pytorch / pytorch Public

20 Pull requests merged by 13 people

• Merge pytorch master into lazy_tensor_staging

  #66449 merged Oct 12, 2021

• [LTC] Code-gen native_layer_norm

  #66192 merged Oct 11, 2021

• fix mean out shape formula

  #66397 merged Oct 11, 2021

• [JIT] Initialize CUDA context before launching fused kernel (#65064)

  #66349 merged Oct 8, 2021

• [LTC] Code-gen dropout

  #66115 merged Oct 8, 2021

• Fix cosine similarity dim checks

  #66214 merged Oct 8, 2021

• [ONNX] Deprecate various args

  #65962 merged Oct 8, 2021

• Convert generator attached to Sampler back to lazily construction

  #65926 merged Oct 8, 2021

• Revert "Added option to update parameters using state_dict in AveragedModel (#65495)"

  #66308 merged Oct 8, 2021

• [LTC] Code-gen gelu and gelu_backward

  #66106 merged Oct 8, 2021

• broadcasting doc fix

  #66029 merged Oct 7, 2021

• [ONNX] Support conv-bn fusion in blocks

  #66152 merged Oct 7, 2021

• [ONNX] Use Reciprocal operator instead of Div(1, x).

  #65382 merged Oct 7, 2021

• [ONNX] Add dim argument to all symbolic

  #66093 merged Oct 7, 2021

• add a script to merge LTC compile_commands.json with pytorch's

  #66243 merged Oct 7, 2021

• Added option to update parameters using state_dict in AveragedModel (#65495)

  #65755 merged Oct 6, 2021

• Tweak `file_diff_from_base` for release/1.10 branch

  #66202 merged Oct 6, 2021

• [DataPipe] DataPipe Fix and Deprecation Warnings for Release 1.10

  #65932 merged Oct 6, 2021

• [iOS][CI] Update dev certs (#66004)

  #66188 merged Oct 6, 2021

• Fix backward compatibility tests

  #66186 merged Oct 6, 2021

132 Pull requests opened by 71 people

• [ONNX] Update value name copying logic for onnx

  #66170 opened Oct 5, 2021

• Expose MobileCode to python

  #66171 opened Oct 5, 2021

• [ONNX] Relax check on Prim::PythonOp nodes for ONNX_FALLTHROUGH

  #66172 opened Oct 5, 2021

• [ONNX] Support tensorList when Infers shape and type uninitialized output (#60534)

  #66175 opened Oct 5, 2021

• Refactoring `aten::copy_` to use dispatcher

  #66193 opened Oct 6, 2021

• Fix strict aliasing rule violation in bitwise_binary_op

  #66194 opened Oct 6, 2021

• [numpy] add decimals argument to round

  #66195 opened Oct 6, 2021

• [FX] Specifies a default value when possible for placeholders created…

  #66213 opened Oct 6, 2021

• Refactor cub namespace handling

  #66219 opened Oct 6, 2021

• [PyTorch Edge] Clean up Quantize Tensor code

  #66220 opened Oct 6, 2021

• Implement derivative for narrow_copy

  #66225 opened Oct 6, 2021

• Add `warn_only` kwarg to `use_deterministic_algorithms`

  #66233 opened Oct 6, 2021

• use irange for loops

  #66234 opened Oct 6, 2021

• Autograd: Delay warnings until the end of backward execution

  #66235 opened Oct 6, 2021

• assert no duplicate yaml keys in codegen

  #66238 opened Oct 7, 2021

• [Bazel] Add CUDA build to CI

  #66241 opened Oct 7, 2021

• [qat]Make sure the bn statistics are the same in the unit test.

  #66244 opened Oct 7, 2021

• Fix common device computation for comparison ops.

  #66245 opened Oct 7, 2021

• Accept 0-dim channel inputs in convolution layer

  #66256 opened Oct 7, 2021

• Formatting changes

  #66257 opened Oct 7, 2021

• [Bootcamp][Pytorch Core] Add testing for complex non-vanilla SGD

  #66261 opened Oct 7, 2021

• Add uniform to onnx export

  #66263 opened Oct 7, 2021

• GHA: Stop skipping periodic jobs

  #66264 opened Oct 7, 2021

• [fxtrace] avoid N^2 selecting module name

  #66265 opened Oct 7, 2021

• Back out "Revert D30599136: [Pytorch Edge][tracing-based] build tracer in OSS"

  #66270 opened Oct 7, 2021

• dropout update in autodiff

  #66273 opened Oct 7, 2021

• [qat]Removed outdated context manager in unit test.

  #66274 opened Oct 7, 2021

• [torchelastic] Set correct default rdzv_id

  #66280 opened Oct 7, 2021

• Fix ubsan in concat_split_op.h

  #66283 opened Oct 7, 2021

• Fix ubsan resize in concat_split_op.h

  #66285 opened Oct 7, 2021

• [JIT] UseVariadicOp handles multiple lists

  #66288 opened Oct 7, 2021

• [Static Runtime][DI] Add variadic grouped_accessor_op

  #66289 opened Oct 7, 2021

• Add support for vmap+fwdAD for basic out-of-place op

  #66291 opened Oct 7, 2021

• Address case when layout of tangent is not same as base

  #66292 opened Oct 7, 2021

• Add internal assert for tangent layout mismatch for view ops

  #66293 opened Oct 7, 2021

• Add OpInfo test and fix a couple cases

  #66294 opened Oct 7, 2021

• [jit] Refcounting spot fixes in alias_analysis

  #66295 opened Oct 7, 2021

• add frozon_numpy as a builtin library to torch::deploy

  #66297 opened Oct 7, 2021

• [PyTorchEdge] make separate targets

  #66298 opened Oct 7, 2021

• [Draft PR] Run CI for OneDNN Graph Preview2

  #66299 opened Oct 8, 2021

• Syncbatchnorm 0 batch dim

  #66303 opened Oct 8, 2021

• Initialize operator versioning tracker

  #66304 opened Oct 8, 2021

• [hpc infer] Backout D31360987 and D31293307

  #66306 opened Oct 8, 2021

• [POC] functorch reimplemented using python mode stacks

  #66310 opened Oct 8, 2021

• [nnc] quantization: quant, dequant, add, conv2d

  #66313 opened Oct 8, 2021

• WIN-NNC

  #66314 opened Oct 8, 2021

• [Draft] Introduce `torch.distributed.nn.utils.init_meta()`

  #66317 opened Oct 8, 2021

• More irange

  #66318 opened Oct 8, 2021

• [nnc] quantization

  #66325 opened Oct 8, 2021

• [nnc] quantization conv2d prepack op

  #66326 opened Oct 8, 2021

• nnc - WIP

  #66327 opened Oct 8, 2021

• [PyTorch Edge] Use a RW Lock in RWSafeLeftRightWrapper<T> to protect concurrent read-write for on-device cases

  #66332 opened Oct 8, 2021

• [quant][embedding qat] Add eager QAT test for EmbeddingBag+Linear model

  #66334 opened Oct 8, 2021

• Use kernel_func_name from aotCompiler

  #66337 opened Oct 8, 2021

• [hpc infer] fix bench perf number

  #66339 opened Oct 8, 2021

• [jit] shape propagation for quantization

  #66343 opened Oct 8, 2021

• [nnc] quantization

  #66344 opened Oct 8, 2021

• [quant][embedding qat] Enable quint4 in EmbeddingBag QAT workflow

  #66348 opened Oct 8, 2021

• [sharded_tensor] easier initialization for Shard

  #66351 opened Oct 8, 2021

• add interactive_embedded_interpreter.cpp to the OSS build

  #66352 opened Oct 8, 2021

• OpInfos for: normal, bernoulli, multinomial

  #66358 opened Oct 9, 2021

• [fx2trt] open source tests for converters

  #66361 opened Oct 9, 2021

• Cherry-pick JIT trace feature from #59949

  #66367 opened Oct 9, 2021

• Add python bindings for JIT trace.

  #66368 opened Oct 9, 2021

• Add some graph-rewrite passes to prepare BT model for compilation.

  #66369 opened Oct 9, 2021

• [TensorExpr] Fix lowering for aten::softmax for the case when dtype parameter is None.

  #66370 opened Oct 9, 2021

• [TensorExpr] Add lowering for aten::conv1d.

  #66371 opened Oct 9, 2021

• [TensorExpr] Add lowering for aten::embedding.

  #66372 opened Oct 9, 2021

• [TensorExpr] Add lowering for aten::max (reduction).

  #66373 opened Oct 9, 2021

• Add example script for compiling bt model.

  #66374 opened Oct 9, 2021

• Cumsum Converter

  #66376 opened Oct 10, 2021

• [deploy] remove `self` from IntepreterSession

  #66377 opened Oct 10, 2021

• [LTC] Code-gen clamp

  #66383 opened Oct 10, 2021

• Strided masked reduction: amin.

  #66385 opened Oct 10, 2021

• Strided masked reduction: prod.

