Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/66056
keep running into this unrelated failure when landing diffs regarding the gpu inference project,
disabling this operator unit test in gpu because it doesn't exist
RuntimeError: [enforce fail at operator.cc:277] op. Cannot create operator of type 'SmartDecaySparseAdam' on the device 'CUDA'. Verify that implementation for the corresponding device exist. It might also happen if the binary is not linked with the operator implementation code. If Python frontend is used it might happen if dyndep.InitOpsLibrary call is missing. Operator def: input: "param" input: "mom1" input: "mom2" input: "last_seen" input: "indices" input: "grad" input: "lr" input: "iter" output: "param" output: "mom1" output: "mom2" output: "last_seen" name: "" type: "SmartDecaySparseAdam" arg { name: "beta1" f: 0 } arg { name: "beta2" f: 0.9 } arg { name: "epsilon" f: 1e-05 } device_option { device_type: 1 }
https://www.internalfb.com/intern/testinfra/diagnostics/5910974579962988.562949996565057.1633122845/
Test Plan: sandcastle
Reviewed By: jianyuh
Differential Revision: D31364731
fbshipit-source-id: 7fbd994cbe7f6ca116f5f34506a1ed7f14759bdf
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/62135
The initial implementation of Adam with Smart Decay had an off-by-one error. This was in the summation of the geometric series used to calculate how much built-up momentum would have been discharged in skipped minibatches.
The unit tests should have caught these, but the testing strategy missed this because k, the "number of skipped minibatches" was always either 0 or so high that the impact of the bug was too small. The impact of the bug was proportional to 1/k. The testing strategy has also been adjusted to cover this bug.
Differential Revision: D29889309
fbshipit-source-id: b086c0efed5c27f621061e726533c73658daffc6
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/61548
We want to decay learning parameters properly. Previously this was not done when a parameter is absent from a minibatch. We fix this by keeping track of missed minibatches and making decay catch up accordingly.
The exponential moving averages (EMA) for the first and second moments used in Adam are updated only for parameters seen in a minibatch. Actually, for these parameters, 0 should be added to the EMAs and the EMAs should then be decayed by multiplying by beta1 and beta2 respectively.
To avoid the computational overhead of touching every parameter for every minibatch, we:
* keep track of the last time a parameter is seen
* instead of decaying the EMAs by multiplying by beta1 and beta2, we multiply by beta1^k and beta2^k, where k is the number of minibatches since the parameter was last seen
* we calculate the amount of momentum that would have been discharged over the missed minibatches and update the weight accordingly.
Differential Revision: D29654246
fbshipit-source-id: 7a6cd7966eb1f31116d99dfce79a78b2d3ee9e3e
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/61488
We want to decay learning parameters properly. Previously this was not done when a parameter is absent from a minibatch. We fix this by keeping track of missed minibatches and making decay catch up accordingly.
The exponential moving averages (EMA) for the first and second moments used in Adam are updated only for parameters seen in a minibatch. Actually, for these parameters, 0 should be added to the EMAs and the EMAs should then be decayed by multiplying by beta1 and beta2 respectively.
To avoid the computational overhead of touching every parameter for every minibatch, we:
* keep track of the last time a parameter is seen
* instead of decaying the EMAs by multiplying by beta1 and beta2, we multiply by beta1^k and beta2^k, where k is the number of minibatches since the parameter was last seen.
