pytorch / pytorch

28 Pull requests merged by 7 people

• Merged #36334 Use counter instead of vector of futures in `_parallel_run` Apr 9, 2020
• Merged #36286 [v1.5.0] repr and _*state_dict for qRNN (#31540) Apr 9, 2020
• Merged #36289 [v1.5.0] Revert two AccumulateGrad sparse changes Apr 9, 2020
• Merged #36274 [v1.5.0] fix is_float_scale_factor warning (python and c++) Apr 9, 2020
• Merged #36283 [release/1.5] .circleci: Add binary builds/tests to run on release branches Apr 8, 2020
• Merged #36273 [v1.5.0][pytorch] Add error when PyTorch used with Python 2 (#36151) Apr 8, 2020
• Merged #36278 Revert "Update docs for 1.5 to remove Python 2 references (#36116)" Apr 8, 2020
• Merged #36245 [1.5 cherrypick] C++ Adam optimizer - corrected messages for check of default options Apr 8, 2020
• Merged #36256 [release/1.5] Remove determine_from from `test/run_test.py` Apr 8, 2020
• Merged #36251 [release/1.5] Exclude `torch/csrc/cuda/*nccl*` from `clang-tidy` Apr 8, 2020
• Merged #36229 [release/1.5] Move all nccl from torch_python to torch_cuda Apr 8, 2020
• Merged #36238 [v1.5.0] s/repo.continuum.io/repo.anaconda.com/ Apr 8, 2020
• Merged #36185 [v1.5.0][ONNX] fix size for opset 11 (#35984) Apr 8, 2020
• Merged #36195 [1.5] Move test to test_jit_py3.py Apr 8, 2020
• Merged #36165 [v1.5.0] Group libraries in TOC and add PyTorch Elastic Apr 8, 2020
• Merged #36163 [v1.5.0][BugFix] Fix compare_exchange_weak in DispatchStub.h (#35794) Apr 8, 2020
• Merged #36192 [1.5.0] Attempt to fix the pytorch_cpp_doc_push build by pinning brea… Apr 8, 2020
• Merged #36116 Update docs for 1.5 to remove Python 2 references Apr 7, 2020
• Merged #36141 fix max_pool2d cuda version Dimension out of range issue(#36046) Apr 7, 2020
• Merged #36111 [1.5 cherrypick][JIT] Fix fake_range() Apr 7, 2020
• Merged #35919 [1.5 cherrypick] Improve C++ API autograd and indexing docs Apr 7, 2020
• Merged #36006 Fix long line splitting issue in python_print Apr 7, 2020
• Merged #36162 [release/1.5] Remove python2 support from tests and torch.jit (#35042) Apr 7, 2020
• Merged #36148 [release/1.5] Add more alternative filters in places people forgot to add them Apr 7, 2020
• Merged #36109 [v1.5.0] Pin Sphinx to 2.4.4 (take 2), fix docs CIs (#36072) Apr 7, 2020
• Merged #35907 [release/1.5] Reduce manywheel binary size Apr 6, 2020
• Merged #36079 [1.5.0] Fix TopK by splitting files by dtype and half fix Apr 6, 2020
• Merged #35960 [1.5 cherrypick] Refactor and turn on C++ API parity test in CI Apr 6, 2020

122 Pull requests proposed by 74 people

• Proposed #36097 [quant] Adding dynamic qRNN Apr 6, 2020
• Proposed #36098 `ivalue::Future::markCompleted()` to call notify under lock Apr 6, 2020
• Proposed #36101 Adding symbolic sizes, contiguity, stride indices Apr 6, 2020
• Proposed #36121 enable mkldnn conv2d backward to support mkldnn tensor input Apr 7, 2020
• Proposed #36122 add mkldnn linear backward Apr 7, 2020
• Proposed #36134 Fix prelu_backward TensorIterator split Apr 7, 2020
• Proposed #36135 Completely kill CUDAApplyUtils.cuh Apr 7, 2020
• Proposed #36139 [Test] Windows nightly builds on CircleCI Apr 7, 2020
• Proposed #36142 Remove extern C for TH_API Apr 7, 2020
• Proposed #36149 Optimize torch.nn.Transformer.generate_square_subsequent_mask method Apr 7, 2020
• Proposed #36158 add build files to fakelowp Apr 7, 2020
• Proposed #36166 increase timeout, enable test_backend_group Apr 7, 2020
• Proposed #36171 Make custom_fwd a no-op when not executed under autocast Apr 7, 2020
• Proposed #36181 remove serial_exec from scatter/gather kernel Apr 7, 2020
• Proposed #36183 Support forward and forward pre-hooks in Modules Apr 7, 2020
• Proposed #36186 [PyTorch Numeric Suite] Add weight compare API Apr 7, 2020
• Proposed #36187 Create a new bionic image with clang9 Apr 8, 2020
• Proposed #36188 Fix long line splitting issue in python_print Apr 8, 2020
• Proposed #36197 [TensorPipe/RPC] Serialize and deserialize message Apr 8, 2020
• Proposed #36202 Enable torch_speed_benchmark to accept different memory formats. Apr 8, 2020
• Proposed #36213 [WIP] dummy commit Apr 8, 2020
• Proposed #36215 Add RecordFunctionGuard Apr 8, 2020
• Proposed #36225 Implementation and operator test for STORM optimizer Apr 8, 2020
• Proposed #36232 replace Generator arguments with c10::optional<Generator> Apr 8, 2020
• Proposed #36237 [TensorExpr] WIP rfactor Apr 8, 2020
• Proposed #36240 Expunge TensorId from all DispatchKey names. Apr 8, 2020
• Proposed #36250 Unconditionally register schema even for manual registration. Apr 8, 2020
• Proposed #36252 qnnpack hardswish - LUTs Apr 8, 2020
• Proposed #36258 Switch to pybind11 style registration function API. Apr 8, 2020
• Proposed #36260 ci: Change file_diff_from_base to be dynamic Apr 8, 2020
• Proposed #36261 [TEST] Output some diagnostic information Apr 8, 2020
• Proposed #36264 Propagate and report the file/line number of registrations on errors. Apr 8, 2020
• Proposed #36265 Add support of Dict as output when connecting script and tracing Apr 8, 2020
• Proposed #36276 [WIP] dummy commit2 Apr 8, 2020
• Proposed #36277 [jit] tracer flag to guard risky behaivors Apr 8, 2020
• Proposed #36288 Use temporary variable to store input parameters in loop. Apr 9, 2020
• Proposed #36290 Make DispatchKeyExtractor forget about TensorOptions Apr 9, 2020