  #66386 opened Oct 10, 2021

• Remove THCNumerics

  #66388 opened Oct 10, 2021

• Remove THCTensorMathReduce.cuh

  #66389 opened Oct 10, 2021

• Migrate USE_MAGMA config macro to ATen

  #66390 opened Oct 10, 2021

• Remove THCGeneral.cpp

  #66391 opened Oct 10, 2021

• Back out "Revert D31299350: Back out "Revert D31005792: [NCCL] Init dummy NCCL comms in constructor""

  #66393 opened Oct 11, 2021

• Build tracer and Enable tracing-based build with tracer output

  #66395 opened Oct 11, 2021

• ATTEMPT 2 : BUG FIX: Assign should copy base class fields as well

  #66398 opened Oct 11, 2021

• [opinfo] elemwise parcel : isfinite, isinf, isposinf, isneginf, isnan, isreal

  #66400 opened Oct 11, 2021

• Sparse CSR CUDA: Support mixed memory format input for triangular_solve

  #66401 opened Oct 11, 2021

• ibm z14/15 SIMD support

  #66407 opened Oct 11, 2021

• CI for model tracer

  #66409 opened Oct 11, 2021

• [sparsity] Fix and enable the pruning tests

  #66411 opened Oct 11, 2021

• [sparsity] Fix GPU training for sparsity

  #66412 opened Oct 11, 2021

• [sparsity] Make pruner follow the sparsifier API

  #66413 opened Oct 11, 2021

• [sparsity] Remove mask from modules

  #66414 opened Oct 11, 2021

• [sparsity] Add axis support to the pruning parametrization

  #66415 opened Oct 11, 2021

• [sparsity] Add weight norm pruner

  #66416 opened Oct 11, 2021

• Updated error message for nn.functional.interpolate

  #66417 opened Oct 11, 2021

• Exposed `recompute_scale_factor` into nn.Upsample

  #66419 opened Oct 11, 2021

• [quant][embedding qat] eager mode QAT for Embeddings

  #66429 opened Oct 11, 2021

• log_sigmoid: Use log1p for improved precision

  #66441 opened Oct 11, 2021

• [torch.fx] Fix replace pattern mechanism

  #66442 opened Oct 11, 2021

• .circleci: re-add builds for bionic-cuda10.2

  #66448 opened Oct 11, 2021

• [tensorexpr] Add a stmt recorder to obtain stmt PCs

  #66450 opened Oct 11, 2021

• [tensorexpr] Add BufMap instruction to reuse the memory of dest buf for src buf

  #66451 opened Oct 11, 2021

• [tensorexpr] Add memory planning to reuse intermediate buffers

  #66452 opened Oct 11, 2021

• [tensorexpr] Remove Alloc insertion pass in LoopNest

  #66453 opened Oct 11, 2021

• DO NOT MERGE, removing previous check for docker images

  #66454 opened Oct 11, 2021

• use OpInfo to verify that LTC report correct shapes (shape formulas) …

  #66455 opened Oct 11, 2021

• fix lint error on line 1317

  #66457 opened Oct 11, 2021

• Introduce torch.distributed._shard package.

  #66459 opened Oct 11, 2021

• Fix for failure in test_autocast.py

  #66461 opened Oct 11, 2021

• Rename shard_lengths to shard_sizes to be more inline with Tensor sizes.

  #66464 opened Oct 12, 2021

• [qnnpack] Remove usage of conv_param_t in deconv-run.cc

  #66465 opened Oct 12, 2021

• Rename tensorexpr::Value so that it can coexist with torch::jit::Value

  #66467 opened Oct 12, 2021

• Use libcudacxx instead of thrust for complex when possible

  #66469 opened Oct 12, 2021

• [SR] Fix bug in ValueGroup

  #66470 opened Oct 12, 2021

• [FX] Fix cases when we should not fuse due to more than one users of intermediate node

  #66472 opened Oct 12, 2021

• [vulkan] Release GPU resources when vTensor::View is destroyed

  #66477 opened Oct 12, 2021

• [vulkan] Remove the persistent resource pool

  #66478 opened Oct 12, 2021

• Migrate THCState to ATen

  #66480 opened Oct 12, 2021

• [WIP][ONNX] Allow registration of custom symbolics for aten namespace

  #66481 opened Oct 12, 2021

• fx quant: enable linear-bn1d fusion for PTQ

  #66484 opened Oct 12, 2021

• Sparse CSR CUDA: fix input checks for `addmm` and `mm`

  #66485 opened Oct 12, 2021

• Use `torch.empty()` instead of `torch.tensor()` in `torch.nn.Parameter`

  #66486 opened Oct 12, 2021

• Make permission errors more human readable

  #66492 opened Oct 12, 2021

• Enable all dtype combinations in `torch.Tensor.view(dtype)`

  #66493 opened Oct 12, 2021

• Refactor saving jit::Module to mobile .pt in 2 steps:

  #66494 opened Oct 12, 2021

• [fx2trt][code quality] Add type annotation and docstring to utils functions in acc_ops_converters.py

  #66496 opened Oct 12, 2021

• Make ContainerHash compatible with const& types

  #66497 opened Oct 12, 2021

• Use torch/lazy hash instead of lazy_tensors::util

  #66498 opened Oct 12, 2021

• add backend_config_dict to checkGraphModeFxOp

  #66499 opened Oct 12, 2021

• WIP: Clean up the LazyTensor class

  #66500 opened Oct 12, 2021

• [Bootcamp][Pytorch]Add testing for complex parameters in Adagrad optimizer

  #66501 opened Oct 12, 2021

• [Testing] Error when dispatch occurs in parallel region

  #66505 opened Oct 12, 2021

• Fix Dispatching not considering List[Optional[Tensor]] for dispatch

  #66506 opened Oct 12, 2021

• [Static Runtime] Add FuseListUnpackV2

  #66509 opened Oct 12, 2021

68 Issues closed by 30 people

• [ONNX] Error when returning multiple outputs from nested conditional statements

  #65540 closed Oct 12, 2021

• PythonProcessGroupTest.test_collectives is flaky

  #66439 closed Oct 12, 2021

• Difference in `torch.normal` between pytorch 1.8.1 and pytorch 1.9.0

  #65709 closed Oct 12, 2021

• How to implement dynamic sampling of training data?

  #66418 closed Oct 12, 2021

• lstsq not working with autograd

  #66420 closed Oct 12, 2021

• NNC should rely on schema rather than on node kind in lowerings

  #64963 closed Oct 12, 2021

• Support memoryview(tensor) and buffer protocol on tensors

  #66382 closed Oct 12, 2021

• torch.mm unexpected behavior

  #66458 closed Oct 12, 2021

• Segfault in nightly binaries while creating a tensor form a list if numpy is missing

  #66353 closed Oct 12, 2021

• distribute的hang

  #66468 closed Oct 12, 2021

• `torch.nn.utils.parametrizations.spectral_norm` cannot use twice

  #66363 closed Oct 11, 2021

• GHA CI failure intermittently with "local git repository checkout dirty"

  #63613 closed Oct 11, 2021

• [feature request] Implement CI testing of installation commands

  #62590 closed Oct 11, 2021

• Investigate why github runners have multiple workers work on the same workflow

  #64217 closed Oct 11, 2021

• `pytorch_linux_xenial_py3_clang7_asan_test2` times out

  #65727 closed Oct 11, 2021

• Document CI label and what they do

  #62585 closed Oct 11, 2021

• [docathon] restructure CI documentation

  #63175 closed Oct 11, 2021

• clang-1300 hits internal error while compiling PyTorch from source on MacOS

  #65578 closed Oct 11, 2021

• Improve CI documentation

  #64192 closed Oct 11, 2021

• Document how to disable python tests on CI through issues

  #64094 closed Oct 11, 2021

• Checkout PyTorch failure: Error: fatal: not a git repository: ...third_party/NNPACK_deps/psimd

  #66269 closed Oct 11, 2021

• [ROCM]: Sometimes test_multiprocessing_context_poll_raises_exception fails in CI

  #62595 closed Oct 11, 2021

• win-vs2019-cuda10.2-py3 builds fail with non-deterministic pattern

  #65612 closed Oct 11, 2021

• No GHA workflows are scheduled for PR

  #66124 closed Oct 11, 2021

• [deploy] GraphModule cannot be made into a moveable object

  #61373 closed Oct 11, 2021

• [package/fx] make custom tracers more pluggable

  #54573 closed Oct 11, 2021

• `pinv` could be differentiable on a wider range of inputs

  #65911 closed Oct 11, 2021

• ERROR: could not find a version that satisfies the requirement torch==1.9.1+cu111(from versions 0.1.2, 0.1.2post1, 0.1.2post2 ERROR: No matching distribution found for torch==1.9.1+cu111

  #66399 closed Oct 11, 2021

• discuss.pytorch.org is down

  #66375 closed Oct 11, 2021

• Conda installs CPU ver despite cudatoolkit is provided

  #66047 closed Oct 11, 2021

• Breaking change in torch.nn.functional.gelu

  #66248 closed Oct 11, 2021

• `requires_update==False`: Partially pausing gradient updates of specific nn.Parameter s

  #51009 closed Oct 9, 2021

• Doubts about masked_scatter gradient

  #66360 closed Oct 9, 2021

• Upsample Module exist bug

  #65200 closed Oct 9, 2021

• Distributed Elastic Training Got Stuck on Rendezvous Stage With Multiple NICs

  #64926 closed Oct 9, 2021

• Precision issue in matrix-vector multiplication

  #66212 closed Oct 9, 2021

• Weird Transpose CNN behavior

  #66231 closed Oct 9, 2021

• [quant] Add support for Embedding/EmbeddingBag quantization via dynamic quant APIs

  #65185 closed Oct 8, 2021

• CUDA extension gets build error when no GPU/driver is present.

  #66333 closed Oct 8, 2021

• CosineAnnealingWarmRestarts η_max

  #66345 closed Oct 8, 2021

• [jit] Support for union types

  #41412 closed Oct 8, 2021

• How to find the source kernel code of cumsum (gpu)

  #66309 closed Oct 8, 2021

• Impossible to launch operators on an external CUDA stream

  #65822 closed Oct 8, 2021

• mobile_optimizer.optimize_for_mobile fail when backend is Metal

  #66307 closed Oct 8, 2021

• TestTensorCreationCPU.test_trilu_indices_cpu crashes in ParallellTBB mode

  #64571 closed Oct 8, 2021

• [DistributedDataParallel] Cannot modify in-place due to DDPSink

  #61982 closed Oct 8, 2021

• [onnx]RuntimeError: Failed to export an ONNX attribute 'onnx::Sub'

  #53594 closed Oct 7, 2021

• Squeeze onnx export issue opset10 vs opset11

  #53922 closed Oct 7, 2021

• Exporting 32bit ONNX models

  #54888 closed Oct 7, 2021

• [pytorch_xla_linux_bionic_py3_6_clang9_build] llvm apt repositories erroring out on fetch (File has unexpected size)

  #66268 closed Oct 7, 2021

• Please verify 1.9 ONNX release candidate on TestPyPI

  #55605 closed Oct 7, 2021

• `binary_macos_wheel_3_7_cpu_build` are broken

  #66252 closed Oct 7, 2021

• -m torch.distributed.launch --nproc_per_node=NUM_GPUS_YOU_HAVE

  #65678 closed Oct 6, 2021

• [ONNX] Invalid onnx exported: input without type information

  #25414 closed Oct 6, 2021

• Backwards compatibility is broken between 1.9.1 and nightly for `torch.cosine_similarity`

  #66086 closed Oct 6, 2021

• ONNX export with EmbeddingBag

  #25469 closed Oct 6, 2021

• Bug while exporting onnx model from greedy decoding attention ScriptModule

  #24788 closed Oct 6, 2021

• Symbolic represention of log1p

  #24808 closed Oct 6, 2021

• [ONNX] export of scripted module including nn.Dropout fails

  #24824 closed Oct 6, 2021

• Some workflows does not seem to be triggered by GHA

  #66110 closed Oct 6, 2021

• test_import_stats is flaky

  #66184 closed Oct 6, 2021

• [RFC] Support local SGD and variants for DDP communication optimization

  #59699 closed Oct 6, 2021

• Segmentation fault after removing hook within its associated hook function.

