rid of this as soon as you get rid of the travis downloads
TRAVIS_DL_URL_PREFIX="https://s3.amazonaws.com/travis-python-archives/binaries/ubuntu/16.04/x86_64"
TRAVIS_PYTHON_VERSION=3.8
GCC_VERSION=7
# Do not install PROTOBUF, DB, and VISION as a test
;;
pytorch-linux-xenial-py3.6-gcc4.8)
ANACONDA_PYTHON_VERSION=3.6
GCC_VERSION=4.8
Expand Down Expand Up @@ -88,22 +97,6 @@ case "$image" in
DB=yes
VISION=yes
;;
pytorch-linux-xenial-cuda8-cudnn7-py2)
CUDA_VERSION=8.0
CUDNN_VERSION=7
ANACONDA_PYTHON_VERSION=2.7
PROTOBUF=yes
DB=yes
VISION=yes
;;
pytorch-linux-xenial-cuda8-cudnn7-py3)
CUDA_VERSION=8.0
CUDNN_VERSION=7
ANACONDA_PYTHON_VERSION=3.6
PROTOBUF=yes
DB=yes
VISION=yes
;;
pytorch-linux-xenial-cuda9-cudnn7-py2)
CUDA_VERSION=9.0
CUDNN_VERSION=7
Expand All @@ -119,7 +112,6 @@ case "$image" in
PROTOBUF=yes
DB=yes
VISION=yes
KATEX=yes
;;
pytorch-linux-xenial-cuda9.2-cudnn7-py3-gcc7)
CUDA_VERSION=9.2
Expand Down Expand Up @@ -147,6 +139,7 @@ case "$image" in
PROTOBUF=yes
DB=yes
VISION=yes
KATEX=yes
;;
pytorch-linux-xenial-py3-clang5-asan)
ANACONDA_PYTHON_VERSION=3.6
Expand Down Expand Up @@ -186,6 +179,7 @@ tmp_tag="tmp-$(cat /dev/urandom | tr -dc 'a-z' | fold -w 32 | head -n 1)"
# Build image
docker build \
--no-cache \
--build-arg "TRAVIS_DL_URL_PREFIX=${TRAVIS_DL_URL_PREFIX}" \
--build-arg "BUILD_ENVIRONMENT=${image}" \
--build-arg "PROTOBUF=${PROTOBUF:-}" \
--build-arg "THRIFT=${THRIFT:-}" \
Expand Down
6 changes: 2 additions & 4 deletions .circleci/docker/common/install_conda.sh
Original file line number Diff line number Diff line change
Expand Up @@ -65,9 +65,7 @@ if [ -n "$ANACONDA_PYTHON_VERSION" ]; then
# DO NOT install cmake here as it would install a version newer than 3.5, but
# we want to pin to version 3.5.
conda_install numpy pyyaml mkl mkl-include setuptools cffi typing future six
if [[ "$CUDA_VERSION" == 8.0* ]]; then
conda_install magma-cuda80 -c pytorch
elif [[ "$CUDA_VERSION" == 9.0* ]]; then
if [[ "$CUDA_VERSION" == 9.0* ]]; then
conda_install magma-cuda90 -c pytorch
elif [[ "$CUDA_VERSION" == 9.1* ]]; then
conda_install magma-cuda91 -c pytorch
Expand All @@ -88,7 +86,7 @@ if [ -n "$ANACONDA_PYTHON_VERSION" ]; then
# 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==0.28.0
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
2 changes: 1 addition & 1 deletion .circleci/docker/common/install_travis_python.sh
Original file line number Diff line number Diff line change
Expand Up @@ -14,7 +14,7 @@ if [ -n "$TRAVIS_PYTHON_VERSION" ]; then