• Proposed #36291 Make profiler thread local Apr 9, 2020
• Proposed #36293 Add trivial reduce for Cuda Apr 9, 2020
• Proposed #36295 Add documentation for FeatureAlphaDropout Apr 9, 2020
• Proposed #36300 Creates "Versioned Symbol" pattern to preserve serialized Torchscript semantics Apr 9, 2020
• Proposed #36301 Canonicalize includes in aten, and add tests for it Apr 9, 2020
• Proposed #36303 Canonicalize includes in torch, and add tests for it Apr 9, 2020
• Proposed #36305 Sparse softmax support [WIP] Apr 9, 2020
• Proposed #36306 Add the one-block multi-thread global reduction support. Apr 9, 2020
• Proposed #36308 Add `torch.logcumsumexp` Apr 9, 2020
• Proposed #36316 modify `select_equals_backward` to propage only to a single value Apr 9, 2020
• Proposed #36320 qnnpack hardswish - pytorch op integration Apr 9, 2020
• Proposed #36335 Added complex types to get_all_dtypes and turned on masked_fill for complex Apr 9, 2020
• Proposed #36336 Update docs for master to remove Python 2 references Apr 9, 2020
• Proposed #36338 Update docs for 1.5 to remove Python 2 references Apr 9, 2020
• Proposed #36339 Use bazel build rules from fbgemm Apr 9, 2020
• Proposed #36344 silence attributes.h warning Apr 9, 2020
• Proposed #36345 [jit] speed up alias analysis Apr 9, 2020
• Proposed #36346 [shape inference] use max_seq_size as max_feature_len in SLS and LengthsRangeFill inference Apr 9, 2020
• Proposed #36347 Fix `torch.min` docs (#36319) Apr 9, 2020
• Proposed #36348 [TensorExpr] Add simplification of length 0 and 1 For loops to IR Simplifier Apr 9, 2020
• Proposed #36349 CSPRNG PoC Apr 9, 2020
• Proposed #36351 hardsigmoid: add cuda kernels Apr 9, 2020
• Proposed #36355 Optimize grouped Conv3d performance Apr 10, 2020
• Proposed #36357 Python binding api to optimize for mobile model on script module. Apr 10, 2020
• Proposed #36358 Fix multiple issues with type annotations Apr 10, 2020
• Proposed #36362 Remove useless copy on zip file load Apr 10, 2020
• Proposed #36364 D20959297 Apr 10, 2020
• Proposed #36379 PR for double backwards of nn.Fold and nn.Unfold (issue #33452) Apr 10, 2020
• Proposed #36382 Fixing SyncBN dgrad Apr 10, 2020
• Proposed #36388 Update internal invariants in the world of manuallyBoxedKernel Apr 10, 2020
• Proposed #36389 Update reference to RegisterOperators in error message in Convolution Apr 10, 2020
• Proposed #36390 Port Resize.cpp to use new registration API. Apr 10, 2020
• Proposed #36392 track ddp api usage from python Apr 10, 2020
• Proposed #36396 Utilise the vector version for sinh and cosh (UnaryOpsKernel) Apr 10, 2020
• Proposed #36399 Implementation of STORM optimizer caffe2 python wrapper Apr 10, 2020
• Proposed #36400 Convert C casts to static casts (UnaryOpsKernel) Apr 10, 2020
• Proposed #36401 order includes in torch/csrc/autograd/init.cpp Apr 10, 2020
• Proposed #36404 [JIT] Dont optimize shape peepholes on inline Apr 10, 2020
• Proposed #36405 [quant] fixing torch scripting in conv1d Apr 10, 2020
• Proposed #36411 Use unittest assertWarns instead Apr 10, 2020
• Proposed #36412 Move unboxing logic to after dispatch for special cases Apr 10, 2020
• Proposed #36413 Add `torch._C._set_watchdog()` that spawns a pthread-watchdog Apr 10, 2020
• Proposed #36416 Fix a segfault in DeviceThreadHandlePool and PoolWindow Apr 10, 2020
• Proposed #36418 Enable more tests in fbcode Apr 11, 2020
• Proposed #36420 hardswish: add backards pass test Apr 11, 2020
• Proposed #36421 Fix signed integer overflow in rng_test.h Apr 11, 2020
• Proposed #36422 Make patterns in test_unmatched_annotations more flexible Apr 11, 2020
• Proposed #36430 Support ComplexFloat dtype for torch.gather Apr 11, 2020
• Proposed #36431 native_functions.yaml: reset_grad_accumulator Apr 11, 2020
• Proposed #36435 New CUDA Fuser: Unrolling support, interface refactor Apr 11, 2020
• Proposed #36438 Update contribution_guide.rst Apr 11, 2020
• Proposed #36439 Add version number to bytecode. Apr 11, 2020
• Proposed #36442 [quant] qtensor resize Apr 11, 2020
• Proposed #36445 torch.isclose for complex dtype Apr 12, 2020
• Proposed #36447 [WIP] Scatter reductions cpu. Apr 12, 2020
• Proposed #36449 [quant] Generalizing _calculate_dynamic_qparams in quantized test Apr 12, 2020
• Proposed #36450 [quant] quantized reflection_pad1d Apr 12, 2020
• Proposed #36456 [WIP] Implements torch.isclose for complex tensors Apr 12, 2020
• Proposed #36458 Port TH cum{sum,prod}_cuda to ATen Apr 12, 2020
• Proposed #36465 move test_abs to device-generic tests Apr 12, 2020
• Proposed #36467 Teach the tensorexpr vectorizer to handle nested For loops. Apr 13, 2020
• Proposed #36468 [caffe2] Enable copying for caffe2::Tensor Apr 13, 2020
• Proposed #36473 [DON'T REVIEW AND MERGE]bf16 test Apr 13, 2020
• Proposed #36474 always use resize_ in max/min with out= kwarg Apr 13, 2020
• Proposed #36475 [iOS] Fix the import path for benchmarking code Apr 13, 2020
• Proposed #36476 Fixed check for the buffer overflow in assert Apr 13, 2020
• Proposed #36477 Implement kl divergence for Cauchy distribution Apr 13, 2020
• Proposed #36479 Fix for num_threads==1 in OpenMP "parallel for" Apr 13, 2020
• Proposed #36480 Make trivial thread-idx for degenerate statements without thread-idx. Apr 13, 2020