  #66037 closed Oct 6, 2021

• GPU cannot load the video frames

  #65925 closed Oct 6, 2021

• [dist] log node's hostname plus rank in the exception message

  #63174 closed Oct 6, 2021

• Add Hash for Lazy Node to PyTorch

  #65636 closed Oct 5, 2021

• ./aten/src/ATen/cpu/vec/vec256/vsx/vec256_double_vsx.h:192:12: error: cout is not a member of std

  #65956 closed Oct 5, 2021

• DISABLED jfkdjgfkldjgldg.gdg [IGNORE THIS]

  #66173 closed Oct 5, 2021

79 Issues opened by 57 people

• Identify if there are more people struggling with #66434

  #66510 opened Oct 12, 2021

• Add retries to docker sign-in and docker pull for all CI

  #66508 opened Oct 12, 2021

• Periodic jobs have been skipped on trunk for a long long time now

  #66507 opened Oct 12, 2021

• BatchNorm runtimeError: one of the variables needed for gradient computation has been modified by an inplace operation

  #66504 opened Oct 12, 2021

• DISABLED test_timer_tiny_fast_snippet (__main__.TestBenchmarkUtils)

  #66503 opened Oct 12, 2021

• ONNX Runtime backend tests on CI are disabled

  #66502 opened Oct 12, 2021

• index_put_ internal assert on CUDA but not on CPU

  #66495 opened Oct 12, 2021

• Normalize handling of scalar arguments

  #66491 opened Oct 12, 2021

• [ci] Tracking issue: GHA cancellations

  #66490 opened Oct 12, 2021

• The uniform operator param names in the C++ impl use python keywords

  #66489 opened Oct 12, 2021

• DISABLED test_ddp_comparison_uneven_inputs (__main__.TensorPipeDdpComparisonTest)

  #66488 opened Oct 12, 2021

• TestBenchmarkUtils.test_timer_tiny_fast_snippet raises divide by 0 error

  #66487 opened Oct 12, 2021

• Quantization - Linear-Bn1d fusion should support QAT

  #66483 opened Oct 12, 2021

• BAR1 memory of GPU is not released when main process is killed.

  #66482 opened Oct 12, 2021

• nn_upsample_***

  #66479 opened Oct 12, 2021

• 0INTERNAL ASSERT FAILED, We don't have an op for aten::eq but it isn't a special case

  #66476 opened Oct 12, 2021

• Exporting the operator roll to ONNX opset version 12 is not supported

  #66475 opened Oct 12, 2021

• ONNX export produce wrong result on nightly build

  #66474 opened Oct 12, 2021

• nccl comm init needs a global barrier.

  #66473 opened Oct 12, 2021

• No win cuda tests are running on PRs?

  #66471 opened Oct 12, 2021

• documentation build run in CI is different from documentation build run on master

  #66466 opened Oct 12, 2021

• Deprecating Python 3.6 support

  #66462 opened Oct 11, 2021

• Inconsistencies in LSTM outputs when processing sequence stepwise on CPU

  #66460 opened Oct 11, 2021

• Enable backward pass comms for sharded linear.

  #66446 opened Oct 11, 2021

• `torch.load()` fails under `enable_python_mode()`

  #66444 opened Oct 11, 2021

• CircleCI workflows consistently erring for some exported PRs with permission issues

  #66434 opened Oct 11, 2021

• torch/onnx/utils.py", line 501, in _model_to_graph params_dict = torch._C._jit_pass_onnx_constant_fold(graph, params_dict, RuntimeError: expected scalar type Long but found Float

  #66425 opened Oct 11, 2021

• Add Python-3.10 support

  #66424 opened Oct 11, 2021

• DISABLED test_send_recv (__main__.PythonProcessGroupTest)

  #66423 opened Oct 11, 2021

• DISABLED test_collectives (__main__.PythonProcessGroupTest)

  #66422 opened Oct 11, 2021

• CIFAR10 doesn't work on M1 MacBook

  #66410 opened Oct 11, 2021

• [docs] Source code pages should include link to GitHub file

  #66408 opened Oct 11, 2021

• IBM z14/15 simd support

  #66406 opened Oct 11, 2021

• [opinfo] Confusing interface for `ops` decorator

  #66403 opened Oct 11, 2021

• torch.isfinite incorrect output on cpu

  #66402 opened Oct 11, 2021

• test_Conv2d_naive_groups_cuda_float16 (__main__.TestNNDeviceTypeCUDA) is flaky on ROCm

  #66378 opened Oct 10, 2021

• Add constant padding_mode for grid_sample

  #66366 opened Oct 9, 2021

• Unsupported prim::Constant kind: `ival`

  #66365 opened Oct 9, 2021

• Hope to obtain authorization for Chinese translation of official tutorials.

  #66364 opened Oct 9, 2021

• question on installing GPU-enabled Pytorch

  #66359 opened Oct 9, 2021

• OpInfos disabled for batched forward grad computation

  #66357 opened Oct 9, 2021

• test_nccl_timeout should not rely on time.sleep()

  #66354 opened Oct 9, 2021

• Invalid handling of out args with type Tensor[] in native_functions.yaml

  #66341 opened Oct 8, 2021

• `torch.fx.symbolic_trace()` loses module class information

  #66335 opened Oct 8, 2021

• Should the Complexity of GPU Cumsum Be O(n)?

  #66330 opened Oct 8, 2021

• Add _reduce_scatter_base and _allgather_base to processGroupWrapper

  #66329 opened Oct 8, 2021

• Better (optional) tensor list handling in native kernels

  #66328 opened Oct 8, 2021

• Audit BinaryUfuncInfos

  #66322 opened Oct 8, 2021

• DISABLED test_create_store_multi (__main__.DistributedUtilTest)

  #66319 opened Oct 8, 2021

• Modify convolution kernels for ops

  #66305 opened Oct 8, 2021

• Android and mobile CI build on CircleCI timed out for seemingly no reason

  #66287 opened Oct 7, 2021

• Memory leak issue with pytorch_java_only 1.9.0 and libtorch 1.9.0+cpu

  #66284 opened Oct 7, 2021

• Rearrange row logic seems not to work correctly in _handle_col_wise_sharding

  #66281 opened Oct 7, 2021

• Windows build in CI failing due to mt.exe manifest issue

  #66278 opened Oct 7, 2021

• 1.10.0-rc1 fails to compile with gcc 6.4.0

  #66271 opened Oct 7, 2021

• [bootcamp task] Add forward AD support for specific ops

  #66266 opened Oct 7, 2021

• TorchBench v1.1 score regresses 2% on 20210605

  #66259 opened Oct 7, 2021

• Should the drop_last parameter of Dataloader be mutually exclusive with the batch samplers?

  #66251 opened Oct 7, 2021

• [ONNX] Still use onnx.optimizer from ONNX even though it has been moved out since ONNX 1.9

  #66250 opened Oct 7, 2021

• Latency issue with torch.sin

  #66249 opened Oct 7, 2021

• Floating point exception in mkl_vml_serv_GetMinN () on a specific computer

  #66247 opened Oct 7, 2021

• Why BatchNorm need to distinguish ndim

  #66246 opened Oct 7, 2021

• [RFC] Establishing PyTorch Test Owners

  #66232 opened Oct 6, 2021

• Clarify hook support in DDP

  #66229 opened Oct 6, 2021

• Virtualize `FloatStorage` and other `<type>Storage` classes

  #66228 opened Oct 6, 2021

• [feature request][quant] Support FakeQuant qconfigs in `test_module_init`

  #66224 opened Oct 6, 2021

• Data Loader tests hang when run in ASAN test job

  #66223 opened Oct 6, 2021

• Track models with SyncBN in DDP

  #66215 opened Oct 6, 2021

• Add LTC-TS support for ops used in SGD optimizer

  #66208 opened Oct 6, 2021

• DISABLED test_create_store_single_server (__main__.DistributedUtilTest)

  #66207 opened Oct 6, 2021

• Load LSTM layer in one GPU but other GPUS' memory usage also increase when inferencing

  #66203 opened Oct 6, 2021

• An exported onnx model can't reduce on dim with value of 0 if 'keepdims' is false

  #66200 opened Oct 6, 2021

• Move the `queue_callback()` API out of `Variable._execution_engine` and into a public API

  #66199 opened Oct 6, 2021

• `torch.fx.replace_pattern` doesn't work with untraceable wrapped functions

  #66197 opened Oct 6, 2021

• Specifying USE_VULKAN=0 in launching build_android.sh does not disable Vulkan

  #66196 opened Oct 6, 2021

• Error on unsupported key in dispatch in native_functions.yaml

  #66190 opened Oct 6, 2021

• Revise `gather()` method of `ShardedTensor`

  #66187 opened Oct 6, 2021

• Add tests to cover default args + concrete_args

  #66180 opened Oct 6, 2021

• [Meta] CI Revert Tracker

  #66178 opened Oct 6, 2021

425 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• Add torch.nn.init.uniform_ operator to ShardedTensor.

  #63997 commented on Oct 12, 2021 • 44 new comments

• Strided masked reductions: sum, amax. Testing of masked reductions.

