Summary:
Resubmit #20698 which got messed up.
Idea is that when PyTorch is used in a custom build environment (e.g. Facebook), it's useful to track usage of various APIs centrally. This PR introduces a simple very lightweight mechanism to do so - only first invocation of a trigger point would be logged. This is significantly more lightweight than #18235 and thus we can allow to put logging in e.g. TensorImpl.
Also adds an initial list of trigger points. Trigger points are added in such a way that no static initialization triggers them, i.e. just linking with libtorch.so will not cause any logging. Further suggestions of what to log are welcomed.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/20745
Differential Revision: D15429196
Pulled By: dzhulgakov
fbshipit-source-id: a5e41a709a65b7ebccc6b95f93854e583cf20aca
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/20108
Add cpp runs for c2, hooked up via pybinds. Print output to terminal. This is not hooked up with the pep output yet because I'd like to verify the numbers first.
Note that this isn't quite the same mechanism as the pytorch cpp hookup, which uses cpp_python_extensions. If I can use the same mechanism to pull all the inputs for c2 through cpp and do FeedBlobs in cpp, then I'll switch to that.
Reviewed By: zheng-xq
Differential Revision: D15155976
fbshipit-source-id: 708079dacd3e19aacfe43d70c5e5bc54da2cf9e3
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/19154
I recently saw some weird workflow error due to empty but set net_type. Maybe we should just fallback to simple net in this case.
Reviewed By: dzhulgakov
Differential Revision: D14890072
fbshipit-source-id: 4e9edf8232298000713bebb0bfdec61e9c5df17d
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/15041
Adding an alternative implementation of a task graph based on TBB
Reviewed By: dmudiger
Differential Revision: D13412517
fbshipit-source-id: f5efedd680bbe0072bf38d504e5682ab51dd630f
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/13320
simple_async has been deprecated via the network override rule for a while,
and we should be able to safely remove it.
This also clears up 2 tech debts:
(1) in rnn executor, rely on the executor override to get the right net.
(2) clearly mark checkExecutorOverride as a potential change to net_type by making it c++ style guide compliant.
Reviewed By: dzhulgakov
Differential Revision: D12840709
fbshipit-source-id: 667702045fa024f5bdc87a9c28ea1786c78432b3
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12714
This is a short change to enable c10 namespace in caffe2. We did not enable
it before due to gflags global variable confusion, but it should have been
mostly cleaned now. Right now, the plan on record is that namespace caffe2 and
namespace aten will fully be supersets of namespace c10.
Most of the diff is codemod, and only two places of non-codemod is in caffe2/core/common.h, where
```
using namespace c10;
```
is added, and in Flags.h, where instead of creating aliasing variables in c10 namespace, we directly put it in the global namespace to match gflags (and same behavior if gflags is not being built with).
Reviewed By: dzhulgakov
Differential Revision: D10390486
fbshipit-source-id: 5e2df730e28e29a052f513bddc558d9f78a23b9b
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12295
Add ability to use custom implementations of thread pool instead of TaskThreadPool
Reviewed By: yinghai
Differential Revision: D10046685
fbshipit-source-id: 95da8c938b8e60b728484c520319b09b0c87ff11
Summary:
This does 6 things:
- add c10/util/Registry.h as the unified registry util
- cleaned up some APIs such as export condition
- fully remove aten/core/registry.h
- fully remove caffe2/core/registry.h
- remove a bogus aten/registry.h
- unifying all macros
- set up registry testing in c10
Also, an important note that we used to mark the templated Registry class as EXPORT - this should not happen, because one should almost never export a template class. This PR fixes that.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/12077
Reviewed By: ezyang
Differential Revision: D10050771
Pulled By: Yangqing
fbshipit-source-id: 417b249b49fed6a67956e7c6b6d22374bcee24cf
* [fix] fixup the bias multiplier data access issue
Hotfix for failues in conv_transpose
* [D2][Easy]: lint regularizer
lint with black
* [GanH]: Split mu in adaptive weight for diagnose
* [Dper] Add the ability to split FC weights into multiple smaller ones
* fix SumReduceLikeOp for empty blob
as desc.
* add ctc_greedy_decoder for caffe2
ctc_greedy_decoder same as tf's
* Update event callback handling
Allow multiple callbacks per event
* Add WeightedSum layer
The motivation is to do weighted sum in HoNet/crossnet, in the next diff, I'll replace model.Add with model.WeightedSum in
honet: https://fburl.com/f4rmolg2
crossnet: https://fburl.com/v7awn8se, https://fburl.com/63filbnm
* Replicate DAG's behavior
Some callers expect RunAsync to block, replicate that behavior in case of
explicit 'dag' net type
* [dper] layernorm layer
as title
* Override dag, async_dag, async_polling
Overriding dag, async_dag and async_polling with async_scheduling
* Name the thread pools
Caffe thread pools currently inherit the thread names from the thread that starts them, which can be misleading. Give them an explicit name instead.
* [Caffe2] FilleOp should support int64_t dimensions
Change argument type to int64_t for shape argument of FillerOp (used in ConstantFill, XavierFill, etc)
* Remove caffe2/caffe2/contrib/torch/
It's not used anywhere and depends on old lua torch that conflicts with Aten. Given PT1 it's not relevant any more (though it was nice and clever code!)
#accept2ship
* Fix linearWarmup multiplier check
The multiplier needs to be non-negative, not strictly positive.
* Revert D3314316
This is after 2 years and we do not seem to have a use case for this one, so
for the sake of clean API design we should potentially remove this. This would
allow us to potentially pass in arguments to optionally construct an object,
although it is indeed a little bit unclear how we can reuse existing objects if
constructor arguments are passed in. In any case, we may want to remove this
dangling feature.
* Speedup generate proposals by partial_sort.
Speedup generate proposals by partial_sort.
FACEBOOK:
- Saw speed improvement for training with this op.