• Proposed #36481 doc note on deterministic/non-deterministic gradient for min/max/median Apr 13, 2020
• Proposed #36483 Port `equal` from THC to ATen (CUDA) Apr 13, 2020
• Proposed #36485 Vectorize non-persistent Softmax kernels Apr 13, 2020
• Proposed #36486 Move to using MemoryFormat::ChannelsLast for avgpool2d. Apr 13, 2020
• Proposed #36487 Add quantized adaptive avgpool. Apr 13, 2020
• Proposed #36488 Replace empty_affine_quantizer with new_qtensor_cpu. Apr 13, 2020
• Proposed #36489 Add channel shuffle op fp32 + quantized. Apr 13, 2020
• Proposed #36490 Enable supplying memory format directly to set_sizes_contiguous. Apr 13, 2020
• Proposed #36491 [Mobile GPU][Integration][proto] Vulkan/GL Integration prototype Apr 13, 2020
• Proposed #36492 Remove duplicate bindings from torch/csrc/jit/python/init.cpp. Apr 13, 2020
• Proposed #36494 [quant] Update qbatch_norm name to qbatch_norm2d Apr 13, 2020
• Proposed #36495 [TensorBoard] fix #36471 Apr 13, 2020
• Proposed #36496 [TensorBoard] fix #34954 Apr 13, 2020
• Proposed #36497 [TensorBoard] fix #33140 Apr 13, 2020
• Proposed #36498 [Shape Inference] Infer input(1) from input(0) in elementwise ops Apr 13, 2020
• Proposed #36501 [ONNX] Adding 'numel' and 'to' export for script module Apr 13, 2020

77 Issues closed by 48 people

• Closed #33494 Complex number printing inconsistent with float Apr 13, 2020
• Closed #36500 torch.blkdiag for sparse tensors Apr 13, 2020
• Closed #33910 [jit] constant prop works for ops that output Objects Apr 13, 2020
• Closed #31932 torch.blkdiag [A way to create a block-diagonal matrix] Apr 13, 2020
• Closed #33226 SummaryWriter.add_video() creates videos that don't match input tensors Apr 13, 2020
• Closed #36482 torch.argmax() returns wrong indices for very big GPU tensors Apr 13, 2020
• Closed #36133 OneCycleLr : which optimizer Apr 13, 2020
• Closed #36443 For margin-type loss used in multiclass classification problem, why torch.nn.functional only has MultiMarginLoss instead of the more important margin loss? Is it because the latter are more difficult to optimize? Why? Apr 13, 2020
• Closed #36387 Can't import iOS libtorch headers when used as a dependency of another Pod. Apr 13, 2020
• Closed #36410 RuntimeError for randn on cuda Apr 13, 2020
• Closed #36327 torch.eye cpu not supported for complex Apr 13, 2020
• Closed #29706 ONNX can not export Unfold Apr 13, 2020
• Closed #36457 RuntimeError: Cowardly refusing to serialize non-leaf tensor which requires_grad, since autograd does not support crossing process boundaries Apr 13, 2020
• Closed #36454 TracerWarning: Converting a tensor to a Python number might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs! Apr 13, 2020
• Closed #36464 How do pytorch calculate image gradient Apr 13, 2020
• Closed #36470 one of the variables needed for gradient computation has been modified by an inplace operation Apr 13, 2020
• Closed #36472 LBFGS Optimizer is non-deterministic? Apr 13, 2020
• Closed #36359 /app/install/libtorch/examples-cpp/cpp/example-app/example-app.cpp:9:24: error: 'indexing' is not a namespace-name Apr 13, 2020
• Closed #36361 torch.onnx.symbolic_helper' has no attribute '_get_interpolate_attributes' Apr 13, 2020
• Closed #36384 [quantization] how to quantize model which include not support to quantize layer Apr 13, 2020
• Closed #36385 GatherElements Apr 13, 2020
• Closed #36367 how to ensure the quality of pytorch framework? Apr 13, 2020
• Closed #36429 return multiple tensors when inferencing on a torch script Apr 13, 2020
• Closed #36423 Is there any loss that can count number of distinct values in tensor? Apr 13, 2020
• Closed #36460 Help me PLEASE!!!!! OSError: [WinError 127] Apr 12, 2020
• Closed #36322 masked_fill not implemented for complex Apr 12, 2020
• Closed #36433 Difference in performance after loading a trained custom LSTM module (torch 1.4.0) Apr 12, 2020
• Closed #36441 Numerous Warnings and Errors occurs when compiling if including <torch/torch.h> in .cu files Apr 12, 2020
• Closed #36221 stuck when using python multiprocess.Process with Pytorch model Apr 11, 2020
• Closed #34431 Linking error in torch_shm_manager near end of compilation Apr 11, 2020
• Closed #36298 DistributedDataParallel on multiple GPU nodes slower than one GPU node. Apr 11, 2020
• Closed #32930 TupleUnpack outputs not used in the correct places in the trace graph Apr 10, 2020
• Closed #19925 ProcessGroupNCCL.cpp:260, unhandled cuda error, when using 2 nodes with 4 GPUs each Apr 10, 2020
• Closed #33054 Migrate `masked_select` from TH to ATen (CUDA) Apr 10, 2020
• Closed #32342 NCCL_ROOT_DIR in Cmake FindNCCL Apr 10, 2020
• Closed #36311 RuntimeError: cuda runtime error (710) : device-side assert triggered at /pytorch/aten/src/THC/THCGeneral.cpp:313 Apr 10, 2020
• Closed #36372 Hook remove not work Apr 10, 2020
• Closed #36034 Freeze when importing CARLA and Torchvision (jit.script related) Apr 10, 2020
• Closed #36279 Extra whitespace issue when printing complex tensors Apr 10, 2020
• Closed #35020 `device_count()` returns 1 while `torch._C._cuda_getDeviceCount()` returns 2? Apr 10, 2020
• Closed #36285 When running DistributedDataParallel, GPU0 sometimes goes to 0 utilization. Apr 9, 2020
• Closed #24615 Migrate `remainder` and `remainder_` from the TH to Aten (CUDA) Apr 9, 2020