  #65990 commented on Oct 11, 2021 • 38 new comments

• OpInfo for `*_like` functions

  #65941 commented on Oct 11, 2021 • 28 new comments

• Sparse CSR CUDA: add `addmv_out`

  #61407 commented on Oct 12, 2021 • 26 new comments

• Handle shared memory cases in MathBitFallback

  #63602 commented on Oct 12, 2021 • 23 new comments

• [PyTorch] Support additional arguments in Python record function

  #65736 commented on Oct 12, 2021 • 20 new comments

• Add a Functionalization pass in core

  #64432 commented on Oct 6, 2021 • 19 new comments

• Refactor cuDNN Convolution memory format and Conv-Bias-Relu code

  #65594 commented on Oct 9, 2021 • 17 new comments

• OpInfos for some Tensor dtype conversion methods

  #64282 commented on Oct 11, 2021 • 16 new comments

• ONNX export: propagate node metadata across passes

  #45256 commented on Oct 8, 2021 • 15 new comments

• [v.1.10.0] Release Tracker

  #65438 commented on Oct 12, 2021 • 14 new comments

• [fix] fix scalar behavior for `torch.where`

  #62084 commented on Oct 12, 2021 • 14 new comments

• [Pytorch Edge] Make Metal Ops Selective

  #65987 commented on Oct 12, 2021 • 13 new comments

• Make ShardedTensor.size() equivalent to torch.Tensor.size() (#65087)

  #66012 commented on Oct 12, 2021 • 13 new comments

• Make a python-level FunctionalTensor()

  #66101 commented on Oct 6, 2021 • 12 new comments

• [POLL][RFC] DataParallel Deprecation

  #65936 commented on Oct 11, 2021 • 11 new comments

• Add fp16/fp32 autocasting to JIT/TorchScript

  #63939 commented on Oct 12, 2021 • 11 new comments

• [sharded_tensor] simplify init_from_local_shards API

  #64481 commented on Oct 9, 2021 • 11 new comments

• [WIP] wrap XLATensors into Python XLA wrapper class

  #65841 commented on Oct 12, 2021 • 11 new comments

• [JIT] Fix type annotations of pooling modules

  #65847 commented on Oct 12, 2021 • 11 new comments

• [ONNX] Export nn.Module call as ONNX local function (#63589)

  #66140 commented on Oct 12, 2021 • 11 new comments

• torch.sparse.sum result has wrong dtype when reducing over all dimensions

  #65392 commented on Oct 8, 2021 • 10 new comments

• Remove process group barrier and all_reduce function calls from tensorpipe agent

  #65946 commented on Oct 7, 2021 • 10 new comments

• [numpy] add torch.argwhere

  #64257 commented on Oct 7, 2021 • 9 new comments

• Parameterless model still has extra inputs when exported to ONNX + insufficient checking of argument export_params

  #55144 commented on Oct 8, 2021 • 8 new comments

• GPU acceleration for Apple's M1 chip?

  #47702 commented on Oct 9, 2021 • 8 new comments

• Implement Tanh Gelu Approximation

  #61439 commented on Oct 11, 2021 • 8 new comments

• [feature request] Reinterpret tensor as different dtype

  #29013 commented on Oct 8, 2021 • 7 new comments

• Add InplaceOrView boxed kernel

  #63878 commented on Oct 12, 2021 • 7 new comments

• Remove native_functions.yaml dependency from Activation.cu

  #64499 commented on Oct 12, 2021 • 7 new comments

• Enable BatchNorm to use running mean/variance during train

  #64730 commented on Oct 7, 2021 • 6 new comments

• NVFuser Examples to Help with Integration with Lazy Tensor Core

  #65837 commented on Oct 11, 2021 • 6 new comments

• Potential strict aliasing rule violation in bitwise_binary_op (on ARM/NEON)

  #66119 commented on Oct 12, 2021 • 6 new comments

• Added Adabound.py

  #47957 commented on Oct 12, 2021 • 6 new comments

• [NNAPI] added elu and hardswish operators

  #62181 commented on Oct 8, 2021 • 6 new comments

• Are tensors on same device?

  #62653 commented on Oct 12, 2021 • 6 new comments

• fix type error in hipify_python.py

  #66164 commented on Oct 6, 2021 • 6 new comments

• Impossible to load ScriptModule on CUDA with libtorch 1.8.1 (Windows)

  #55735 commented on Oct 9, 2021 • 5 new comments

• Promote integral inputs to floating for `torch.logsumexp`

  #63393 commented on Oct 9, 2021 • 5 new comments

• Create linalg.tensordot

  #63478 commented on Oct 7, 2021 • 5 new comments

• Micro-optimisations for matmul

  #64387 commented on Oct 12, 2021 • 5 new comments

• Anomaly detection: Error detected in CudnnRnnBackward0

  #65301 commented on Oct 12, 2021 • 4 new comments

• jit trace

  #59949 commented on Oct 11, 2021 • 4 new comments

• Implement Python Array API `asarray` function.

  #60627 commented on Oct 11, 2021 • 4 new comments

• `torch.linalg.lstsq`: forward/backward AD support

  #65054 commented on Oct 12, 2021 • 4 new comments

• Add `SourceView` which doesn't own source text as base class of `Source`

  #65309 commented on Oct 6, 2021 • 4 new comments

• async compilation

  #65543 commented on Oct 7, 2021 • 4 new comments

• [TensorExpr] Add helper passes for AOT pipeline.

  #65555 commented on Oct 12, 2021 • 4 new comments

• torch.utils._pytree -> stable

  #65761 commented on Oct 6, 2021 • 3 new comments

• Set warning or error with `use_deterministic_algorithms`

  #64883 commented on Oct 6, 2021 • 3 new comments

• JIT makes model run x14 times slower in pytorch nightly

  #52286 commented on Oct 7, 2021 • 3 new comments

• [doc] Error "You've reached a dead end" when opening torch.__config__ in the docs

  #48287 commented on Oct 8, 2021 • 3 new comments

• Use old and new fft in one program

  #49695 commented on Oct 9, 2021 • 3 new comments

• Tensor with read only memory mapped numpy array causes seg fault.

  #46741 commented on Oct 11, 2021 • 3 new comments

• DataLoader num_workers > 0 causes CPU memory from parent process to be replicated in all worker processes

  #13246 commented on Oct 11, 2021 • 3 new comments

• [Feature request] Let DistributedSampler take a Sampler as input

  #23430 commented on Oct 11, 2021 • 3 new comments

• Port triangular_solve to structured kernel

  #61857 commented on Oct 12, 2021 • 3 new comments

• [PyTorch Edge][type] Add type support for NamedTuple custom class (import)

  #63130 commented on Oct 6, 2021 • 3 new comments

• Optimize PReLU (float32) and enable PReLU BFloat16 support in CPU path

  #63634 commented on Oct 12, 2021 • 3 new comments

• [codegen] teach C++ reference types how to make copies of themselves

  #64431 commented on Oct 6, 2021 • 3 new comments

• Add manual target determination

  #64950 commented on Oct 11, 2021 • 3 new comments

• Clean up `DictLiteral` and `DictComprehension` emission logic

  #64953 commented on Oct 8, 2021 • 3 new comments

• Implement histogramdd on CPU

  #65318 commented on Oct 12, 2021 • 3 new comments

• [Static Runtime] Enable manage_output_tensors for ptvsc2_predictor_bench

  #65598 commented on Oct 12, 2021 • 3 new comments

• Make empty* and *_like factory functions respect tensor subclasses

  #65677 commented on Oct 12, 2021 • 3 new comments

• [not4land][quant][fx][graphmode] lower reference linear module example

  #65723 commented on Oct 12, 2021 • 3 new comments

• Enable CUPTI for kineto by default on windows

  #65891 commented on Oct 11, 2021 • 3 new comments

• Clarify topk doc

  #65938 commented on Oct 11, 2021 • 3 new comments

• add a note on numerical accuracy

  #65947 commented on Oct 12, 2021 • 3 new comments

• [jit][edge] Export maybe-used interface methods from modules.

  #65966 commented on Oct 12, 2021 • 3 new comments

• torch.sparse.sum result has wrong dtype when reducing over all dimensions

  #66153 commented on Oct 12, 2021 • 3 new comments

• Add input node id tracking to autograd profiler along with input tensor dimensions

  #66105 commented on Oct 6, 2021 • 2 new comments

• Profiler is stuck for DDP training

  #51279 commented on Oct 6, 2021 • 2 new comments

• Deprecation: Remove nn.functional.tanh

  #65909 commented on Oct 6, 2021 • 2 new comments

• libtorch: collate_fn equivalent

  #65984 commented on Oct 7, 2021 • 2 new comments

• Batched torch.cholesky_solve is incorrect for CUDA inputs in 1.9.1

  #66072 commented on Oct 7, 2021 • 2 new comments

• Ability to explicitly close/dispose a DataLoader

  #64766 commented on Oct 8, 2021 • 2 new comments

• Couldn't export Python operator when operator_export_type set to ONNX_FALLTHROUGH

  #65853 commented on Oct 8, 2021 • 2 new comments

• [docs] Missing docs for torch.__version__ and torch.version

  #32864 commented on Oct 8, 2021 • 2 new comments

• [pytorch v1.9.0] for torch sync.pipeline in a single GPU, passing tensor layer by layer is better or using the skip-connection API is better?

  #65038 commented on Oct 9, 2021 • 2 new comments

• THE C++ FRONTEND Example Can not run correctly

  #65388 commented on Oct 9, 2021 • 2 new comments

• compiling pytorch on windows Unrecognized CMAKE_SYSTEM_NAME = Windows

  #55224 commented on Oct 10, 2021 • 2 new comments

• [doc] Document general guidelines to work with CUDA async copying and streams

  #19571 commented on Oct 11, 2021 • 2 new comments

• Inplace error of BatchNorm layer in DistributedDataParallel module

  #65907 commented on Oct 12, 2021 • 2 new comments

• Cuda Out of Memory - MMDMatic

  #66162 commented on Oct 12, 2021 • 2 new comments

• Including AdaBound in the list of Optimizers.