- Yanghan benchmarked the op on a small dataset and see consistent 100% improvement on speed (6ms -> 3ms) on 420 input resolution. See next diff for details.
* More parallel processing friendly for CPP version of GenerateProposals.
More parallel processing friendly for CPP version of GenerateProposals.
* [DT] [43/n] Lift stop conditions inside reader code back to flow control
1. Split multi_reader function into local_reader and remote_reader
2. Lifted stop conditions inside Limiter back to flow control
3. Split epoch flow building logic into 3 cases:
- single machine (1 reader, 1 trainer on trainer0 node, no PS)
- (1 reader + 1 trainer) on trainer0 node, has PS
- multiple readers, readers do not share nodes with trainers, might have PS or not
* Resolve conflicts for torch/_thnn/utils.py
* [Caffe2] Handle image decoding errors
Image decoding errors can make the whole training fail. This diff is to handle them
1.Catch imdecode exceptions and check if decoded image has zero columns or rows. This is counted as decoding errors.
2.Replace the image with empty in case of error
3.Count the number of errors and throw runtime exception if the rate reaches given number
The empty image data is kept. It might introduce noise in the training data.
* Update MKL exporter to IDEEP ops
TSIA
* [Caffe2] GlobalInit is thread safe, fixing the comment
With the mutex and lock, GlobalInit is thread safe.
Update the comments.
* Back out "Add support for generating ATen files during fbcode build"
Original commit changeset: 28970ddba353
@override-unit-failures
(Note: this ignores all push blocking failures!)
* [DT]: fix predictor save
similar to D6610058, here we add the fix for distributed online training
* Remove net_singlethread_async_gpu.cc
Closes https://github.com/caffe2/caffe2/pull/2528
This removes net_singlethread_async_gpu.cc as part of our effort to clean
CUDAContext and the net executors.
* Inline DFS task execution
Add a DFS inline task execution mode in executor
* Add c10 folder to fbcode
This adds the c10 folder and its test cases to fbcode. Build flags are mostly taken from aten.
* add dependencies for online trainer
Add some dependencies so that the online model can use DataPipeline and PredictionTransform operators
Relevent post: https://fb.intern.facebook.com/groups/1324375037655677/permalink/1740993462660497/
* Resolve conflicts for tools/jit/gen_jit_dispatch.py
* [Fix] sparse regularization in distributed training
* Support advanced pooling options in sum processor
* support advanced pooling options in sum processor
* remove redundant code
* support attention in sum processor
* Improve shard logging in net tracing code
Make it handle arbitrary shard ids instead of just one digit ids.
* [Caffe2] Call GlobalInit in predictor only in mobile
FACEBOOK:
Calling GlobalInit long after the program starts may not be safe. There are issues if the following happens:
User does not call GlobalInit and initFacebook after program starts
User sets a flag manually: https://fburl.com/mcsumw7d
User calls OSS predictor.
OSS predictor calls GlobalInit
GlobalInit calls initFacebook
initFacebook resets all flags: https://fburl.com/tolszha1
Thus, the user manually set flags are overwritten
This would happen anytime GlobalInit is called long after the program starts.
I suppose the intention of the user in this case is not to call GlobalInit throughout the program,
but use Caffe2 regardless (is that desired?)
But adding GlobalInit in the OSS predictor would automatically call GlobalInit when using Caffe2.
This issue doesn't exist in mobile, since initFacebook is not called on mobile.
For now, guard the GlobalInit in predictor for mobile only.
May want to ensure the GlobalInit is always called at the start of the program. @[3501714:kutta] has seen weird issues when not calling GlobalInit at the start of the program on server side. He has made some progress on this.
* resolve conflicts for caffe2/core/logging_is_google_glog.h and test/test_torch.py
* Add empty fix for SumLikeReduceOp
Add empty fix for SumLikeReduceOp
* Revert D7962948: [caffe2][nomnigraph] Concat elim for sparseNN
This reverts commit f7f434dc5c34ca6058b9765d2ef615453d2276a9
@bypass-lint
An infra SEV is better than not reverting this diff.
If you copy this password, see you in SEV Review!
@cause_a_sev_many_files
* Remove Declarations.yaml
* Include common.h
* Change std::stoi to caffe2::stoi
* Add thread_name.cc to the CMake file
* No need to subtract 1. Fix test segfaults
* Fix NetTest, ObserverTest
Fix tests
(cherry picked from commit 3767e66c3f365596cba3d46d3e7322c933a0ab41)
* CTCGreedyDecoderOp only has CPU implementation, test should only run on CPU
* Add a variable to avoid conversion resizing issue
* [fix] fixup the bias multiplier data access issue
Hotfix for failues in conv_transpose
* [D2][Easy]: lint regularizer
lint with black
* [GanH]: Split mu in adaptive weight for diagnose
* [Dper] Add the ability to split FC weights into multiple smaller ones
* fix SumReduceLikeOp for empty blob
as desc.
* add ctc_greedy_decoder for caffe2
ctc_greedy_decoder same as tf's
* Update event callback handling
Allow multiple callbacks per event
* Add WeightedSum layer
The motivation is to do weighted sum in HoNet/crossnet, in the next diff, I'll replace model.Add with model.WeightedSum in
honet: https://fburl.com/f4rmolg2
crossnet: https://fburl.com/v7awn8se, https://fburl.com/63filbnm
* Replicate DAG's behavior
Some callers expect RunAsync to block, replicate that behavior in case of
explicit 'dag' net type
* [dper] layernorm layer
as title
* Override dag, async_dag, async_polling
Overriding dag, async_dag and async_polling with async_scheduling
* Name the thread pools
Caffe thread pools currently inherit the thread names from the thread that starts them, which can be misleading. Give them an explicit name instead.
* [Caffe2] FilleOp should support int64_t dimensions
Change argument type to int64_t for shape argument of FillerOp (used in ConstantFill, XavierFill, etc)
* Remove caffe2/caffe2/contrib/torch/
It's not used anywhere and depends on old lua torch that conflicts with Aten. Given PT1 it's not relevant any more (though it was nice and clever code!)