• Closed #36242 Thread Joining Consistency in RPC Agent Apr 9, 2020
• Closed #32345 Add alltoall collective communication support to torch.distributed module Apr 9, 2020
• Closed #16012 _grad_input_padding does not consider the dilation. Apr 9, 2020
• Closed #36315 PyTorch no longer supports this GPU because it is too old. Apr 9, 2020
• Closed #36045 Crash when creating a tensor from a tensor on a different device Apr 9, 2020
• Closed #36271 python script to cuda Apr 9, 2020
• Closed #36219 Non contiguous output of meshgrid Apr 8, 2020
• Closed #36218 Deserialize BERT Sequence Lassifier Quantized Model & Inferencing Issue #3695 Apr 8, 2020
• Closed #36227 CMakeFiles/Makefile2:923: recipe for target 'caffe2/CMakeFiles/caffe2_gpu.dir/all' failed Apr 8, 2020
• Closed #36210 What kind of function does USE_QNNPACK do? Apr 8, 2020
• Closed #36132 how to get cuda stream in torch 1.5? Apr 8, 2020
• Closed #36270 nn.Dropout layer TypeError: ‘<’ not supported between instances of ‘Dropout’ and ‘int’ Apr 8, 2020
• Closed #35858 Question/Possible bug with ddp/DistributedDataParallel accessing a parameter.data field Apr 8, 2020
• Closed #35863 test_grad_copy_sparse_indices_extra_ref is flaky Apr 8, 2020
• Closed #36246 [JIT] Cannot do type refinement on class attribute Apr 8, 2020
• Closed #36060 C++ Adam optimizer Apr 8, 2020
• Closed #35470 Failing to replicate model on MultipleGPUs Apr 8, 2020
• Closed #36055 [JIT] After scripting generator variables affects outer local variables Apr 8, 2020
• Closed #33830 Error when calling `prepare_qat` on model with PerChannelQuant support Apr 8, 2020
• Closed #36207 python setup.py egg_info leads to a build Apr 8, 2020
• Closed #34895 Delete as_variable_ref Apr 8, 2020
• Closed #35378 Deadlock on GCC 5.4 when running test_nn.py Apr 8, 2020
• Closed #35940 commit version 35715a56a93363bc61fdf924175d4c4d6352f813 cause DLRM benchmark performance drop Apr 8, 2020
• Closed #36204 got cmakelists error Apr 8, 2020
• Closed #36117 Jitscript doesnot support tensor to list conversion? Apr 8, 2020
• Closed #36200 misspelling `algoriths`, while the true word should be `algorithms`. Apr 8, 2020
• Closed #36156 Multiprocessing context not compatible with ProcessPoolExecutor Apr 7, 2020
• Closed #36046 IndexError: Dimension out of range (expected to be in range of [-3, 2], but got -4) Apr 7, 2020
• Closed #36089 padding_mode Apr 7, 2020
• Closed #36143 Bitwise XOR inconsistent behavior depending on input dtype Apr 7, 2020
• Closed #34984 Rename CPUGenerator to CPUGeneratorImpl and CUDAGenerator to CUDAGeneratorImpl Apr 7, 2020
• Closed #34987 Hide Generator's mutex behind a method Apr 7, 2020
• Closed #30379 The guidelines for loading a PyTorch model in C++ do not work on Windows Apr 7, 2020
• Closed #31846 test_debug_info is flaky Apr 6, 2020
• Closed #36069 Cannot convert model to onnx Apr 6, 2020

85 Issues created by 66 people

• Opened #36499 `F.logsigmoid(input, out=blah)` crashes Apr 13, 2020
• Opened #36493 torch.cdist gradients are NAN for p<1 and very small differencens in a given dimension (0<delta<~e-45) Apr 13, 2020
• Opened #36484 Reduce functions don't work with Android Java Wrapper Apr 13, 2020
• Opened #36478 JIT ONNX: List indexing fails when converting ScriptModule to ONNX Apr 13, 2020
• Opened #36471 [Tensorboard] Incorrect (stale) docs for `add_scalars()` method in `torch.utils.tensorboard.SummaryWriter` class Apr 13, 2020
• Opened #36469 Wrong results for multiplication of non-finite complex numbers with real numbers Apr 13, 2020
• Opened #36466 torch.cuda.amp.GradScaler: RuntimeError Apr 13, 2020
• Opened #36463 Autograd.grad not making higher order graph for scaling functions Apr 12, 2020
• Opened #36462 torch.isclose for complex tensors Apr 12, 2020
• Opened #36461 Pretrained model in libtorch gives different results each time Apr 12, 2020
• Opened #36459 nn.Parallel similar to nn.Sequential Apr 12, 2020
• Opened #36455 torch.Tensor.new_full ignores requires grad Apr 12, 2020
• Opened #36452 Curreent SyncBatchNorm (a synchronized implementation) slow down DDP training Apr 12, 2020
• Opened #36451 torch.isclose is not implemented for cpu half tensors Apr 12, 2020
• Opened #36448 nn.utils.rnn.pack_padded_sequence is broken Apr 12, 2020
• Opened #36444 Comparison ops for Complex Tensors Apr 12, 2020
• Opened #36437 Issue when linking C++ code with libtorch_cpu: cuda not detected Apr 11, 2020
• Opened #36436 After `create_graph=True`, calculating `backward()` on sparse Tensor fails Apr 11, 2020
• Opened #36434 Intermittent hang in BuildLivenessSets after test_svd_cuda_float64 on CUDA 10.1 configurations Apr 11, 2020
• Opened #36432 TorchScript doesn't parse floats that are not inf-y. Apr 11, 2020
• Opened #36428 AMP autocast error in cnn-lstm forward Apr 11, 2020
• Opened #36426 Any reference to LPPool2d Apr 11, 2020
• Opened #36425 [AdaptiveMaxPool3d] operator export to onnx error Apr 11, 2020
• Opened #36424 nn.Conv2d doesn't handle well padding in the borders of matrix Apr 11, 2020
• Opened #36408 Intermittent segfaults during libtorch shutdown + Proposed Fix Apr 10, 2020
• Opened #36407 Unable to build wheel from RC3 Apr 10, 2020
• Opened #36403 Incorrect view behavior for split/split_with_sizes/chunk Apr 10, 2020
• Opened #36386 Docs of distributed module do not include the full documentation for torch.distributed.launch Apr 10, 2020