# Download Python binary from Travis
pushd tmp
as_jenkins wget --quiet https://s3.amazonaws.com/travis-python-archives/binaries/ubuntu/14.04/x86_64/python-$TRAVIS_PYTHON_VERSION.tar.bz2
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
Expand Down
1 change: 1 addition & 0 deletions .circleci/docker/ubuntu/Dockerfile
Original file line number Diff line number Diff line change
Expand Up @@ -46,6 +46,7 @@ RUN bash ./install_gcc.sh && rm install_gcc.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
Expand Down
6 changes: 4 additions & 2 deletions .circleci/scripts/binary_ios_upload.sh
Original file line number Diff line number Diff line change
Expand Up @@ -17,8 +17,10 @@ cd ${ZIP_DIR}/install/lib
target_libs=(libc10.a libclog.a libcpuinfo.a libeigen_blas.a libpytorch_qnnpack.a libtorch_cpu.a libtorch.a)
for lib in ${target_libs[*]}
do
libs=(${ARTIFACTS_DIR}/x86_64/lib/${lib} ${ARTIFACTS_DIR}/arm64/lib/${lib})
lipo -create "${libs[@]}" -o ${ZIP_DIR}/install/lib/${lib}
if [ -f "${ARTIFACTS_DIR}/x86_64/lib/${lib}" ] && [ -f "${ARTIFACTS_DIR}/arm64/lib/${lib}" ]; then
libs=("${ARTIFACTS_DIR}/x86_64/lib/${lib}" "${ARTIFACTS_DIR}/arm64/lib/${lib}")
lipo -create "${libs[@]}" -o ${ZIP_DIR}/install/lib/${lib}
fi
done
# for nnpack, we only support arm64 build
cp ${ARTIFACTS_DIR}/arm64/lib/libnnpack.a ./
Expand Down
6 changes: 4 additions & 2 deletions .circleci/scripts/should_run_job.py
Original file line number Diff line number Diff line change
Expand Up @@ -13,13 +13,13 @@
# Selected oldest Python 2 version to ensure Python 2 coverage
'pytorch-linux-xenial-py2.7.9',
# PyTorch CUDA
'pytorch-linux-xenial-cuda9-cudnn7-py3',
'pytorch-linux-xenial-cuda10.1-cudnn7-py3',
# PyTorch ASAN
'pytorch-linux-xenial-py3-clang5-asan',
# PyTorch DEBUG
'pytorch-linux-xenial-py3.6-gcc5.4',
# LibTorch
'pytorch-libtorch-linux-xenial-cuda9-cudnn7-py3',
'pytorch-libtorch-linux-xenial-cuda10.1-cudnn7-py3',

# Caffe2 CPU
'caffe2-py2-mkl-ubuntu16.04',
Expand Down Expand Up @@ -58,8 +58,10 @@
# Pytorch iOS builds
'pytorch-ios-11.2.1-x86_64_build',
'pytorch-ios-11.2.1-arm64_build',
'pytorch-ios-11.2.1-arm64_custom_build',
# PyTorch Mobile builds
'pytorch-linux-xenial-py3-clang5-mobile-build',
'pytorch-linux-xenial-py3-clang5-android-ndk-r19c-mobile-code-analysis',

# Pytorch backward compatibility check
'pytorch-linux-backward-compatibility-check-test',
Expand Down
2 changes: 1 addition & 1 deletion .circleci/validate-docker-version.py
Original file line number Diff line number Diff line change
Expand Up @@ -23,7 +23,7 @@ def load_tags_for_projects(workflow_config):


def check_version(job, tags, expected_version):
valid_versions = [int(v) for v in tags[job].split(",")]
valid_versions = [v for v in tags[job].split(",")]
if expected_version not in valid_versions:
raise RuntimeError(
"We configured {} to use Docker version {}; but this "
Expand Down
12 changes: 3 additions & 9 deletions .circleci/verbatim-sources/header-section.yml
Original file line number Diff line number Diff line change
@@ -1,15 +1,9 @@
# WARNING: DO NOT EDIT THIS FILE DIRECTLY!!!
# See the README.md in this directory.