#accept2ship
* Fix linearWarmup multiplier check
The multiplier needs to be non-negative, not strictly positive.
* Revert D3314316
This is after 2 years and we do not seem to have a use case for this one, so
for the sake of clean API design we should potentially remove this. This would
allow us to potentially pass in arguments to optionally construct an object,
although it is indeed a little bit unclear how we can reuse existing objects if
constructor arguments are passed in. In any case, we may want to remove this
dangling feature.
* Speedup generate proposals by partial_sort.
Speedup generate proposals by partial_sort.
FACEBOOK:
- Saw speed improvement for training with this op.
- Yanghan benchmarked the op on a small dataset and see consistent 100% improvement on speed (6ms -> 3ms) on 420 input resolution. See next diff for details.
* More parallel processing friendly for CPP version of GenerateProposals.
More parallel processing friendly for CPP version of GenerateProposals.
* [DT] [43/n] Lift stop conditions inside reader code back to flow control
1. Split multi_reader function into local_reader and remote_reader
2. Lifted stop conditions inside Limiter back to flow control
3. Split epoch flow building logic into 3 cases:
- single machine (1 reader, 1 trainer on trainer0 node, no PS)
- (1 reader + 1 trainer) on trainer0 node, has PS
- multiple readers, readers do not share nodes with trainers, might have PS or not
* Resolve conflicts for torch/_thnn/utils.py
* [Caffe2] Handle image decoding errors
Image decoding errors can make the whole training fail. This diff is to handle them
1.Catch imdecode exceptions and check if decoded image has zero columns or rows. This is counted as decoding errors.
2.Replace the image with empty in case of error
3.Count the number of errors and throw runtime exception if the rate reaches given number
The empty image data is kept. It might introduce noise in the training data.
* Update MKL exporter to IDEEP ops
TSIA
* [Caffe2] GlobalInit is thread safe, fixing the comment
With the mutex and lock, GlobalInit is thread safe.
Update the comments.
* Back out "Add support for generating ATen files during fbcode build"
Original commit changeset: 28970ddba353
@override-unit-failures
(Note: this ignores all push blocking failures!)
* [DT]: fix predictor save
similar to D6610058, here we add the fix for distributed online training
* Remove net_singlethread_async_gpu.cc
Closes https://github.com/caffe2/caffe2/pull/2528
This removes net_singlethread_async_gpu.cc as part of our effort to clean
CUDAContext and the net executors.
* Inline DFS task execution
Add a DFS inline task execution mode in executor
* Add c10 folder to fbcode
This adds the c10 folder and its test cases to fbcode. Build flags are mostly taken from aten.
* add dependencies for online trainer
Add some dependencies so that the online model can use DataPipeline and PredictionTransform operators
Relevent post: https://fb.intern.facebook.com/groups/1324375037655677/permalink/1740993462660497/
* Resolve conflicts for tools/jit/gen_jit_dispatch.py
* [Fix] sparse regularization in distributed training
* Support advanced pooling options in sum processor
* support advanced pooling options in sum processor
* remove redundant code
* support attention in sum processor
* Improve shard logging in net tracing code
Make it handle arbitrary shard ids instead of just one digit ids.
* [Caffe2] Call GlobalInit in predictor only in mobile
FACEBOOK:
Calling GlobalInit long after the program starts may not be safe. There are issues if the following happens:
User does not call GlobalInit and initFacebook after program starts
User sets a flag manually: https://fburl.com/mcsumw7d
User calls OSS predictor.
OSS predictor calls GlobalInit
GlobalInit calls initFacebook
initFacebook resets all flags: https://fburl.com/tolszha1
Thus, the user manually set flags are overwritten
This would happen anytime GlobalInit is called long after the program starts.
I suppose the intention of the user in this case is not to call GlobalInit throughout the program,
but use Caffe2 regardless (is that desired?)
But adding GlobalInit in the OSS predictor would automatically call GlobalInit when using Caffe2.
This issue doesn't exist in mobile, since initFacebook is not called on mobile.
For now, guard the GlobalInit in predictor for mobile only.
May want to ensure the GlobalInit is always called at the start of the program. @[3501714:kutta] has seen weird issues when not calling GlobalInit at the start of the program on server side. He has made some progress on this.
* resolve conflicts for caffe2/core/logging_is_google_glog.h and test/test_torch.py
* Add empty fix for SumLikeReduceOp
Add empty fix for SumLikeReduceOp
* Revert D7962948: [caffe2][nomnigraph] Concat elim for sparseNN
This reverts commit f7f434dc5c34ca6058b9765d2ef615453d2276a9
@bypass-lint
An infra SEV is better than not reverting this diff.
If you copy this password, see you in SEV Review!
@cause_a_sev_many_files
* Remove Declarations.yaml
* Include common.h
* Change std::stoi to caffe2::stoi
* Add thread_name.cc to the CMake file
* No need to subtract 1. Fix test segfaults
* Fix NetTest, ObserverTest
Fix tests
(cherry picked from commit 3767e66c3f365596cba3d46d3e7322c933a0ab41)
* CTCGreedyDecoderOp only has CPU implementation, test should only run on CPU
* Add a variable to avoid conversion resizing issue
* Remove the code per soumith's comments
* Remove the code per soumith's comments
* Remove blank lines in the end of file
* Resolve conflicts for torch/_thnn/utils.py
* Update MKL exporter to IDEEP ops
TSIA
* Back out "Add support for generating ATen files during fbcode build"
Original commit changeset: 28970ddba353
@override-unit-failures
(Note: this ignores all push blocking failures!)