• Opened #36383 CheckpointFunctionBackward not called if the output of the checkpointed function is an in-place modified view Apr 10, 2020
• Opened #36381 Low cache hit count for conda builds on Windows Apr 10, 2020
• Opened #36380 Population Count Op Apr 10, 2020
• Opened #36378 CMake targets wrongly forward unknown options to NVCC (v1.5+) Apr 10, 2020
• Opened #36375 torch.utils.data.DataLoader multiprocess hanging when matplotlib.pyplot imported after torch Apr 10, 2020
• Opened #36374 Min and Max with complex inputs exhibit behavior incompatible with NumPy Apr 10, 2020
• Opened #36370 Jit doesn't match schema like eager mode does Apr 10, 2020
• Opened #36369 jit creates double tensors from float scalars Apr 10, 2020
• Opened #36368 The jit misinterprets complex values as integers Apr 10, 2020
• Opened #36363 Accept tensor-like objects in place of tensor inputs Apr 10, 2020
• Opened #36360 Create helper function for RPC profiling in _invoke_rpc and remote() Apr 10, 2020
• Opened #36356 Memory leak issue still exists in CI Apr 10, 2020
• Opened #36341 torch.quasirandom.SobolEngine(..., scramble=True) does not respect torch.manual_seed Apr 9, 2020
• Opened #36340 torch.gather doesn't work for complex tensors Apr 9, 2020
• Opened #36333 Illegal memory access / CUDNN_STATUS_MAPPING_ERROR on Quadro 8000 Apr 9, 2020
• Opened #36332 DISABLED test_check_failed_messages (__main__.FaultyAgentRpcTestWithSpawn) Apr 9, 2020
• Opened #36328 index_fill not supported for complex tensors Apr 9, 2020
• Opened #36326 Support quant fusion with custom ops Apr 9, 2020
• Opened #36323 torch.sign doesn't work for complex tensors Apr 9, 2020
• Opened #36318 Docs should say that standard arithmetic operations for torch.half are not supported Apr 9, 2020
• Opened #36317 Updating a ModelDict instance with another ModelDict instance generates error. Apr 9, 2020
• Opened #36314 [discussion] Refactor spectral_norm to use the newly merged lowrank solvers and proposal for Linear Algebra Cookbook page Apr 9, 2020
• Opened #36313 local variable 'values' referenced before assignment Apr 9, 2020
• Opened #36312 Cannot build pytorch inside rocm/pytorch:rocm3.3_ubuntu16.04_py3.6_pytorch contianer Apr 9, 2020
• Opened #36309 torch.ceil, torch.floor should accept a dtype argument Apr 9, 2020
• Opened #36307 [docs] Matrix transpose missing in formula in nn.functional.bilinear docs Apr 9, 2020
• Opened #36304 why the quantized weights are showed in wrong values? Apr 9, 2020
• Opened #36294 NCCL Connection Failed Using PyTorch Distributed with Error ProcessGroupNCCL.cpp:290, unhandled system error Apr 9, 2020
• Opened #36272 Add support for delays in FaultyProcessGroupAgent Apr 8, 2020
• Opened #36268 DistributedDataParallel don't work at nightly build(1.6.0.dev20200408+cu101) Apr 8, 2020
• Opened #36253 [quantization] Error in QNNPACK: failed to create convolution with 0.6145506 input scale, 1.624538 kernel scale, and 0.2153073 output scale: convolution scale 4.636909 is greater or equal to 1.0 Apr 8, 2020
• Opened #36247 [JIT] Doesn't support dict() / list() Apr 8, 2020
• Opened #36217 Rectify docs for MultiLabelSoftMarginLoss Apr 8, 2020
• Opened #36216 [doc] Part of the JIT code is not highlighted Apr 8, 2020
• Opened #36214 Pruning doesn't affect speed nor memory usage Apr 8, 2020
• Opened #36211 [JIT] support self.named_buffers and self.named_parameters in TorchScript Apr 8, 2020
• Opened #36191 Calling onnx export hangs using multiprocessing Apr 8, 2020
• Opened #36182 ## 🐛 Bug: QNNPACK tests failing on master on Nexus 6 Apr 7, 2020
• Opened #36176 len(DataLoader) doesn't take into account DataLoader.batch_size when using IterableDataset Apr 7, 2020
• Opened #36175 New dtype ComplexPolarFloat Apr 7, 2020
• Opened #36169 [question] Is amp CUDA-specific? Apr 7, 2020
• Opened #36160 Decouple DDP from CUDA Apr 7, 2020
• Opened #36155 Conv3d performance improvement Apr 7, 2020
• Opened #36152 Seg Fault: import vaex with torch Apr 7, 2020
• Opened #36147 Ownership and lifetime of `c10::ivalue::Future` Apr 7, 2020
• Opened #36145 Update the MSVC toolchain for Windows release CIs Apr 7, 2020
• Opened #36140 Inconsistent ProcessGroupMPI work data structure for send/recv and collectives Apr 7, 2020
• Opened #36136 Ctrl+C exit leaves unhandled processes using multiprocessing Apr 7, 2020
• Opened #36129 test_avg_pool3d_nhwc is failing on parallel builds (like pytorch_parallelnative_linux_xenial_py3_6_gcc5_4_test) Apr 7, 2020
• Opened #36127 test_qadd_scalar_relu is failing on pytorch_macos_10_13_py3_test Apr 7, 2020
• Opened #36125 build Pytorch-1.5.0-rc1 from source fail Apr 7, 2020
• Opened #36120 Calling sparse_coo_tensor in accumulateGrad makes it much slower Apr 7, 2020
• Opened #36108 Add a flow_sample function to sample optical flows and displacement fields Apr 6, 2020
• Opened #36107 Allow grid_sample to accept pixel units (absolute coordinates) Apr 6, 2020
• Opened #36104 minor formatting issue in the quantization doc Apr 6, 2020
• Opened #36099 In AutogradContext, get_saved_variables() should be renamed to get_saved_tensors() Apr 6, 2020
• Opened #36091 [quantization] torch.quantized_lstm and torch.quantized_gru not documented Apr 6, 2020

211 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.