# IMPORTANT: To update Docker image version, please first update
# https://github.com/pytorch/ossci-job-dsl/blob/master/src/main/groovy/ossci/pytorch/DockerVersion.groovy and
# https://github.com/pytorch/ossci-job-dsl/blob/master/src/main/groovy/ossci/caffe2/DockerVersion.groovy,
# and then update DOCKER_IMAGE_VERSION at the top of the following files:
# * cimodel/data/pytorch_build_definitions.py
# * cimodel/data/caffe2_build_definitions.py
# And the inline copies of the variable in
# * verbatim-sources/job-specs-custom.yml
# (grep for DOCKER_IMAGE)
# IMPORTANT: To update Docker image version, please follow
# the instructions at
# https://github.com/pytorch/pytorch/wiki/Docker-image-build-on-CircleCI

version: 2.1

Expand Down
23 changes: 13 additions & 10 deletions .circleci/verbatim-sources/job-specs-custom.yml
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@
environment:
BUILD_ENVIRONMENT: pytorch-python-doc-push
# TODO: stop hardcoding this
DOCKER_IMAGE: "308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-xenial-cuda9-cudnn7-py3:405"
DOCKER_IMAGE: "308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-xenial-py3.6-gcc5.4:07597f23-fa81-474c-8bef-5c8a91b50595"
resource_class: large
machine:
image: ubuntu-1604:201903-01
Expand Down Expand Up @@ -47,7 +47,7 @@
pytorch_cpp_doc_push:
environment:
BUILD_ENVIRONMENT: pytorch-cpp-doc-push
DOCKER_IMAGE: "308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-xenial-cuda9-cudnn7-py3:405"
DOCKER_IMAGE: "308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-xenial-py3.6-gcc5.4:07597f23-fa81-474c-8bef-5c8a91b50595"
resource_class: large
machine:
image: ubuntu-1604:201903-01
Expand Down Expand Up @@ -205,7 +205,7 @@
pytorch_android_gradle_build:
environment:
BUILD_ENVIRONMENT: pytorch-linux-xenial-py3-clang5-android-ndk-r19c-gradle-build
DOCKER_IMAGE: "308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-xenial-py3-clang5-android-ndk-r19c:405"
DOCKER_IMAGE: "308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-xenial-py3-clang5-android-ndk-r19c:07597f23-fa81-474c-8bef-5c8a91b50595"
PYTHON_VERSION: "3.6"
resource_class: large
machine:
Expand Down Expand Up @@ -291,7 +291,7 @@
pytorch_android_publish_snapshot:
environment:
BUILD_ENVIRONMENT: pytorch-linux-xenial-py3-clang5-android-ndk-r19c-gradle-publish-snapshot
DOCKER_IMAGE: "308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-xenial-py3-clang5-android-ndk-r19c:405"
DOCKER_IMAGE: "308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-xenial-py3-clang5-android-ndk-r19c:07597f23-fa81-474c-8bef-5c8a91b50595"
PYTHON_VERSION: "3.6"
resource_class: large
machine:
Expand Down Expand Up @@ -327,7 +327,7 @@
pytorch_android_gradle_build-x86_32:
environment:
BUILD_ENVIRONMENT: pytorch-linux-xenial-py3-clang5-android-ndk-r19c-gradle-build-only-x86_32
DOCKER_IMAGE: "308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-xenial-py3-clang5-android-ndk-r19c:405"
DOCKER_IMAGE: "308535385114.dkr.ecr.us-east-1.amazonaws.com/pytorch/pytorch-linux-xenial-py3-clang5-android-ndk-r19c:07597f23-fa81-474c-8bef-5c8a91b50595"
PYTHON_VERSION: "3.6"
resource_class: large
machine:
Expand Down Expand Up @@ -437,6 +437,12 @@
chmod a+x ${PROJ_ROOT}/scripts/build_ios.sh
echo "IOS_ARCH: ${IOS_ARCH}"
echo "IOS_PLATFORM: ${IOS_PLATFORM}"