  #46809 commented on Oct 12, 2021 • 2 new comments

• nn.TransformerEncoder - all nan values issues when src_key_padding_mask provided

  #64525 commented on Oct 12, 2021 • 2 new comments

• Update ops.cpp

  #58807 commented on Oct 11, 2021 • 2 new comments

• Remove some unnecessary python functional wrappers

  #61608 commented on Oct 6, 2021 • 2 new comments

• [ONNX] Adding optional type in ONNX exporter

  #61938 commented on Oct 11, 2021 • 2 new comments

• Add BFloat16 support for GLU, and randperm operators on CPU

  #61944 commented on Oct 8, 2021 • 2 new comments

• Enable BFloat16 ELU, SELU and CELU in CPU path

  #62546 commented on Oct 12, 2021 • 2 new comments

• Add BFloat16 support for smooth_l1_loss on CPU

  #62558 commented on Oct 8, 2021 • 2 new comments

• Add example of FiniteTensor with __torch_dispatch__

  #62819 commented on Oct 8, 2021 • 2 new comments

• Add softplus support to autodiff

  #63942 commented on Oct 12, 2021 • 2 new comments

• Add tensor.{adjoint(),H,mT,mH} methods and properties

  #64179 commented on Oct 11, 2021 • 2 new comments

• [quant] Member variable access in quantizable LSTM

  #64833 commented on Oct 6, 2021 • 2 new comments

• Logic to categorize all the commits

  #64929 commented on Oct 11, 2021 • 2 new comments

• Clean up `ListLiteral` and `ListComprehension` emission logic

  #64952 commented on Oct 8, 2021 • 2 new comments

• Matrix multiplication for CUDA batched and non-batched int32/int64 tensors.

  #65133 commented on Oct 8, 2021 • 2 new comments

• Use a Python wrapper for C++ tests

  #65514 commented on Oct 12, 2021 • 2 new comments

• Automated submodule update: FBGEMM

  #65595 commented on Oct 12, 2021 • 2 new comments

• Add error reporting when debugging on Windows

  #65695 commented on Oct 8, 2021 • 2 new comments

• [sharded_tensor] enable gloo/mpi backend to be used in sharded tensor tests

  #65855 commented on Oct 12, 2021 • 2 new comments

• Porting index_add to structured kernels, add an out variant

  #65993 commented on Oct 7, 2021 • 2 new comments

• initial implementation of PartitionedDataset

  #65995 commented on Oct 11, 2021 • 2 new comments

• Compute input gradient only if required (CPU)

  #66069 commented on Oct 6, 2021 • 2 new comments

• [ONNX] Update sum symbolic to handle dtypes (#64289)

  #66141 commented on Oct 12, 2021 • 2 new comments

• [ONNX] Add warning for inplace updates on tensor.shape in tracing mode (#63170)

  #66142 commented on Oct 12, 2021 • 2 new comments

• ONNX: Delete or document skipped ORT tests (#64470)

  #66143 commented on Oct 12, 2021 • 2 new comments

• [ONNX] Add diagonal symbolic (#64454)

  #66144 commented on Oct 12, 2021 • 2 new comments

• [ONNX] Deprecate fold_if pass (#65697)

  #66145 commented on Oct 12, 2021 • 2 new comments

• [ONNX] Modify softplus symbolic to support beta!=1 (#65001)

  #66146 commented on Oct 12, 2021 • 2 new comments

• [ONNX] Symbolic: dynamic input for OneHot, bool for Einsum (#65940)

  #66147 commented on Oct 12, 2021 • 2 new comments

• [ONNX] Support exporting with Apex O2 (#65374)

  #66148 commented on Oct 12, 2021 • 2 new comments

• zip file serialization endianness detection is broken

  #65300 commented on Oct 5, 2021 • 1 new comment

• How to use `MASTER_ADDR` in a distributed training script

  #65992 commented on Oct 6, 2021 • 1 new comment

• `__torch_dispatch__` can result in returning `None` for an op that should return Tensors.

  #66026 commented on Oct 6, 2021 • 1 new comment

• Deprecation: Remove nn.functional.sigmoid

  #65910 commented on Oct 6, 2021 • 1 new comment

• SummaryWriter `add_hparams` should support adding new hyperparameters

  #39250 commented on Oct 6, 2021 • 1 new comment

• [RFC] Low-level speed optimizations for PowerSGD

  #65813 commented on Oct 6, 2021 • 1 new comment

• Batch matmul with sparse matrix, dense vector

  #14489 commented on Oct 6, 2021 • 1 new comment

• F.conv2d() causes RuntimeError: cuDNN error: CUDNN_STATUS_INTERNAL_ERROR

  #45769 commented on Oct 6, 2021 • 1 new comment

• Complex backward returns NaN values

  #46155 commented on Oct 6, 2021 • 1 new comment

• Support matrix operations between complex and real tensors

  #61909 commented on Oct 6, 2021 • 1 new comment

• ONNX export of MaxUnpool2d is not supported

  #25088 commented on Oct 6, 2021 • 1 new comment

• PicklingError when saving a ddp module with torch.save()

  #66046 commented on Oct 7, 2021 • 1 new comment

• [docs] Disable google indexing of old docs and of master docs and of function/module summary pages

  #3428 commented on Oct 7, 2021 • 1 new comment

• [ONNX] export torch.zeros() inside torch.jit.script

  #25501 commented on Oct 7, 2021 • 1 new comment

• [docs] Strange torch.unique function signature

  #58139 commented on Oct 7, 2021 • 1 new comment

• Support complex type in ONNX export

  #59246 commented on Oct 7, 2021 • 1 new comment

• ONNX export of torch.nn.functional.linear with weight dimension 1

  #64978 commented on Oct 7, 2021 • 1 new comment

• NNC Lowering bugs found by new OpInfo tests

  #58772 commented on Oct 7, 2021 • 1 new comment

• Exporting the operator roll to ONNX opset version 12 is not supported. Please open a bug to request ONNX export support for the missing operator.

  #56355 commented on Oct 7, 2021 • 1 new comment

• Import ONNX model to Pytorch

  #21683 commented on Oct 8, 2021 • 1 new comment

• RuntimeError: Exporting the operator quantize_per_tensor to ONNX opset version 10 is not supported.

  #47217 commented on Oct 8, 2021 • 1 new comment

• PyPy support

  #17835 commented on Oct 8, 2021 • 1 new comment

• Drop _stacklevel from argspecs of F.softmax, F.softmin, F.log_softmax (for implicit dim has been long deprecated)

  #36524 commented on Oct 8, 2021 • 1 new comment

• Unification of model initialization methods / naming across domain libraries + support of skip_init

  #65153 commented on Oct 8, 2021 • 1 new comment

• Immediate mode API (with functional flavor) for optimizers

  #38471 commented on Oct 9, 2021 • 1 new comment

• a problem happened in torch.randperm

  #63726 commented on Oct 9, 2021 • 1 new comment

• [feature request] BatchNorm frozen mode in core

  #66073 commented on Oct 9, 2021 • 1 new comment

• Feature Request: torch.{fft, rfft} to torch.fft.{fft, rfft} porting guide

  #49637 commented on Oct 9, 2021 • 1 new comment

• Improve this error message: RuntimeError: Expected from <= to to be true, but got false. (Could this error message be improved? If so, please report an enhancement request to PyTorch.)

  #65980 commented on Oct 9, 2021 • 1 new comment

• RuntimeError: Exporting the operator atan2 to ONNX opset version 9 is not supported. Please open a bug to request ONNX export support for the missing operator.

  #51334 commented on Oct 10, 2021 • 1 new comment

• torch.unique signature is not descriptive

  #65949 commented on Oct 11, 2021 • 1 new comment

• Error with build PyTorch iOS libraries from source

  #61348 commented on Oct 11, 2021 • 1 new comment

• NCCL 2.7.8 errors on PyTorch distributed process group creation

  #47885 commented on Oct 11, 2021 • 1 new comment

• Broadcasting for torch.cross

  #39656 commented on Oct 11, 2021 • 1 new comment

• forward-mode AD: Check to see that generated code for output_differentiability: [false] is fine

  #61926 commented on Oct 11, 2021 • 1 new comment

• Exception thrown in final autograd callback (queue_callback) not caught if not on CPU thread: terminate called after throwing an instance of 'python_error'

  #65366 commented on Oct 11, 2021 • 1 new comment

• Per-overload torch.ops API

  #65997 commented on Oct 11, 2021 • 1 new comment

• [LTC] Code-gen IRs and TorchScript lowerings needed by running full TorchBench

  #65576 commented on Oct 11, 2021 • 1 new comment

• [Poll] Support DistributedDataParallel (DDP) in PyTorch C++ API (libtorch)

  #48959 commented on Oct 12, 2021 • 1 new comment

• [FR] Support for non-Parameter, non-Buffer objects in state_dict

  #62094 commented on Oct 12, 2021 • 1 new comment

• Support memoryview() method on torch.Tensor

  #19143 commented on Oct 12, 2021 • 1 new comment

• [feature request] Sum (torch.add / torch.logaddexp) of several tensors

  #27522 commented on Oct 12, 2021 • 1 new comment

• Fast Cosine Transform

  #51965 commented on Oct 12, 2021 • 1 new comment

• Inefficient conversion between COO and CSR formats

  #56959 commented on Oct 12, 2021 • 1 new comment

• Different behavior when trace model.