• 34 new comments Open #36073 Save view_fn for inplace update on view tensors Apr 13, 2020
• 33 new comments Open #35524 Add core of c10::complex Apr 10, 2020
• 32 new comments Open #34830 ONNX Export Support for CrossEntropyLoss Apr 11, 2020
• 32 new comments Open #36032 Mobile CPU allocator. Apr 13, 2020
• 29 new comments Open #35055 [profiler] Extend record_function ctx manager to profile futures Apr 13, 2020
• 27 new comments Open #33344 [WIP] Parametrization Functionality Apr 12, 2020
• 23 new comments Open #35744 [ONNX] Adding a pass to replace interpolate function with aten::__interpolate Apr 13, 2020
• 21 new comments Open #35871 Returns float tensors for complex inputs to abs Apr 12, 2020
• 19 new comments Open #36029 Implement Complex Support that allows Complex Tests to run in test_torch.py Apr 12, 2020
• 17 new comments Open #35699 backward for tensor.min() and tensor.min(dim=0) behaves differently Apr 9, 2020
• 16 new comments Open #33763 Add support for bool/byte `attn_mask` tensor in MultiheadAttention/Transformer modules Apr 8, 2020
• 16 new comments Open #35923 [quant] ConvPackedParams with TorchBind Apr 13, 2020
• 14 new comments Open #33430 Bmm sparse dense Apr 13, 2020
• 14 new comments Open #35855 Implement avg_pool2d kernel for channels_last Apr 9, 2020
• 13 new comments Open #33057 Move profiler to a dispatch wrapper Apr 13, 2020
• 12 new comments Open #35009 [v1.5.0] Release Tracker Apr 10, 2020
• 12 new comments Open #35483 [Tensorpipe/RPC] tensorpipe RPC agent Apr 12, 2020
• 10 new comments Open #35463 QuantizedCUDA implementation Apr 13, 2020
• 10 new comments Open #35771 Python API for Complex Storage [CPU only] Apr 7, 2020
• 10 new comments Open #36052 Changes warnings generated in cpp to show point of Python origination Apr 7, 2020
• 9 new comments Open #32916 from torch._C import * (ImportError: DLL load failed) Apr 10, 2020
• 9 new comments Open #34059 serialization: validate sparse tensors after loading Apr 11, 2020
• 9 new comments Open #35896 Add XNNPACK Clamp. Apr 13, 2020
• 8 new comments Open #35561 Pytorch 1.4 much slower than 1.3 on Ubuntu 18.04 Apr 9, 2020
• 8 new comments Open #30898 Migrate addmv and mv from legacy to ATen native (CUDA & CPU) Apr 10, 2020
• 8 new comments Open #35172 bfloat16: enable basic math function Apr 13, 2020
• 8 new comments Open #35429 Port masked_select cuda from TH to ATen Apr 13, 2020
• 8 new comments Open #35569 Add Inverse Short Time Fourier Transform in ATen native Apr 13, 2020
• 8 new comments Open #35679 set_: reset AccumulateGrad ptr Apr 10, 2020
• 6 new comments Open #31975 TensorIterator unrolling and vectorized load for GPU loop Apr 7, 2020
• 6 new comments Open #32284 Memory leak in MaxPool2d Apr 13, 2020
• 6 new comments Open #33269 port masked_select from TH to ATen and optimize perf on CPU Apr 12, 2020
• 6 new comments Open #34577 Bucketization (#7284) Apr 13, 2020
• 6 new comments Open #34768 DataLoader: properly diagnose exceeding file descriptor limit Apr 13, 2020
• 6 new comments Open #35790 Enable stateless XNNPACK convolutions. Apr 13, 2020
• 6 new comments Open #35833 [WIP][JIT] Add JIT backend registration API Apr 11, 2020
• 6 new comments Open #35952 Enforce tensorboard minimum version as 1.15 Apr 13, 2020
• 5 new comments Open #35878 Use libtorch to load module failed Apr 8, 2020
• 5 new comments Open #36071 [RFC] Async User Function for RPC Apr 8, 2020
• 5 new comments Open #20643 RuntimeError: ONNX export failed: Couldn't export operator aten::softmax Apr 9, 2020
• 5 new comments Open #35446 [JIT] dropout fails on legacy mode Apr 13, 2020
• 5 new comments Open #31860 [jit] Add dictionary as output of tracer Apr 13, 2020
• 5 new comments Open #34501 [PyTorch][Dist] Trigger hooks of output function nodes under distributed autograd Apr 13, 2020
• 5 new comments Open #34787 Remove attributes and method of submodules in frozen module Apr 9, 2020
• 4 new comments Open #36035 nn.Parameter not copied to gpus in forward pass when nn.DataParallel is used Apr 7, 2020
• 4 new comments Open #36061 [JIT] support list of nn.Module in torchscript Apr 8, 2020
• 4 new comments Open #35942 Build PyTorch-1.4.0 from source failed Apr 9, 2020
• 4 new comments Open #33452 2nd derivative for im2col and col2im not implemented Apr 9, 2020
• 4 new comments Open #32575 Interrupting a DDP worker while using the CUDA MPS server causes CUDA to hang until reboot Apr 10, 2020
• 4 new comments Open #35954 torch.finfo doesn't work for complex Apr 11, 2020
• 4 new comments Open #36038 commit version 618104185bce6591c4697063b1faf2027c3761e7 cause DLRM benchmark performance drop Apr 13, 2020
• 4 new comments Open #31278 BTRS algorithm for fast/efficient binomial sampling Apr 13, 2020
• 4 new comments Open #35029 Enable uint8 upsampling 2 Apr 9, 2020
• 4 new comments Open #35263 Fix for test_backward_node_failure: Error Handling in RPC Agent Apr 7, 2020
• 4 new comments Open #35638 Port `histc` cpu from TH to ATen Apr 8, 2020
• 4 new comments Open #35941 [pytorch] improve code analyzer to dump ops called from c++ functions Apr 8, 2020
• 4 new comments Open #36016 Allowing casting str to int in JIT Apr 11, 2020