* add dependencies for online trainer
Add some dependencies so that the online model can use DataPipeline and PredictionTransform operators
Relevent post: https://fb.intern.facebook.com/groups/1324375037655677/permalink/1740993462660497/
* Resolve conflicts for tools/jit/gen_jit_dispatch.py
* Support advanced pooling options in sum processor
* support advanced pooling options in sum processor
* remove redundant code
* support attention in sum processor
* resolve conflicts for caffe2/core/logging_is_google_glog.h and test/test_torch.py
* Revert D7962948: [caffe2][nomnigraph] Concat elim for sparseNN
This reverts commit f7f434dc5c34ca6058b9765d2ef615453d2276a9
@bypass-lint
An infra SEV is better than not reverting this diff.
If you copy this password, see you in SEV Review!
@cause_a_sev_many_files
* Remove Declarations.yaml
* Include common.h
* Change std::stoi to caffe2::stoi
* [caffe2] uprade IDEEP and hotfix for conv op accuracy issue (#8364)
* [IDEEP] Upgrade IDEEP version
Signed-off-by: Gu, Jinghui <jinghui.gu@intel.com>
* [IDEEP] Fix accuracy issue in conv op
Signed-off-by: Gu, Jinghui <jinghui.gu@intel.com>
* Fix build error due to lack of src in CMakeLists
Signed-off-by: Gu, Jinghui <jinghui.gu@intel.com>
* Remove the code per soumith's comments
* [ONNX] Add an ATen fallback pathway for ONNX export (#8273)
* ATen fallback for ONNX export
* Move to enum
* Fix model test
* Add comment
* Address comments
BC interface
* Remove imaginary file (#8415)
* [Caffe2] Enable AMD/MIOPEN ops for Caffe2 (#8306)
* Add hip support for caffe2 core
* Add MIOPEN header/wrapper to caffe2 core
* Add HIP device into caffe2 PB
* top level makefile change for rocm/hip
* makefile scaffolding for AMD/RocM/HIP
* Makefile scafodding for AMD/RocM/HIP; add makefile/utility for HIP files
* caffe2 PB update for AMD/ROCM HIP device
* Add AMD/RocM/Thrust dependency
* HIP threadpool update
* Fix makefile macro
* makefile fix: duplicate test/binary name
* makefile clean-up
* makefile clean-up
* add HIP operator registry
* add utilities for hip device
* Add USE_HIP to config summary
* makefile fix for BUILD_TEST
* merge latest
* Fix indentation
* code clean-up
* Guard builds without HIP and use the same cmake script as PyTorch to find HIP
* Setup rocm environment variables in build.sh (ideally should be done in the docker images)
* setup locale
* set HIP_PLATFORM
* Revert "set HIP_PLATFORM"
This reverts commit 8ec58db2b390c9259220c49fa34cd403568300ad.
* continue the build script environment variables mess
* HCC_AMDGPU_TARGET
* Cleanup the mess, has been fixed in the lastest docker images
* Assign protobuf field hip_gpu_id a new field number for backward compatibility
* change name to avoid conflict
* Fix duplicated thread pool flag
* Refactor cmake files to not add hip includes and libs globally
* Fix the wrong usage of environment variables detection in cmake
* Add MIOPEN CNN operators
* Revert "Add MIOPEN CNN operators"
This reverts commit 6e89ad4385b5b8967a7854c4adda52c012cee42a.
* Add MIOPEN pooling operator
* Add MIOPEN activation operator
* Add MIOPEN softmax operator
* Add MIOPEN spatial batch norm operator
* Add MIOPEN loacl response normalization operator
* Add MIOPEN conv operator
* Clean-up LRN ops
* enable fp16 in MIOPEN pool ops
* Enable fp16 for MIOPEN relu op
* Enable fp16 for MIOPEN spatial batch norm op
* code clean-up
* revert float16 support
* Create Caffe2 python binding for AMD/ROCM/HIP
* Add op fallback for HIP operator
* add hip src/test files in cmake
* exclude hip src/test files
* fix python binding for hip backend
* fix MIOPEN pooling op workspace
* hack to compile miopen operators
* fix include path for MIOPEN ops
* Fix include path
* Add HIP math utilities
* Fix path for HIP math utils
* cmake fix
* Cmake fix / hipcc for hip files
* suppress hipcc warning
* cmake fix /replcae USE_HIP with USE_ROCM
* revert LoadHIP.cmake change
* fix include for thrust/cub-hip
* include path fix for conversion.h
* Updated with latest upstream changes
* clang format fixes
* Context_hip updates
* Fixed typo in rocblas handle get function
* Updated hipified math utils
* Updated math hip test util
* Updated context hip test
* Updated common_hip
* Updated net async dag for HIP
* Added MIOPEN in operator hip test
* fix
* C2 dependencies clean-up
* fix include path for building custom protobuf
* Decouple miopen pool op and conv_pool_op base
* cmake refactor
* fix operator_hip_test
* move all hip/miopen ops files into caffe2/operators/hip
* sanitize cmake
* permission issue
* remove extra parenthesis
* remove artifact from resolving merge conflict
* cont. sanitize cmake files
* fix syntax error
* sanitize conversion.h
* .
* Revert "."
This reverts commit 56020cb0e996a31ae27bf1f8f491955ed0b121b9.
* clang-format
* Enable some reduce operators' ONNX backend tests (#8418)
* fix old comment to point to the right file (#8416)
* Stop pinning nccl version. (#8421)
Signed-off-by: Edward Z. Yang <ezyang@fb.com>
* Expose logsumexp docs and mark log_sum_exp in distributions for internal use (#8428)
* Enable some of the ONNX backend test on broadcasting (#8423)
* Enable some of the ONNX backend test on broadcasting
* enable gemm broadcast
* Expose proto utils and ONNX (#8073)
* Expose proto utils and ONNX from PyTorch libcaffe2.so
* Try to use protobuf from _C.so
* Fix ONNX proto header include
* Adjust order of imports for ONNX until nanopb goes away
* Set and use ONNX_NAMESPACE for PyTorch builds
* Show protobuf summary for all builds
* Add ONNX_NAMESPACE for cpp_build
* Statically link libprotobuf.a into libtorch.so
* Set ONNX_NAMESPACE on Windows build
* Move core/dispatch up as well
* Add /MD flag for Windows build of _C
* Potential Windows fix for ONNX and protobuf
* Add direct linkage from _C to ONNX on Windows
* Only include protobuf wrapper for PyTorch
* Pass extra_compile_args to _nvrtc ext build
* Remove installation of .a files
* Rebase creates some weird situations, revert them manually
* Remove more weird changes due to rebase
* Need to add thread_name.cc after merge
* Fix some signed/unsigned mismatches
* Skip unused result warning
* Explict fallthrough for murmur hash
* Enable aligned new support to eliminate warning
* Switch to int instead of unsigned in some cases
* [mpscnn] MPSCNNChannelShuffle
att
* [Easy] Adding tags as an argument to the functional layer
Without it "tags" would be added as an argument to the operator.