  #19349 commented on Oct 12, 2021 • 1 new comment

• Non-persistent modules

  #65774 commented on Oct 12, 2021 • 1 new comment

• CUDA ROCm grad tests sometimes leak memory

  #65797 commented on Oct 12, 2021 • 1 new comment

• Delete TensorImpl::GetDevice()

  #17668 commented on Oct 8, 2021 • 1 new comment

• Add sphinx class tag

  #27502 commented on Oct 11, 2021 • 1 new comment

• Add sphinx class tag

  #27503 commented on Oct 11, 2021 • 1 new comment

• Update sphinx writer to restore old documentation references

  #41555 commented on Oct 11, 2021 • 1 new comment

• [cuDNN v8 API] cuDNN benchmark, convolution bwd / transposed convolution fwd, `bfloat16`, conv-bias-activation fusion

  #60755 commented on Oct 6, 2021 • 1 new comment

• [PyTorch Edge][type] Add type support for NamedTuple custom class (export)

  #62612 commented on Oct 6, 2021 • 1 new comment

• fake LTS 1.8 hotfix (DO NOT MERGE)

  #62837 commented on Oct 7, 2021 • 1 new comment

• Add BFloat16 support for grid_sample on CPU

  #62875 commented on Oct 12, 2021 • 1 new comment

• run rpc only

  #63019 commented on Oct 10, 2021 • 1 new comment

• [WIP] Enable module profiling in backwards pass

  #63072 commented on Oct 10, 2021 • 1 new comment

• Add BFloat16 support for logsigmoid, hardsigmoid, hardshrink, softshrink, hardswish and softplus on CPU

  #63134 commented on Oct 8, 2021 • 1 new comment

• [tensorexpr] Disable inlining for unhandled cases to avoid runtime error

  #63167 commented on Oct 12, 2021 • 1 new comment

• Array API: Add torch.linalg.cross

  #63285 commented on Oct 8, 2021 • 1 new comment

• `torch.linalg.householder_product` faster backward

  #63880 commented on Oct 6, 2021 • 1 new comment

• Add native_dropout

  #63937 commented on Oct 11, 2021 • 1 new comment

• Deprecate x.T on tensors of dimension other than 0 or 2

  #64180 commented on Oct 11, 2021 • 1 new comment

• Prefer mT and mH over transpose(-2, -1) and transpose(-2, -1).conj()

  #64181 commented on Oct 11, 2021 • 1 new comment

• Fixed custom op sizes not appearing in profiler trace file or execution graph.

  #64325 commented on Oct 12, 2021 • 1 new comment

• [TEST ONLY] Replacing codegen with boxed for testing

  #64654 commented on Oct 12, 2021 • 1 new comment

• [aot_compiler] Add passes, fix long dim

  #64755 commented on Oct 8, 2021 • 1 new comment

• [quant] Correct name for quantized LSTM

  #64834 commented on Oct 6, 2021 • 1 new comment

• [jit] Add MutableTypePtrHelper::mapTypeToBorrowedAliasTypeSet

  #65344 commented on Oct 9, 2021 • 1 new comment

• [WIP] add cudaMallocAsync as an alternative backend for the CUDA allocator

  #65365 commented on Oct 11, 2021 • 1 new comment

• Clean up test running scripts

  #65508 commented on Oct 11, 2021 • 1 new comment

• Initialize operator versioning tracking files

  #65518 commented on Oct 7, 2021 • 1 new comment

• Fix attr_to_scope's key in `torch/utils/tensorboard/_pytorch_graph.py`

  #65692 commented on Oct 6, 2021 • 1 new comment

• [nn] draft : module_info cpp api parity infra

  #65807 commented on Oct 12, 2021 • 1 new comment

• Add OpInfo for torch.bucketize

  #65821 commented on Oct 12, 2021 • 1 new comment

• [PyTorch Edge] Optimize Dequantize Tensor with Intrinsics

  #65844 commented on Oct 11, 2021 • 1 new comment

• [PyTorch Edge] Parallelize Quantize and Dequantize Tensor

  #65845 commented on Oct 11, 2021 • 1 new comment

• ufunc codegen

  #65851 commented on Oct 7, 2021 • 1 new comment

• Refactor Node class to move Shape class usage to new derived TsNode

  #65869 commented on Oct 12, 2021 • 1 new comment

• Refactoring named tensor handling

  #65912 commented on Oct 6, 2021 • 1 new comment

• Add memory_format support for torch.zeros

  #65930 commented on Oct 12, 2021 • 1 new comment

• [Rocm]: Reduce severity of detected possible memory leak from assertion to warning

  #65973 commented on Oct 7, 2021 • 1 new comment

• Add output_mask

  #66068 commented on Oct 6, 2021 • 1 new comment

• Compute input gradient only if required (CUDA)

  #66070 commented on Oct 6, 2021 • 1 new comment

• Add no-input-grad-needed cases to test_grid_sample

  #66071 commented on Oct 6, 2021 • 1 new comment

• Fix gcc11 build error in ir_emitter.cpp

  #66089 commented on Oct 12, 2021 • 1 new comment

• [JIT] Freeze allows preservation of submodule attributes

  #66102 commented on Oct 11, 2021 • 1 new comment

• Mild refactor of native_functions.yaml dispatch parsing

  #66109 commented on Oct 7, 2021 • 1 new comment

• [LTC] Migrate the CI from linux-xenial-cuda11.1 to linux-xenial-cuda11.3

  #66160 commented on Oct 12, 2021 • 1 new comment

• Support scatter and gather for ProcessGroupNCCL

  #65571 commented on Oct 6, 2021 • 0 new comments

• Function request: np.around (alias with torch.round)

  #65908 commented on Oct 6, 2021 • 0 new comments

• Beta distribution yields wrong results for a,b >1 and x=0 or x=1

  #65811 commented on Oct 6, 2021 • 0 new comments

• [RFC] Model Sharding for distributed training

  #55207 commented on Oct 6, 2021 • 0 new comments

• [feature request] Allow broadcasting for F.huber_loss and potentially other losses

  #66083 commented on Oct 6, 2021 • 0 new comments

• Alternate Model Loading - Android

  #63305 commented on Oct 6, 2021 • 0 new comments

• I am unable to detect torch.Tensor._version change during __torch_dispatch__ call

  #64874 commented on Oct 6, 2021 • 0 new comments

• Complex Numbers Support

  #33152 commented on Oct 6, 2021 • 0 new comments

• ONNX build uses system protoc headers instead of bundled one

  #65614 commented on Oct 6, 2021 • 0 new comments

• Export torch.cat to ONNX with Dynamic shape does not work on GPU

  #24899 commented on Oct 6, 2021 • 0 new comments

• Warnings during backward execution may not be turned into Python warnings, leading to console spam

  #50209 commented on Oct 6, 2021 • 0 new comments

• Label tracking meta-issue (edit me to get automatically CC'ed on issues! cc bot)

  #24422 commented on Oct 7, 2021 • 0 new comments

• torch.distributed server back-connects to clients on random ports leading to firewall problems

  #44544 commented on Oct 7, 2021 • 0 new comments

• [quant] Support QAT for Embedding/EmbeddingBag

  #61865 commented on Oct 7, 2021 • 0 new comments

• Add support for HF NLP models in torch.package/deploy

  #65305 commented on Oct 8, 2021 • 0 new comments

• Improve F.interpolate error message

  #63845 commented on Oct 11, 2021 • 0 new comments

• [JIT] Fix batchnorm / instance norm alias annotations

  #65563 commented on Oct 11, 2021 • 0 new comments

• API torch.ops.image.read_file reports RuntimeError - No such operator image::read_file

  #65473 commented on Oct 11, 2021 • 0 new comments

• Intellisense on Windows doesn't report real error message by default

  #47900 commented on Oct 11, 2021 • 0 new comments

• [RFC] Extend Autocast to CPU/CUDA with BF16 data type

  #55374 commented on Oct 11, 2021 • 0 new comments

• logsigmoid fails on -inf

  #20972 commented on Oct 11, 2021 • 0 new comments

• Source location range issue for dictionary

  #55577 commented on Oct 11, 2021 • 0 new comments

• [rfc] optimize distributed and RPC tests in CI

  #60344 commented on Oct 11, 2021 • 0 new comments

• Execute C++ based GTest unit tests from a python wrapper

  #64067 commented on Oct 11, 2021 • 0 new comments

• Obscure error when source dir is not writeable

  #65920 commented on Oct 11, 2021 • 0 new comments

• `test_load_commit_from_forked_repo` started to fail with "Rate Limit Exceeded"

  #65252 commented on Oct 11, 2021 • 0 new comments

• [RFC] Plan to deduplicate test_c10d and distributed_test

  #55340 commented on Oct 11, 2021 • 0 new comments

• Cleanup optimizer handling of complex numbers.

  #65711 commented on Oct 12, 2021 • 0 new comments

• Comprehensively test NCCL's `get_future()` API

  #56838 commented on Oct 12, 2021 • 0 new comments

• Remove single device constraint from ProcessGroupNCCL profiling

  #56836 commented on Oct 12, 2021 • 0 new comments

• TestZeroRedundancyOptimizerDistributed.test_multiple_groups is failing intermittently

  #66059 commented on Oct 12, 2021 • 0 new comments

• TensorPipeDistAutogradTest is frequently failing

  #64518 commented on Oct 12, 2021 • 0 new comments

• Store created by dist.new_group doesn't appear to respect timeout

  #54524 commented on Oct 12, 2021 • 0 new comments

• PyTorch CI Stability Action Tracker

  #65439 commented on Oct 12, 2021 • 0 new comments

• [DO NOT DELETE OR MERGE] Nightly release

  #26921 commented on Oct 12, 2021 • 0 new comments

• [DO NOT DELETE OR MERGE] Postnightly release

  #27167 commented on Oct 12, 2021 • 0 new comments

• [WIP] Programmatically disambiguate when to step a `_LRScheduler`

  #37770 commented on Oct 12, 2021 • 0 new comments

• Named tensor serialization support

  #37982 commented on Oct 12, 2021 • 0 new comments

• Choose test affinity based on current affinity

  #44369 commented on Oct 12, 2021 • 0 new comments

• [JIT] Alias Doc Improvements

  #50969 commented on Oct 12, 2021 • 0 new comments

• A ~17x speed optimization of creating tensor from multiple numpy arrays using torch.tensor and torch.Tensor.

  #51731 commented on Oct 12, 2021 • 0 new comments

• [JIT]Add manual argument index support for string format

  #52796 commented on Oct 11, 2021 • 0 new comments

• [WIP][vulkan] FRCNN ops

  #53083 commented on Oct 12, 2021 • 0 new comments

• [caffe2] Make SparseLabelSplit and SparseLabelSplitGradient available

  #53415 commented on Oct 12, 2021 • 0 new comments

• Support mix of int32 and int64 offsets/indices for EmbeddingBag and its variants

  #53655 commented on Oct 12, 2021 • 0 new comments

• Initial version of parallelization framework

  #53915 commented on Oct 12, 2021 • 0 new comments

• [ONNX] Name onnx value name to satisfy C identifier

  #54130 commented on Oct 12, 2021 • 0 new comments

• Implement CPU Array Interface support and testing tensor creations using OpInfo framework

  #54187 commented on Oct 12, 2021 • 0 new comments

• Reuse cosntant table from jit and support exporting old models.