• 3 new comments Open #36036 torch.multinomial is misnamed. Apr 8, 2020
• 3 new comments Open #31895 torch.masked_select Apr 8, 2020
• 3 new comments Open #34715 Maybe gpu_kernel shouldn't ASSERT_HOST_DEVICE_LAMBDA Apr 9, 2020
• 3 new comments Open #33568 Inconsistency between torch.clamp() and numpy.clip() behavior for complex numbers Apr 12, 2020
• 3 new comments Open #18113 Can't convert Upsample to onnx Apr 13, 2020
• 3 new comments Open #32322 deadlock when using mp.spawn multiprocessing Apr 13, 2020
• 3 new comments Open #27542 Make topk sort stable Apr 13, 2020
• 3 new comments Open #30566 Update sampler.py Apr 13, 2020
• 3 new comments Open #35831 Add path to NeuralNet.h when USE_TENSORRT=1 Apr 10, 2020
• 3 new comments Open #35856 Enable float only requantization. Part 1. Apr 13, 2020
• 3 new comments Open #36087 Type Promotion for complex and bool pair Apr 7, 2020
• 2 new comments Open #29724 Failed to convert quantized mobilenet to onnx Apr 7, 2020
• 2 new comments Open #35418 PyTorch / libtorch executables fail when built against libcuda stub library Apr 9, 2020
• 2 new comments Open #32189 Need unittests for the quantization save_state/load_state Apr 9, 2020
• 2 new comments Open #35180 hasSpecialCase INTERNAL ASSERT FAILED: We don't have an op for aten::to but it isn't a special case. Apr 10, 2020
• 2 new comments Open #35901 OOM error where ~50% of the GPU RAM cannot be utilised/reserved Apr 10, 2020
• 2 new comments Open #36058 Half type promotion with Numpy arrays is incorrect Apr 10, 2020
• 2 new comments Open #28478 How to train a torch::jit::script::Module? Apr 12, 2020
• 2 new comments Open #34002 Cannot export ScriptModule models to ONNX format Apr 13, 2020
• 2 new comments Open #13246 CPU memory gradually leaks when num_workers > 0 in the DataLoader Apr 13, 2020
• 2 new comments Open #32898 OneCycleLearningrate is bug in version Apr 13, 2020
• 2 new comments Open #32561 PyTorch 1.4 ONNX Producer Version Mismatch Apr 13, 2020
• 2 new comments Open #33140 Summary writer doesn't support None Apr 13, 2020
• 2 new comments Open #33286 Migrate equal from the TH to Aten (CPU) Apr 11, 2020
• 2 new comments Open #34004 Migrate poisson from CUDA_tensor_apply2 to TensorIterator Apr 9, 2020
• 2 new comments Open #35088 pop warning message for cuda module when asan is built in Apr 8, 2020
• 2 new comments Open #35098 pop warning message for multiprocessing module when asan is built in Apr 8, 2020
• 2 new comments Open #35694 [WIP] [JIT] Consolidate RRef type class and RRef impl class Apr 7, 2020
• 2 new comments Open #35697 Migrate `scatter` and `scatter_` from the TH to Aten (CUDA) Apr 13, 2020
• 1 new comment Open #35872 torch.autograd.set_detect_anomaly(True) does not exist in C++? Apr 6, 2020
• 1 new comment Open #35641 Dimension reducing variants of bitwise operations (bitwise_or, bitwise_and, bitwise_xor) Apr 6, 2020
• 1 new comment Open #25039 Add new interpolation modes to `grid_sample` Apr 6, 2020
• 1 new comment Open #35937 MKLDNN_conv2d 2X slower than the native TH implementation Apr 7, 2020
• 1 new comment Open #28293 RuntimeError: svd_cuda: the updating process of SBDSDC did not converge (error: 1) Apr 7, 2020
• 1 new comment Open #36064 Stop using javasphinx Apr 7, 2020
• 1 new comment Open #35566 nn.Module subclasses don't typecheck due to lack of contravariant input types Apr 7, 2020
• 1 new comment Open #18475 The latest version of onnx-caffe2 does not support “pow” ？ Apr 7, 2020
• 1 new comment Open #35975 [RFC] Target Specifier for CI on Pull Requests Apr 7, 2020
• 1 new comment Open #21457 downsampling with grid_sample doesn't match interpolate Apr 7, 2020
• 1 new comment Open #31111 Python stub files for torch.nn seem misconfigured Apr 7, 2020
• 1 new comment Open #36040 [JIT] Huge delay (1274s vs 0.031s) when running scripted model Apr 7, 2020
• 1 new comment Open #35990 [JIT] Tensor method API behavior discrepancy, Tensor.detach(..) Apr 7, 2020
• 1 new comment Open #14219 C++ model load error Apr 7, 2020
• 1 new comment Open #34977 `conv2d` is slow with specific shapes of channels_last tensors Apr 7, 2020
• 1 new comment Open #24005 Using `torch.utils.checkpoint.checkpoint_sequential` and `torch.autograd.grad` breaks when used in combination with `DistributedDataParallel` Apr 7, 2020
• 1 new comment Open #35987 [JIT] Tensor method API schema discrepancy, Tensor.requires_grad_(..) Apr 7, 2020
• 1 new comment Open #25883 Python/C++ API Parity: torch.nn modules and functional Apr 8, 2020
• 1 new comment Open #33343 How to convert the model to onnx in libtorch? Apr 8, 2020
• 1 new comment Open #32838 Improve docs search engine indexing Apr 8, 2020
• 1 new comment Open #34032 Documentation is super heavy Apr 8, 2020
• 1 new comment Open #21819 RuntimeError: CUDA error: an illegal memory access was encountered Apr 8, 2020
• 1 new comment Open #21683 Import ONNX model to Pytorch Apr 8, 2020
• 1 new comment Open #35316 Bazel Support Apr 8, 2020
• 1 new comment Open #10370 Different results when using function argmax or argmin for data type ".cuda()" and ".cpu()" Apr 8, 2020