Differential Revision: D27978269
fbshipit-source-id: e47524101ddfcb281c46c505b9b7a8f0835bc64a
Summary:
Context: https://github.com/pytorch/pytorch/pull/53299#discussion_r587882857
These are the only hand-written parts of this diff:
- the addition to `.github/workflows/lint.yml`
- the file endings changed in these four files (to appease FB-internal land-blocking lints):
- `GLOSSARY.md`
- `aten/src/ATen/core/op_registration/README.md`
- `scripts/README.md`
- `torch/csrc/jit/codegen/fuser/README.md`
The rest was generated by running this command (on macOS):
```
git grep -I -l ' $' -- . ':(exclude)**/contrib/**' ':(exclude)third_party' | xargs gsed -i 's/ *$//'
```
I looked over the auto-generated changes and didn't see anything that looked problematic.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/53406
Test Plan:
This run (after adding the lint but before removing existing trailing spaces) failed:
- https://github.com/pytorch/pytorch/runs/2043032377
This run (on the tip of this PR) succeeded:
- https://github.com/pytorch/pytorch/runs/2043296348
Reviewed By: walterddr, seemethere
Differential Revision: D26856620
Pulled By: samestep
fbshipit-source-id: 3f0de7f7c2e4b0f1c089eac9b5085a58dd7e0d97
Summary:
There is a module called `2to3` which you can target for future specifically to remove these, the directory of `caffe2` has the most redundant imports:
```2to3 -f future -w caffe2```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/45033
Reviewed By: seemethere
Differential Revision: D23808648
Pulled By: bugra
fbshipit-source-id: 38971900f0fe43ab44a9168e57f2307580d36a38
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/20501
Fixing unit tests related to optimizer related operators and tests
Reviewed By: hx89
Differential Revision: D15307410
fbshipit-source-id: e5400c26e08f26191ee542fe6b02e0a69bc4e1ae
* Fix handling of empty batches in SumReduceDimsOp
As titled
* Deferrable async_scheduling finishRun fix
Proper order of finishing run operations in deferrable_async_scheduling net
* Simplify exception handling in async_scheduling
Simplify exception handling, no need to busy wait, thread that processes the
last task can finish the run
* [C2]worker_coordinator_memorize_worker_ids
As titled. This is related to T28689868, where the number of blobs we want to create is equal to the number of worker ids
* Add unit test for nets with no type set
* Ignore total length argument in sympolic_pad_packed_sequence
1- There was a mistake in the code that total_length was added to the wrong symbolic function (pack_padded_sequence) instead of (pad_packed_sequence)
2- No need to throw an exception if total_length is given since it is only used to enable data_parallel training on multi-gpus and doesn't have anything to do with onnx export, so just ignore it. https://fburl.com/tk4gciqp
* Add support for MKLDNN to async_scheduling
Just add MKLDNN as a possible CPU option to async_scheduling's pool function
* [AuFL][ensemble] support branch output for prediction
This diff supports using predictions from different branches and thus enables model ensembling (not fully independent).
* Fix a bug in add_loss in layer_model_helper
As titled.
* Support lradaption for adam
1.lr adaption operator
2.apply to dense adam
* Perf tweaks for async_scheduling
Restore single pool option + remove unnecessary (no-ops) calls
* add quantization to SparseSimdAdagradOp
add a bunch of quantization signatures to SparseSimdAdagradOp, implementations to come next
* [sr] [codemod] Change all SR callsites to use new API
@allow-large-files
This diff refactors all callsites of SR to use the slightly changed API introduced in the diff below. Really what this means is that you need to include the correct header. Also if you were using `ClientFactory::newFactory` you need to not prefix it with `ClientFactory::`.
```
cd ~/fbsource/fbcode
find ./ -type f -exec sed -i -e 's:#include "servicerouter/client/cpp2/ClientFactory.h":#include "servicerouter/client/cpp2/ServiceRouter.h":' -e 's:#include <servicerouter/client/cpp2/ClientFactory.h>:#include <servicerouter/client/cpp2/ServiceRouter.h>:' -e 's/ClientFactory::newFactory(/newFactory(/g' {} \;
```
Also manually fixed spots that couldn't be done automatically (or broke because they depended on transitive includes).
* Back out "Fix handling of empty batches in SumReduceDimsOp"
Original commit changeset: 282da1730cc2 This commit is blocking the
Github->fbcode sync, which really needs to get merged ASAP. D7881937 which this
diff depends on will be reverted in the sync D7990948 which causes this to
break. The sync diff cannot be patched with this reversion because it must be
landed against base revision 5c8c099 , and D7881937 must not be included in the
sync diff because it is breaking GPU tests that are not available in sandcastle
: https://ci.pytorch.org/jenkins/job/caffe2-builds/job/py2-cuda8.0-cudnn6-ubuntu16.04-test/3638/console
for one example.
* Add the flow to support operator benchmark
1) generate model with the operator 2) upload to everstore 3) generate model spec into json file 4) start running the benchmark
* [tum][gpu] Connect DPM trainer with flow and unit tests
This diff:
- Fix some small bugs for Yiming's recent changes to parallelizer, so it suits real use cases.