  #54306 commented on Oct 12, 2021 • 0 new comments

• implement cdf for gamma distribution

  #54496 commented on Oct 12, 2021 • 0 new comments

• Update persons_of_interest.rst

  #54567 commented on Oct 12, 2021 • 0 new comments

• Make hipified paths relative to build directory because setuptools doesn't allow absolute paths

  #54801 commented on Oct 12, 2021 • 0 new comments

• [testing] clamp: test with either min or max

  #54939 commented on Oct 12, 2021 • 0 new comments

• Added verbosity to C++ learning rate schedulers

  #54946 commented on Oct 12, 2021 • 0 new comments

• Add dependency of libnccl_static.a when do libtorch static build

  #54950 commented on Oct 12, 2021 • 0 new comments

• [please ignore]

  #54957 commented on Oct 12, 2021 • 0 new comments

• [torch] make StreamContext nop (and doesn't call cuda lazy init) when stream is None

  #54963 commented on Oct 12, 2021 • 0 new comments

• cmake: allow to build pytorch as a CMake subproject

  #54978 commented on Oct 12, 2021 • 0 new comments

• [Don't Review][TestOnly] Debug Remote CUDA RPC

  #54993 commented on Oct 12, 2021 • 0 new comments

• [caffe2] Implement gradient for Log1p

  #55018 commented on Oct 12, 2021 • 0 new comments

• Streamcontext nop fix

  #55055 commented on Oct 12, 2021 • 0 new comments

• Make DispatchGlobal consts

  #55068 commented on Oct 12, 2021 • 0 new comments

• [WIP] ugly fix to the bias type mismatch in linear

  #55089 commented on Oct 12, 2021 • 0 new comments

• [ci experiment] AlwaysCall Backend Fallback

  #55101 commented on Oct 12, 2021 • 0 new comments

• [WIP] Add torch.package shell

  #55110 commented on Oct 12, 2021 • 0 new comments

• [android] fbjni from prefab dependency 0.2.2

  #55118 commented on Oct 12, 2021 • 0 new comments

• [Static Runtime] Add TORCH_API to functions in static/passes.h

  #55127 commented on Oct 12, 2021 • 0 new comments

• Add some missing types to training_toolkit/torch, make a test name consistent

  #55186 commented on Oct 12, 2021 • 0 new comments

• introduce gain arg for kaiming initialisation

  #55216 commented on Oct 12, 2021 • 0 new comments

• Add some missing HIP error types

  #55233 commented on Oct 12, 2021 • 0 new comments

• [NOT_FOR_LAND][CI-all][android] fbjni from prefab dependency 0.2.2

  #55235 commented on Oct 12, 2021 • 0 new comments

• [Pytorch Edge] Remove run_on_bundled_input

  #55313 commented on Oct 12, 2021 • 0 new comments

• [1/n] Adding Engine, Event, EventHandler for Composable DDP

  #55410 commented on Oct 12, 2021 • 0 new comments

• [WIP][ONNX] Add support for max_unpool op in export

  #55411 commented on Oct 6, 2021 • 0 new comments

• [WIP][ONNX] Update lower tuple pass

  #55473 commented on Oct 6, 2021 • 0 new comments

• [FX][hackathon][not for submit yet] Can we depend on torchtext?

  #55478 commented on Oct 12, 2021 • 0 new comments

• [FX][hackathon][not for submit yet] Can we depend on torchaudio?

  #55480 commented on Oct 12, 2021 • 0 new comments

• [FX][hackathon][not for submit yet] Can we depend on detectron2?

  #55484 commented on Oct 12, 2021 • 0 new comments

• [Hackathon] Add source range highlighting check to test_slice

  #55487 commented on Oct 12, 2021 • 0 new comments

• [FX][hackathon] Initial tests for symbolically tracing all modules for FX

  #55494 commented on Oct 12, 2021 • 0 new comments

• [test][no land]Test PR to reproduce AMD compiler issue for gpu_kernel_multiple_outputs

  #55559 commented on Oct 12, 2021 • 0 new comments

• Bring AVX2 support for float16 at par with that for BFloat16

  #55586 commented on Oct 12, 2021 • 0 new comments

• [WIP][package] Factor out dependency resolution

  #55625 commented on Oct 12, 2021 • 0 new comments

• [WIP] enable batchnorm in autodiff

  #55627 commented on Oct 12, 2021 • 0 new comments

• [vulkan] Add SPIR-V Optimizer to build process

  #55634 commented on Oct 12, 2021 • 0 new comments

• Add acctype in NLL Loss

  #55674 commented on Oct 12, 2021 • 0 new comments

• test_autograd test memory leak

  #55688 commented on Oct 12, 2021 • 0 new comments

• Improve list of nondeterministic ops

  #55708 commented on Oct 12, 2021 • 0 new comments

• [WIP] Migrate _fft_c2c_mkl to structured kernel

  #55730 commented on Oct 12, 2021 • 0 new comments

• [Pytorch] Build lite interpreter as default for iOS

  #55732 commented on Oct 12, 2021 • 0 new comments

• [quant][refactor] Dedup the batch norm code

  #55758 commented on Oct 12, 2021 • 0 new comments

• Support new style in add_scalars

  #55789 commented on Oct 12, 2021 • 0 new comments

• [WIP][ONNX] Extend support for args/kwargs in optional inputs API

  #55792 commented on Oct 6, 2021 • 0 new comments

• Inline expressions into Backward graphs [WIP]

  #55846 commented on Oct 12, 2021 • 0 new comments

• Specify __all__ in cuda/__init__.py

  #55872 commented on Oct 12, 2021 • 0 new comments

• Specify __all__ in jit/__init__.py

  #55873 commented on Oct 12, 2021 • 0 new comments

• Fix passing of jit_opt_level

  #55897 commented on Oct 12, 2021 • 0 new comments

• Deduplicate codegenOutputQuery to query maximum CUDA compute capabilities

  #55901 commented on Oct 12, 2021 • 0 new comments

• Handle JIT test failure when the GPU is newer than the CUDA compiler

  #55904 commented on Oct 12, 2021 • 0 new comments

• [wip] Switch to C++17

  #55938 commented on Oct 12, 2021 • 0 new comments

• [PyTorch] Remove debug_pkl when exporting bytecode

  #55943 commented on Oct 12, 2021 • 0 new comments

• Annotate functions in _jit_internal.py

  #55974 commented on Oct 12, 2021 • 0 new comments

• [vulkan] Fix Vulkan build with C++14

  #55987 commented on Oct 12, 2021 • 0 new comments

• Allow passing cpu to CUDA RPC device maps

  #56043 commented on Oct 12, 2021 • 0 new comments

• fix jit module freeze eval check

  #56341 commented on Oct 11, 2021 • 0 new comments

• fixed numerical instability in beta_grad_alpha_mid

  #57969 commented on Oct 12, 2021 • 0 new comments

• warn appropriate warning for fn in distributed api

  #58046 commented on Oct 11, 2021 • 0 new comments

• [Pytorch] Add non mutator bundled inputs method (#58342)

  #58404 commented on Oct 12, 2021 • 0 new comments

• [TEST] [DO NOT MERGE] Testing Ed's PR: Export d28458898

  #58419 commented on Oct 12, 2021 • 0 new comments

• Fix usage of TORCH_INTERNAL_ASSERT with message

  #58478 commented on Oct 11, 2021 • 0 new comments

• Better handling of OpInfo.sample_inputs

  #58481 commented on Oct 12, 2021 • 0 new comments

• Added statistic related to out variant nodes

  #58686 commented on Oct 11, 2021 • 0 new comments

• [torch][package] Make load_library a no-op inside a package

  #58901 commented on Oct 11, 2021 • 0 new comments

• Document torch.quantile interpolation kwarg

  #59397 commented on Oct 8, 2021 • 0 new comments

• Single agent c10d backend tests.

  #59453 commented on Oct 11, 2021 • 0 new comments

• Skeleton code for filestore rendezvous backend.

  #59601 commented on Oct 11, 2021 • 0 new comments

• Remove composite definition for new_empty_strided

  #59708 commented on Oct 12, 2021 • 0 new comments

• lr_sche_update_OneCycle

  #59930 commented on Oct 12, 2021 • 0 new comments

• Implement a generic Distribution.expand() and use it in a few places

  #59944 commented on Oct 12, 2021 • 0 new comments

• Improve TorchScript printout format to be as close to eager as possible

  #60033 commented on Oct 11, 2021 • 0 new comments

• replace static_cast by dynamic_cast, found by clang-tidy

  #60317 commented on Oct 12, 2021 • 0 new comments

• Rename dataset to data_source in data.distributed

  #60967 commented on Oct 11, 2021 • 0 new comments

• [JIT] Add an api for jit logging output stream selection

  #61178 commented on Oct 11, 2021 • 0 new comments

• Added ScriptFunction pkl exception

  #61381 commented on Oct 11, 2021 • 0 new comments

• Sparse CSR CPU: add `addmv_out`

  #61536 commented on Oct 12, 2021 • 0 new comments

• [distributed] add flock-based wrapper around built-in print

  #61803 commented on Oct 11, 2021 • 0 new comments

• Sparse CSR CUDA: add `triangular_solve_out`

  #61858 commented on Oct 12, 2021 • 0 new comments

• start iwyu project

  #62010 commented on Oct 12, 2021 • 0 new comments

• Implements allreduce_coalesced for ProcessGroupNCCL

  #62140 commented on Oct 11, 2021 • 0 new comments

• Sparse CSR CPU: add `triangular_solve_out`

  #62180 commented on Oct 12, 2021 • 0 new comments

• Make ProcessGroupGloo use CUDA Future

  #62302 commented on Oct 12, 2021 • 0 new comments

• Fix rsub with alpha != 1

  #62397 commented on Oct 12, 2021 • 0 new comments

• [Draft][ONNX] STFT and ISTFT

  #62412 commented on Oct 6, 2021 • 0 new comments

• [jit] Add TORCH_API to [Scalar,Vector]AttributeValue

  #62435 commented on Oct 11, 2021 • 0 new comments

• Refactor RecordFunction to use CustomClass

  #62710 commented on Oct 11, 2021 • 0 new comments

• Create linalg.matrix_exp

  #62715 commented on Oct 8, 2021 • 0 new comments

• Separating quantization test from distributed_test

  #63051 commented on Oct 11, 2021 • 0 new comments

• apy_test: Add check before using rmtree

  #63060 commented on Oct 11, 2021 • 0 new comments

• sync c10d NCCL version parsing with formatting

  #63155 commented on Oct 11, 2021 • 0 new comments

• Test clang build on Windows

  #63288 commented on Oct 12, 2021 • 0 new comments

• Enable Intel® VTune™ Profiler's Instrumentation and Tracing Technology APIs (ITT) to PyTorch