The change here is based on the assumption that there is no operator that takes "tags" as an argument.
* Fix locally_connected_op schema check.
Fix locally_connected_op schema check.
* [C2] Add TypeAndShape inference for few more operators
As desc
* [c2] Shape inference should support 0 as dimension
Tensors can have 0 in their dimension.
* Make MockHiveReader loop over and support max_examples
Replace DatasetReader with RandomDatasetReader.
So that Mock Hive Reader can simulate a large data input using a small sample file as source.
* Utility function to wipe cache between benchmark runs
Caffe2 benchmark does not wipe out cache between runs, and this potentially creates an unrealistically optimistic picture of performance. This diff adds utility function to wipe out the cache.
* Allow caffe2 GlobalInit to be invoked multiple times
Allow caffe2 GlobalInit to be invoked multiple times. Will re-parse gflags and update logging levels on successive invocations, but will not re-run init functions or perform other one-time initialization.
* Add Caffe2 GlobalInitIsCalledGuard to base net and operator classes
Warn if caffe2's GlobalInit function has not been invoked before creating an operator or net object. This is based on discussion here: https://fb.quip.com/kqGIAbmK7vNG
* Rethrow current exception on failure
Rethrow current exception instead of copy constructing a new one on op failure.
* Make `clone()` return subclass of List/Struct
`clone()` is not working correctly when we subclass those classes
* Wipe the cache before the net run
the util function is copied from D7409424
will rebase once D7409424 is landed.
* [Caffe2] [Mobile] Support utils/cast.h::GetCastDataType with LITE_PROTO builds
* Correct includes
async_polling include -> async_base include
* Prepare execution flags for executor migration
Making async_scheduling aware of underlying net type to prepare for executor
migration
* Add operator level observers into async executor
Adding operator level observers into RunAsync operators' calls
* Cleanup TEST_Benchmark
Remove duplicate code and provide default implementation in NetBase
* [C2] Fix type and shape inference for binary comparison ops
As desc.
* Add GlobalInit to predictor to ensure initialization is always done before prediction
FACEBOOK:
Redo D7651453 the correct way.
Now use a static variable for the arguments passed to GLog
* Remove spammy log message
This method is currently used in various places inside Caffe itself.
* Disable events for operators inside a chain
We don't need to use events in operators within a chain because the chain is
always scheduled on a single stream, keeping only first and last event for
scheduling purposes
* Ensure correct finish run order
In rare cases we might call finishRun and trigger net's destruction while
another worker is still holding shared_ptr to a thread pool, that can cause
thread pool destruction from within a worker thread in case no other nets are
using the pool. This diff fixes the order of calling finishRun and also changes
pool() to return raw pointer to keep pool's ownership within the net
* Reduce unnecessary polling
Make sure we don't waste CPU by polling operators that we can set an efficient
callbacks on
* Squash commit of syncing 9506eeb from github to fbcode
Patch xplat buck fix
add virtual destructor to OptimizationPass
add virtual destructor to OptimizationPass
build fixes for sync
build fixes for sync
* Fix net tracing
Fix net tracing from async_scheduling
* Fix logging
* [bootcamp] Improve "Shape" operator to support axes specification
To improve .shape operator of Caffe2 to support x.shape(tensor, axes), which takes an optional int array "axes" as input. For example, x.shape(tensor, [1, 0]) will return the dimension for axis 1 and 0 following the specified order. For current version, "axes" input allows duplications and can have arbitrary length.
* Back out "Add barrier net that runs before training nets"
Original commit changeset: b373fdc9c30f. Need additional changes to some callers to support barrier failures.
* Change warning to verbose log to reduce log spam
The `LOG(WARNING)` was a bit spammy for regular use so lets just make it a `VLOG`.
* Extract the shared code from different caffe2_benchmark binaries
The OSS benchmark and Internal benchmark will share most functions in the benchmark.
* Support MFR in sequence training
As titled.
* Make knowledge distillation work with using logged prediction feature as teacher label.
1) Add loading raw dense feature as teacher label.
2) Optional calibration function for teacher label
3) Add teacher label into generic unit test
4) Deprecated TTSN workflow version using feature_options to config teacher label
* [C2/CUDA]: unjoined cross entropy sigmoid
as desc
* Add async_scheduling executor into deferrable_net_exec_test
Add async_scheduling into tests and fix some exception cases
* Fix Event disabled error
When disabling event in RNN ops make sure we don't call Finish on disabled
event from op's RunAsync
* cuda ensure cpu output op can handle both TensorCPU and TensorCUDA
as desc.
* [C2 Core] Infer input device option in C2 hypothesis_test checkers
Improve how we default input blob device options.
Previously it defaults as where op lives but it is not necessarily the case.