#check the custom build flag
echo "SELECTED_OP_LIST: ${SELECTED_OP_LIST}"
if [ -n "${SELECTED_OP_LIST}" ]; then
export SELECTED_OP_LIST="${PROJ_ROOT}/ios/TestApp/custom_build/${SELECTED_OP_LIST}"
fi
export BUILD_PYTORCH_MOBILE=1
export IOS_ARCH=${IOS_ARCH}
export IOS_PLATFORM=${IOS_PLATFORM}
Expand Down Expand Up @@ -471,12 +477,9 @@
WORKSPACE=/Users/distiller/workspace
PROJ_ROOT=/Users/distiller/project
source ~/anaconda/bin/activate
# Temporarily pin pillow to 6.2.1 as PILLOW_VERSION is replaced by
# _version_ in 7.0.0. Long term fix should be making changes to
# torchvision to be compatible with both < and >= v7.0.0.
pip install pillow==6.2.1
pip install pillow==7.0.0 --progress-bar off
#install the latest version of PyTorch and TorchVision
pip install torch torchvision
pip install torch "torchvision>=0.5.0" --progress-bar off
#run unit test
cd ${PROJ_ROOT}/ios/TestApp/benchmark
python trace_model.py
Expand Down
6 changes: 6 additions & 0 deletions .circleci/verbatim-sources/pytorch-build-params.yml
Original file line number Diff line number Diff line change
Expand Up @@ -29,10 +29,16 @@ pytorch_ios_params: &pytorch_ios_params
ios_platform:
type: string
default: ""
op_list:
type: string
default: ""
environment:
BUILD_ENVIRONMENT: << parameters.build_environment >>
IOS_ARCH: << parameters.ios_arch >>
IOS_PLATFORM: << parameters.ios_platform >>
SELECTED_OP_LIST: << parameters.op_list >>





Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -51,12 +51,17 @@
build_environment: "conda 2.7 cpu"
requires:
- setup
# This job has an average run time of 3 hours o.O
# Now only running this on master to reduce overhead
- binary_mac_build:
name: binary_macos_libtorch_2_7_cpu_build
build_environment: "libtorch 2.7 cpu"
requires:
- setup

filters:
branches:
only:
- master
- binary_linux_test:
name: binary_linux_manywheel_2_7mu_cpu_devtoolset7_test
build_environment: "manywheel 2.7mu cpu devtoolset7"
Expand Down
9 changes: 3 additions & 6 deletions .circleci/verbatim-sources/workflows-docker-builder.yml
Original file line number Diff line number Diff line change
Expand Up @@ -17,12 +17,6 @@
- docker_build_job:
name: "pytorch-linux-xenial-cuda10.1-cudnn7-py3-gcc7"
image_name: "pytorch-linux-xenial-cuda10.1-cudnn7-py3-gcc7"
- docker_build_job:
name: "pytorch-linux-xenial-cuda8-cudnn7-py2"
image_name: "pytorch-linux-xenial-cuda8-cudnn7-py2"
- docker_build_job:
name: "pytorch-linux-xenial-cuda8-cudnn7-py3"
image_name: "pytorch-linux-xenial-cuda8-cudnn7-py3"
- docker_build_job:
name: "pytorch-linux-xenial-cuda9-cudnn7-py2"
image_name: "pytorch-linux-xenial-cuda9-cudnn7-py2"
Expand All @@ -47,6 +41,9 @@
- docker_build_job:
name: "pytorch-linux-xenial-py3.5"
image_name: "pytorch-linux-xenial-py3.5"
- docker_build_job:
name: "pytorch-linux-xenial-py3.8"
image_name: "pytorch-linux-xenial-py3.8"
- docker_build_job:
name: "pytorch-linux-xenial-py3.6-clang7"
image_name: "pytorch-linux-xenial-py3.6-clang7"
Expand Down
2 changes: 1 addition & 1 deletion .circleci/verbatim-sources/workflows-ecr-gc.yml
Original file line number Diff line number Diff line change
Expand Up @@ -10,7 +10,7 @@
- ecr_gc_job:
name: ecr_gc_job_for_pytorch
project: pytorch
tags_to_keep: "271,262,256,278,282,291,300,323,327,347,389,401,402,403,405"
tags_to_keep: "271,262,256,278,282,291,300,323,327,347,389,401,402,403,405,a8006f9a-272d-4478-b137-d121c6f05c83,07597f23-fa81-474c-8bef-5c8a91b50595"
- ecr_gc_job:
name: ecr_gc_job_for_caffe2
project: caffe2
Expand Down