- Add correct tags to the TUM code, so we can do data parallel transform
- pass extra info when instantiation.
- add unit test for using DPM in TUM model
After this diff, we can do simple box, multi-gpu fully-sync trainer for TUM in Fblearner workflow, but may still need to do speed benchmarking.
* w/o normalized lradaption for adam dense only
The previous lr adaption includes a normalization step when performing the dot product operation. This is not exactly same as what is proposed in the paper. I add normalization as an option. Without it, the operator performs exactly what the paper proposed. With the option, we add the normalization step
* [fb] Use SharedPromise in DeferrableAsyncSchedulingNet
This code is to simplify DeferrableAsyncSchedulingNet by removing condition
variable + small fixes
* [tum] implement cuda sparseLengthsMean and LengthsMean
as title
* Adding an optional parameter to allow use of protobufs in InferShapesAndTypes function.
Adding an optional parameter to allow use of protobufs in InferShapesAndTypes function.
* Move feature_to_index to FeatureSpec.feature_to_index
move feature_to_index to FeatureSpec.feature_to_index to avoid override other fields
* [Caffe2] Rename bytes_moved to bytes_written
Just a rename in preparation for supporting bytes_read.
* [c2] fix ReduceFrontSumOp for empty case by setting 0
otherwise, it may use the results from last iteration when it's empty batch.
* [Caffe2] [Int8] Improve Intel CPU performance
* [Easy] Improve PrependDim op logging
as titled
* DBFileReader expand db_path using os.path.expanduser(..)
Since there are a lot of possible use cases of `DBFileReader` to read from user home path, like `~/local/sample.db`, I want to save people's trouble of calling `os.path.expanduser(db_path)` themselves.
* [Caffe2] Add bytes_read to cost structure
We're adding analytical read bytes to cost functions. This extends the structure accordingly for all CostInference defined operators.
Additionally, some small bug fixes were performed:
1) Cost functions now extract type information of operands instead of assuming float
* Fix sleef on aarch64 for hhvm
@bypass-lint
Rename flag
* Remove duplicated part in caffe2/ideep/operators/conv_op.cc
should be sync error
* Rename test helper function test_adagrad_sparse_helper to adagrad_sparse_test_helper to avoid confusing pytest
Summary: The original implementation averaged the momentum across the embedding dimensions, which doesn't make any sense. This meant all the embedding dimensions received the same update, becoming a very memory-expensive one-dimensional embedding.
Differential Revision: D7003135
fbshipit-source-id: ed54e3427bc13895a4e949e96b4b17f6ebfb6d53
Summary: Added the RowWise functionality for SparseAdam, which saves roughly 2/3 memory usage by only keeping one first and second moment term for each row of the parameter tensor, rather than one for each individual parameter.
Differential Revision: D6679342
fbshipit-source-id: ce6fb27e35ce41a890c66f6089cd2748d10e7a44
Summary:
There were no dimensionality constraints to the generated indices
array, causing many examples being generated and filtered out. Instead,
we should ensure the probability of unique indices is high.
There is a better fix for this by using the `unique` keyword argument
to `hypothesis.extra.numpy.arrays`, but this is available only in
hypothesis version 3.28.0 and later.
This is related to #1536 and #1599.
Once this change has proven to be OK, we can modify the other tests
that now have health check suppression enabled as well.
Closes https://github.com/caffe2/caffe2/pull/1686
Reviewed By: Yangqing
Differential Revision: D6651789
Pulled By: pietern
fbshipit-source-id: d80886c9ccf0a7a842a7580a279f33a2d6cca97c
Summary:
These GPU paths are probably even buggier than the CPU paths for sparse gradients with duplicate indices. Both paths cause multiple momentum updates in a single iteration, but only the GPU path is non-deterministic. Depending on how we decide to address the issues on the CPU path, pooyadavoodi has a good idea for how to match dense behavior with the sparse GPU ops.
Closes https://github.com/caffe2/caffe2/pull/254
Reviewed By: bwasti
Differential Revision: D4871680
Pulled By: dzhulgakov
fbshipit-source-id: 220be57a0f699a22ea85ed4f7022d92d362d06b3