For example:
CopyCPUToGPU
* [C2 Op]SplitByLengthsOp CPU/GPU implementation
[C2 Op]SplitByLengthsOp CPU/GPU implementation
* fix undefined symbol error
not sure why we're getting undefined symbol even with link_whole = True
Need to figure out why but need this workaround for now
* Add tools in DAIPlayground platform to help debugging models
Add additional tools to allow Plauground override individual method defined in AnyExp. This will allow user to create module that specificly change certain default method behavior. An example included in this diff is deactivating test model and checkpointing. When debugging any model problems, switching off components helps me quickly narrow down the location of the bug. The technique is extensively used in task T27038712 (Steady memory increase in EDPM, eventually resulting in gloo/cuda.cu:34: out of memory)
* add shape and type inference for int8 conversion operator
* Fix flaky test for group_norm
Fix flaky test for group_norm
* Fix group_norm_op_test flaky
Fix group_norm_op_test flaky
* Implementation of composite learning rate policy
In many state-of-the-arts deep learning works, people use a simple trick to
schedule the learning rate: use a fixed learning rate until error plateaus
and then switch to a different fixed learning rate, and so on. In this diff,
we implemented a simple version of the composite learning rate. The user gives
a set of learning rates policies and corresponding iteration nums, and the
optimizer will change the learning rate policy based on the number of iterations so far.
For example, the user give two learning rate policies, one is FixedLearningRate
and PolyLearningRate, with an iteration number of 1k. Then the first 1k iteration,
we use FixedLearningRate. For the following iterations, we use PolyLearningRate.
* Split two use cases of CachedReader into two classes, DBFileReader and CachedReader
# Use Cases:
1). input: DB file -> output: DatasetReader.
Use DBFileReader.
2). input: Reader -> build cache DB file -> output: DatasetReader.
Use CachedReader.
# Changes to CachedReader:
1). Move db_path to the constructor.
Because in mock reader. cache will always be built ahead.
# Changes to tests:
1). Make a separate TestCase class for CachedReader and DBFileReader.
2). Make it possible to add more test functions by adding setUp, tearDown and _make_temp_path.
3). Make delete db_path more general. `db_path` could be a file for `log_file_db`, but could also be a directory for `leveldb`.
* Back out "On Mobile phones, call GlobalInit with no arguments in predictor in case we need to perform initialization"
Original commit changeset: 4489c6133f11
* Fix LARS bug
Fixed a bug in the LARS implementation which caused all subsequent blobs not using LARS to have the LARS learning rate multiplier applied to them.
* [tum] support sparse init & add uniformFill option
as title
* Propagate exception for async nets
Capture the exception when an exception is thrown in async nets and re-throw it after wait(). This allows exceptions to be propagated up to the caller.
This diff was a part of D7752068. We split the diff so that C2 core files changes are in a separate diff.
* Automatic update of fbcode/onnx to 69894f207dfcd72d1e70497d387201cec327efbc
Previous import was 403ccfbd0161c38f0834413d790bad0874afbf9a
Included changes:
- **[69894f2](https://github.com/onnx/onnx/commit/69894f2)**: Use op schema.all tensor types in random like definitions (#865) <Scott McKay>
- **[b9d6b90](https://github.com/onnx/onnx/commit/b9d6b90)**: Clarify random like operators (#846) <Scott McKay>
- **[fc6b5fb](https://github.com/onnx/onnx/commit/fc6b5fb)**: Refactor shape inference implementation (#855) <anderspapitto>
- **[b7d8dc8](https://github.com/onnx/onnx/commit/b7d8dc8)**: fix cmake warning message (#863) <Eric S. Yu>
- **[f585c5d](https://github.com/onnx/onnx/commit/f585c5d)**: add pytorch-operator test for tile (#831) <Wenhao Hu>
- **[993fe70](https://github.com/onnx/onnx/commit/993fe70)**: add install step (#832) <Eric S. Yu>
- **[68bc26c](https://github.com/onnx/onnx/commit/68bc26c)**: add type inference for traditional ml ops except classifier ops. (#857) <Ke Zhang>
- **[9cc0cda](https://github.com/onnx/onnx/commit/9cc0cda)**: fix string representation of scalar types (#858) <G. Ramalingam>
- **[1078925](https://github.com/onnx/onnx/commit/1078925)**: fix y in pow test case to scalar (#852) <Wenhao Hu>
- **[c66fb6f](https://github.com/onnx/onnx/commit/c66fb6f)**: Add some math function shape inference (#845) <anderspapitto>
- **[ff667d1](https://github.com/onnx/onnx/commit/ff667d1)**: Refactor return type and docs for ONNXIFI_BACKEND_DIRECTX_ID (#853) <Marat Dukhan>
- **[11c6876](https://github.com/onnx/onnx/commit/11c6876)**: clear initializer names when clear initializer (#849) <Wenhao Hu>
- **[73c34ae](https://github.com/onnx/onnx/commit/73c34ae)**: Clarify FeatureVectorizer description. (#843) <Scott McKay>
- **[1befb9b](https://github.com/onnx/onnx/commit/1befb9b)**: Remove useless text in docs (#850) <Lu Fang>
- **[e84788f](https://github.com/onnx/onnx/commit/e84788f)**: Fix SELU attributes' default values (#839) <Lu Fang>
- **[ebac046](https://github.com/onnx/onnx/commit/ebac046)**: Add tile test case (#823) <Wenhao Hu>
- **[8b7a925](https://github.com/onnx/onnx/commit/8b7a925)**: a few more shape inference functions (#772) <anderspapitto>
- **[9718f42](https://github.com/onnx/onnx/commit/9718f42)**: Make the coefficient non optional for LinearClassifier (#836) <Jaliya Ekanayake>
- **[ef083d0](https://github.com/onnx/onnx/commit/ef083d0)**: Add save_tensor and load_tensor functions for Protos (#770) <Lu Fang>
- **[45ceb55](https://github.com/onnx/onnx/commit/45ceb55)**: Check if CMAKE_BUILD_TYPE set before project(). (#812) <Sergii Dymchenko>
- **[4b3d2b0](https://github.com/onnx/onnx/commit/4b3d2b0)**: [WIP] reenable shape inference tests (#834) <anderspapitto>
- **[22d17ee](https://github.com/onnx/onnx/commit/22d17ee)**: RNN tests: LSTM, GRU, SimpleRNN (#739) <Peyman Manikashani>
- **[de65b95](https://github.com/onnx/onnx/commit/de65b95)**: dimension denotation (#443) <Tian Jin>
- **[eccc76e](https://github.com/onnx/onnx/commit/eccc76e)**: fix field number issue in onnx operator proto and enable its build (#829) <Ke Zhang>
- **[d582beb](https://github.com/onnx/onnx/commit/d582beb)**: disable shape inference test to unbreak ci (#830) <Lu Fang>
- **[485b787](https://github.com/onnx/onnx/commit/485b787)**: function proto for composite op. (#802) <Ke Zhang>
- **[cd58928](https://github.com/onnx/onnx/commit/cd58928)**: specify defaults for attributes of Affine op (#820) <G. Ramalingam>
- **[7ee2cf9](https://github.com/onnx/onnx/commit/7ee2cf9)**: merge the dummy backend back into the main one (#743) <anderspapitto>
- **[1c03a5a](https://github.com/onnx/onnx/commit/1c03a5a)**: [Proposal] ONNX Interface for Framework Integration (previously ONNX Backend API) header and docs (#551) <Marat Dukhan>
- **[3769a98](https://github.com/onnx/onnx/commit/3769a98)**: Rename real model test case from VGG-16 to ZFNet (#821) <Lu Fang>
* [C2]ReluN Op
relu n op.