  #63289 commented on Oct 12, 2021 • 0 new comments

• DRAFT: shareded_tensor.ones

  #63358 commented on Oct 12, 2021 • 0 new comments

• Sparse CSR: Add torch.empty

  #63509 commented on Oct 12, 2021 • 0 new comments

• Sparse CSR: Add tensor.resize_ and tensor.copy_

  #63510 commented on Oct 12, 2021 • 0 new comments

• Sparse CSR CUDA: add `torch.addmm` with all inputs sparse

  #63511 commented on Oct 12, 2021 • 0 new comments

• [nvFuser] [WIP] Normalization support

  #63542 commented on Oct 6, 2021 • 0 new comments

• OpInfo for `nn.functional.cross_entropy`

  #63547 commented on Oct 6, 2021 • 0 new comments

• [WIP] Ops in Python

  #63643 commented on Oct 12, 2021 • 0 new comments

• Add xpu legacy profiler for Intel(R) XPU devices

  #63746 commented on Oct 11, 2021 • 0 new comments

• [testing][wip] Use fastcov instead of lcov

  #63908 commented on Oct 12, 2021 • 0 new comments

• specializeGradSumToSize patch - propagate profile_none through profile_ivalue

  #63941 commented on Oct 11, 2021 • 0 new comments

• Sparse CSR CUDA: add `torch.add` with all inputs sparse

  #63948 commented on Oct 12, 2021 • 0 new comments

• [PyTorch] RFC: Add tuple inline storage

  #64066 commented on Oct 12, 2021 • 0 new comments

• [PyTorch Edge][tracing-based] Unify tracer between internal and external

  #64152 commented on Oct 11, 2021 • 0 new comments

• Incubation of Graphical Representation in TF2 of HParams

  #64262 commented on Oct 12, 2021 • 0 new comments

• Sparse CSR CPU: add `torch.add` with all inputs sparse

  #64391 commented on Oct 12, 2021 • 0 new comments

• [PyTorch Edge][tracing-based] tracing-based build libtorch_lite

  #64406 commented on Oct 6, 2021 • 0 new comments

• Added nearest-exact interpolation mode

  #64501 commented on Oct 11, 2021 • 0 new comments

• Convert profiler schedule to a functor for serialization

  #64583 commented on Oct 11, 2021 • 0 new comments

• enable convolution relu fusion path for cpu path

  #64639 commented on Oct 11, 2021 • 0 new comments

• [WIP] working POC of quantization via dynamic tracing

  #64676 commented on Oct 8, 2021 • 0 new comments

• Port `min` and `max` full reductions to structured kernels

  #64735 commented on Oct 9, 2021 • 0 new comments

• WIP-NNC

  #64855 commented on Oct 7, 2021 • 0 new comments

• Back out "[pytorch] deprecate default_op_deps.yaml"

  #64884 commented on Oct 12, 2021 • 0 new comments

• Disable target determination

  #64925 commented on Oct 12, 2021 • 0 new comments

• [nn] MarginRankingLoss : no batch dim

  #64975 commented on Oct 11, 2021 • 0 new comments

• [JIT][Easy] Shape cleanups

  #65148 commented on Oct 8, 2021 • 0 new comments

• [jit] Implement one-element MemoryDAG::mayContainAlias more efficiently

  #65178 commented on Oct 9, 2021 • 0 new comments

• [JIT] Add partial evaluation graph stitching logic

  #65377 commented on Oct 8, 2021 • 0 new comments

• [PyTorch] Adopt faster Tuple::create

  #65381 commented on Oct 12, 2021 • 0 new comments

• Implementing a coalesced allgather for all backends in Python

  #65403 commented on Oct 12, 2021 • 0 new comments

• [PyTorch][Static Runtime] Don't use vector in ProcessedNode

  #65429 commented on Oct 12, 2021 • 0 new comments

• [v1.10.0] Getting signal for release/1.10

  #65434 commented on Oct 8, 2021 • 0 new comments

• [do not merge] Test docs determination

  #65440 commented on Oct 11, 2021 • 0 new comments

• get aten memory ops in the graph

  #65445 commented on Oct 12, 2021 • 0 new comments

• replace boolean_dispatch with Union for `(adaptive|fractional)?max_pool[1-3]d`

  #65456 commented on Oct 12, 2021 • 0 new comments

• add `OpInfo` for `torch.nn.functional.pairwise_distance`

  #65460 commented on Oct 7, 2021 • 0 new comments

• add `OpInfo` for `torch.nn.pixel_shuffle`

  #65467 commented on Oct 7, 2021 • 0 new comments

• add `OpInfo` for `torch.nn.pixel_unshuffle`

  #65468 commented on Oct 7, 2021 • 0 new comments

• add `OpInfo` for `torch.nn.functional.kl_div`

  #65469 commented on Oct 7, 2021 • 0 new comments

• TCPStore timeout condition and extra logging

  #65478 commented on Oct 12, 2021 • 0 new comments

• Small updates to RELEASE.md

  #65489 commented on Oct 11, 2021 • 0 new comments

• [PyTorch] Add IValue::toTupleRef()

  #65504 commented on Oct 12, 2021 • 0 new comments

• [PyTorch] Adopt IValue::toTupleRef() where obvious

  #65505 commented on Oct 12, 2021 • 0 new comments

• Fix bug preventing optimization from firing

  #65573 commented on Oct 8, 2021 • 0 new comments

• Add x + 0 optimization

  #65574 commented on Oct 8, 2021 • 0 new comments

• Add Handling of Cat in Shape Analysis

  #65575 commented on Oct 8, 2021 • 0 new comments

• [wip] Add more C++ tests to Python wrapper

  #65581 commented on Oct 11, 2021 • 0 new comments

• ci: Move windows to build bdist_wheel

  #65587 commented on Oct 12, 2021 • 0 new comments

• Sparse CSR CPU: add `torch.addmm`

  #65606 commented on Oct 12, 2021 • 0 new comments

• add `OpInfo` for `torch.diagflat`

  #65680 commented on Oct 7, 2021 • 0 new comments

• fix `torch.nn.functional.l1_loss`

  #65681 commented on Oct 7, 2021 • 0 new comments

• [nn] MultiLabelSoftMarginLoss : no batch dim support

  #65690 commented on Oct 11, 2021 • 0 new comments

• add a quantized hardsigmoid inplace variant

  #65740 commented on Oct 8, 2021 • 0 new comments

• [DataPipe] Control shuffle settings from DataLoader2

  #65756 commented on Oct 12, 2021 • 0 new comments

• [MonitoredBarrier] Fix some logging

  #65771 commented on Oct 12, 2021 • 0 new comments

• [DDP Logging] Add iteration in error reporting

  #65772 commented on Oct 12, 2021 • 0 new comments

• Add OpInfo for torch.bincount

  #65796 commented on Oct 11, 2021 • 0 new comments

• CPUBlas: Use opmath_type for alpha/beta scalars

  #65839 commented on Oct 12, 2021 • 0 new comments

• Replace issue templates with new issue forms

  #65917 commented on Oct 11, 2021 • 0 new comments

• Scale up

  #65927 commented on Oct 12, 2021 • 0 new comments

• [fx-acc] add optimize_quantization to FX graph opts

  #65929 commented on Oct 12, 2021 • 0 new comments

• [PyTorch] Capture function args from schema by reference

  #65951 commented on Oct 12, 2021 • 0 new comments

• [jit][edge] Add control stack frame to lite interpreter

  #65963 commented on Oct 11, 2021 • 0 new comments

• [jit][edge] Enable CALL instruction in lite interpreter.

  #65964 commented on Oct 12, 2021 • 0 new comments

• [jit] Factor findAllNodes into one place.

  #65965 commented on Oct 8, 2021 • 0 new comments

• [jit] Remove graph() call from abstract Function interface.

  #65967 commented on Oct 7, 2021 • 0 new comments

• [jit] Clean up unneeded virtual methods from Function interface.

  #65968 commented on Oct 7, 2021 • 0 new comments

• [jit] Replace get_executor() with call() in abstract Function interface.

  #65969 commented on Oct 7, 2021 • 0 new comments

• [jit][edge] Implement torch::jit::Function for mobile funciton.

  #65970 commented on Oct 7, 2021 • 0 new comments

• [jit][edge] Load interface methods to corresponding ClassTypes.

  #65971 commented on Oct 7, 2021 • 0 new comments

• [jit][edge] Enable lite interpreter to correctly handle INTERFACE_CALL instruction.

  #65972 commented on Oct 7, 2021 • 0 new comments

• [PyTorch Edge] Don't use LeftRight in mobile

  #66064 commented on Oct 9, 2021 • 0 new comments

• Add support for multi output nodes in partial eval graph stitching

  #66097 commented on Oct 8, 2021 • 0 new comments

• Add support for cat in output stitching

  #66098 commented on Oct 8, 2021 • 0 new comments

• demo backend decoupled from operators

  #66100 commented on Oct 10, 2021 • 0 new comments

• fixed unique arguments documentation

  #66132 commented on Oct 8, 2021 • 0 new comments

• [LTC] Autogen sigmoid and rsqrt for TS backend

  #66133 commented on Oct 11, 2021 • 0 new comments

• [PyTorch][easy] Avoid allocating OperatorName strings in append_operator

  #66134 commented on Oct 12, 2021 • 0 new comments

• Temp

  #66136 commented on Oct 8, 2021 • 0 new comments

• [TestingAChange][NotForReview] remove function_schema from function

  #66156 commented on Oct 6, 2021 • 0 new comments