tf reference: https://www.tensorflow.org/api_docs/python/tf/nn/relu6
* Call destructor when assigning a blob value
* Add executor overrides
Add executor overrides flag to enable migration to async_scheduling executor
* Add barrier net that runs before training nets - attempt #2
Add a synchonize barrier net that is run before training nets. With this net, shards that are faster will wait for other shards before start training. This reduce chances of the faster shards timing out during GLOO AllReduce.
Removed explicit data_parallel_model.py.synchronize call in holmes workflow.
This change was landed previously but caused errors for some EDPM workflows - See https://fb.facebook.com/groups/1426530000692545/permalink/1906766366002237/ - because EDPM assumes any call to CreateOrCloneCommonWorld and Gloo ops are wrapped in exception handlers but in this case exception thrown in the barrier init net is not handled.
To address this issue, we add _CreateOrCloneCommonWorld to the param_init_net instead of a new barrier init net. Since errors for param_init_net run is handled gracefully and re-rendezvous, it should fixes the problem.
* Handle empty nets in async_scheduling
Make sure we don't get stuck on empty nets
* use CUDA_ARCH for conditional compile
* [C2 fix] infer function for ensure_cpu_output_op
* Update group_norm test to reduce flaky test
* Fix lr_multiplier for GPU
* Track checkpoint performance in scuba
As title.
* [C2/CUDA]: fix cross entropy sigmoid with logits
when adding log_d_trick, I forgot to add it to the cuda impl; this diff fixes
it.
* Back out "[caffe2] Unregister MKL fallbacks for NCHW conversions"
Original commit changeset: 8918dd40205a
Will land after @jongsoo's diff https://phabricator.intern.facebook.com/D7596315 lands
* [Easy][C2] Don't add blob to external outputs from output_record if it's already external output
As desc.
* On Mobile phones, call GlobalInit with no arguments in predictor in case we need to perform initialization
FACEBOOK:
The QPL logger needs the initialization code. In the past, the initialization code is put in the pipeline calling Caffe2. However, those places become obsolete quickly, as the product teams change places to call Caffe2 from time to time. We also need to track which teams use Caffe2 so that we can put the initialization code there.
With this diff, the initialization code is put in the predictor constructor, only enabled for mobile phones. This way, we can always enable QPL logging.
Once we do this, we can check how many times Caffe2 inference is called in production, and which models are more popular in production. This way, we can prioritize our effort supporting those models.
Will clean up the old code calling the init in the product in a separate diff.
* add padding op for sparse length tensor
to pad length-based sparse tensor with padding_value
* Add conv_op with cudaconvnet engine
Add conv_op with cudaconvnet engine
* [numa] Fix simple NUMA copy benchmark
Move XavierFill into init_net and also compute BW
* call roundf (device function) instead of round (host function)
* [caffe2_benchmark][observer] Make caffe2_benchmark use its own observer
1. Add ClearGlobalNetObservers()
2. Make caffe2_benchmark use its own observer and observer_reporter
* [detectron] Use roundf instead of round in the detectron module ops
* allow K larger than number of elements in top k op
one use case is to use this op together with PackSegments for sparse tensors, where the number of elements in each slice is not statistically defined.
* add ChannelShuffle DNNLOWP op
* fixup math_cpu.cc break
Summary:
Implementation of polling async net executor.
Notes:
- New net executor async_polling - schedules CPU and GPU ops asynchronously, uses single polling thread
- Events: update to Caffe2 events to support async CPU events, adding new methods:
Query() - non-blocking checking of event states: INITIALIZED -> RECORDED -> SUCCESS/FAILED
ErrorMessage() - when operation runs asynchronously and fails calling this on event will give error message
- Tasks: using existing DAGNet's algorithm to compute CPU and GPU chains, a separate task for each chain
- Polling: using single thread to query state of events - for CPU tasks atomically queries task state, for GPU task - uses cudaEventQuery; using Event
- Scheduling of CPU ops: using global thread pools
- Scheduling of GPU ops: using GPU thread pool per GPU device
Reviewed By: dzhulgakov
Differential Revision: D5985110
fbshipit-source-id: a9de7fcbb71d046a3aa1b573072b89a65dfeee8c
Summary: observer framework can now be used in python + a small writeup of how to use it. this is D6035393 with a fix for ct-scan
Reviewed By: salexspb
Differential Revision: D6066380
fbshipit-source-id: 896c4c580d4387240b81ac2dbbc43db51d4bfeb9
Summary: observer framework can now be used in python + a small writeup of how to use it
Reviewed By: sf-wind
Differential Revision: D6035393
fbshipit-source-id: 4563cf0203095fa979bb2160621cd16dd22ff830
Summary: observer framework can now be used in python + a small writeup of how to use it
Reviewed By: salexspb
Differential Revision: D5905002
fbshipit-source-id: e40ec24a55e08fb73beea9b4f3b68e71fc66ffb1
Summary:
running ##xplat/caffe2/fb_sync.sh##.
Also add two new core sources to the BUCK file, and add ##createSharedBuffer## to NNPACKConvOp.
Reviewed By: ajtulloch
Differential Revision: D5373061
fbshipit-source-id: c030b2629d2715e1d2776c98715f57e2650922c9
Summary: SimpleNet and DAGNetBase are the only two direct subclasses of NetBase. This feature has already been applied to SimpleNet before, with this diff all nets should be covered.
Reviewed By: dzhulgakov
Differential Revision: D5475498
fbshipit-source-id: 339edac31d008ec1e4630d93d2e27d0f518f4ebb
Summary: Constructor should extract everything needed from NetDef instead of keeping it for usage after construction.
Reviewed By: akyrola
Differential Revision: D5469095
fbshipit-source-id: 288ea3243d85061ba9c018d2aef3b4d97485dd00
Summary:
As title. Not sure how did the unit test bug went through -
we should have push blocking test guarding it. Looks like sandcastle
thought that it was already broken
Reviewed By: jamesr66a
Differential Revision: D5340741
fbshipit-source-id: 76b2287fc2f746d85dd732b669ff89808bcbd497
Summary:
Last time I used uuid filled into OperatorDef. And operator_tracebacks was populated using traceback.extract_stack. There were several issues with this approach:
1. A random field in OperatorDef breaks workflows relying on memoization, i.e. when computation is skipped based on already computed result before.
2. Adding one more field revealed RNNs being non forward compatible wrt to new fields in there. prototxt format seems to not allow forward compatibility (thanks jamesr66a for the investigation!). For RNNs we need to swtich them to a more resilient approach. azzolini's proposed change to OperatorDef / NetDef would allow that by just nesting NetDef dirrectly inside OperatorDef without need for extra serialization.
3. traceback.extract_stack is very slow when executable is on a remote filesystem. It does one or more os.stat for each frame on the stack. For some cases it ended up being up to 15 extra minutes on model construction.
In this diff I use a different approach which should fix all those problems above.
1.2. are solved by not adding a new field at all. Instead I report operator idx wrt to a net it runs in. Thanks akyrola and dzhulgakov for the idea. Downside here is that operator list manipulation breaks the logic and separately created ops are not covered at all.
3. I solved this by operating on raw frames without using traceback and inspect modules which end up doing a lot of file system calls. See function extract_stacktace in core.py with additional comments.
Reviewed By: dzhulgakov
Differential Revision: D5286285
fbshipit-source-id: 626dd0f5f6b8b1d86bd6bf519078b122f43ddcaa
Summary: Contains the ObserverBase class and some unittests.
Reviewed By: bwasti, pietern
Differential Revision: D5099367
fbshipit-source-id: fabde126d3281729dfc772d63dbf363e5d649319
Summary: Unneeded for mobile, should go from 90kb to ~30kb or so.
Differential Revision: D4545466
fbshipit-source-id: 47945493895a8f72d17de684b0429c2c7b5564ed
Summary:
We have noticed that the number of chains computed is usually much larger than necessary, when there is a backward pass. For example having a network of 5 FCs with gradient operators (but no parameter updates) should yield only one chain, but instead over 20 were created. After adding parameter updates, the forward pass still should remain one chain, while the backward pass will be splintered.
Analysis showed that the problem was the dependices from forward ops to the gradient computation. But these are redundant since the gradient op is already dependent from the op via the full path over ops. Example:
fc1 -> fc2 ---> fc3 --> loss
| | | |
fc1grad <- fc2grad <- fc3grad <-
Here fc1 and fc1 grad have a direct dependency, but indirect dependency via fc2->fc3->[...]->fc1grad already covers that dependency.
To fix this, I added a pruning step prior to the chain computation. The chain computation is done on the pruned tree, but I do not modify the runtime chains for safety.
Pruning is based on following logic:
- if one of my direct parents is ancestor via an another traversar, I can remove the direct dependency
Pruning is extremely fast, linear in the number of dependencies.
Reviewed By: dzhulgakov
Differential Revision: D4500293
fbshipit-source-id: 0994ae6775c53378ea1e0074365cef041764a1b4
Summary:
This diff brings us to roughly par with Torch on ResNet memory usage. On batch size 32, Resnet-50 took 7497MiB, after this 5010 MiB. This will thus allow us to handle 64 images / GPU, or 256 images / 4 GPUs.
In addition, I added a special argument to DagNet that causes it to run only one thread for the first iteration. This is needed since there are allocations on the first iteration's backward pass due to gradient sharing, and this will cause NCCL to deadlock.
The sharing of gradient buffers requires inferring which gradients can share memory (i.e that they are not used concurrently). Previous memonger code uses topological sort, but rbgirshick showed that it does not work with tree-like models. Thus, I wrote a new optimization algorithm based on DFS. It takes about 0.25 secs / GPU on resnet-50, so is clearly fast enough.
Module data_parallel_model supports this feature natively.
Reviewed By: prigoyal
Differential Revision: D4363209
fbshipit-source-id: 73b11e7610438098bb11bff0af8075ab0cf2c0f1
Summary:
This is needed so that we stick with C++11 instead of 14, which are not well
supported in a few platforms.
Reviewed By: bwasti
Differential Revision: D4377534
fbshipit-source-id: d65d7caaa935a8f16e3b44c838104a576c8f78e4
Summary: provide a easy way to benchmark different dper models.
Differential Revision: D4367258
fbshipit-source-id: 4821645c58ad183becf0c82daae991375d5c6ef4