• 1 new comment Open #35926 RunTimeError while unscaling optimizer gradients Apr 8, 2020
• 1 new comment Open #31283 Compile Error for .cu files when using libtorch (caused by nvcc flags) Apr 9, 2020
• 1 new comment Open #34513 RNN + DataParallel + scripting is broken Apr 9, 2020
• 1 new comment Open #9886 [doc] functionalities not documented Apr 9, 2020
• 1 new comment Open #23110 [RFC] RPC Based Distributed Model Parallel Apr 9, 2020
• 1 new comment Open #34862 [C++ API] Support for CIFAR10 and CIFAR100 Datasets Apr 9, 2020
• 1 new comment Open #2389 sparse.mm(S, D) with autograd Apr 9, 2020
• 1 new comment Open #9674 The state of sparse Tensors Apr 9, 2020
• 1 new comment Open #22400 Sparse allreduce for ProcessGroupNCCL Apr 9, 2020
• 1 new comment Open #24241 [feature request] dim/keepdim/multidim for torch.std_mean and torch.var_mean Apr 9, 2020
• 1 new comment Open #32291 PyTorch-1.4.0 doesn't encode numpy dependency Apr 9, 2020
• 1 new comment Open #35456 Share build file lists between CMake, Bazel and Buck builds Apr 9, 2020
• 1 new comment Open #32867 [feature request] Bit packed tensors utilities Apr 10, 2020
• 1 new comment Open #35998 JITed GRU too slow Apr 10, 2020
• 1 new comment Open #36068 Flask report "RuntimeError: cuda runtime error (3) : initialization error " Apr 11, 2020
• 1 new comment Open #26555 Build failure (sleef.h not found) in some hard-to-understand situations Apr 11, 2020
• 1 new comment Open #18631 FP32 depthwise convolution is slow in GPU Apr 11, 2020
• 1 new comment Open #24557 Migrate `equal` from the TH to Aten (CUDA) Apr 11, 2020
• 1 new comment Open #33048 allclose and isclose should support bool tensors Apr 12, 2020
• 1 new comment Open #18757 [MSVC Static] DeviceGuardImpl for CPU is not available Apr 12, 2020
• 1 new comment Open #34954 [Tensorboard] `RuntimeWarning: invalid value encountered in multiply` when calling `add_embedding` with `label_img`. Apr 12, 2020
• 1 new comment Open #16954 torch.multiprocessing.pool.Pool broken Apr 12, 2020
• 1 new comment Open #35471 torch.isclose() undocumented Apr 13, 2020
• 1 new comment Open #35864 PyTorch and Libtorch for Xcode/IOS header file location/visibility issue. Apr 13, 2020
• 1 new comment Open #31264 Failed to config caffe2_rocksdb in cmake Apr 13, 2020
• 1 new comment Open #9983 [feature request] Add matrix exponential Apr 13, 2020
• 1 new comment Open #35881 Error when loading jit traced FasterRCNN model in C++ Apr 13, 2020
• 1 new comment Open #19911 BatchNorm1d does not support batchsize>65535 in eval mode with 3 dimension (NxCxL), raise CUDNN_STATUS_NOT_SUPPORTED Apr 13, 2020
• 1 new comment Open #17501 Tensor unfold backward is slow Apr 13, 2020
• 1 new comment Open #35049 Compilation error on aarch64 Apr 13, 2020
• 1 new comment Open #31095 pin_memory stuck in DDP/Reducer constructor Apr 13, 2020
• 1 new comment Open #27319 [jit] Allow `@unused` on TorchScript classes Apr 13, 2020
• 1 new comment Open #32876 Adds LogCumSumExp Apr 9, 2020
• 1 new comment Open #35101 [autograd] Avoid holding lock when completing GraphTask futureResult Apr 10, 2020
• 1 new comment Open #35359 fixing build failures with USE_NATIVE_ARCH ON Apr 11, 2020
• 1 new comment Open #35607 Fix CMake errors on systems where {Q/X}NNPACK is not supported Apr 7, 2020
• 1 new comment Open #35774 Merge two function_traits into one Apr 7, 2020
• 1 new comment Open #35791 Enable stateless XNNPACK linear. Apr 13, 2020
• 1 new comment Open #35795 Reset `_BaseDataLoaderIter` instead of creating a new one Apr 8, 2020
• 1 new comment Open #36009 Fix NaN error in dynamic quantization in qLinear, re-enable test_quantized_rnn Apr 10, 2020
• 1 new comment Open #36017 find backtrace by cmake module Apr 13, 2020
• 1 new comment Open #36030 Prevent unbounded growth of sparse tensor in add operation Apr 8, 2020
• 1 new comment Open #36076 [quant][graphmode] Insert quant-dequant for LSTM weights Apr 6, 2020
• 0 new comments Open #36062 RuntimeError CUDA error despite CUDA available and GPU supported Apr 6, 2020
• 0 new comments Open #24870 Recommendations for Grid Sample/Affine Grid/Displacement Fields/Optical Flow Apr 6, 2020
• 0 new comments Open #35667 torch::cat has no documentation? Apr 7, 2020
• 0 new comments Open #3818 Memory leak when doing backward with grad as yourself Apr 7, 2020
• 0 new comments Open #35989 Add timeout parameter for JIT rpc_async call Apr 7, 2020
• 0 new comments Open #32287 Python 3.8 Windows JIT test failure Apr 9, 2020
• 0 new comments Open #32344 missing LAPACK Apr 9, 2020
• 0 new comments Open #32277 pytorch 1.4.0 doesn't compile Apr 9, 2020
• 0 new comments Open #32638 TORCH_CUDA_API export failure on torch::cuda::nccl::detail::throw_nccl_error(ncclResult_t) Apr 9, 2020
• 0 new comments Open #32370 Backward `Functional.conv3d` is slow when cuDNN is enabled Apr 9, 2020
• 0 new comments Open #35014 `torch.cat` should do type promotion Apr 9, 2020
• 0 new comments Open #35216 API doc for torch.max/min is broken Apr 9, 2020
• 0 new comments Open #33281 Uninitialised value was created by a stack allocation, reported by valgrind Apr 10, 2020
• 0 new comments Open #21821 Stream safety of MAGMA functions Apr 12, 2020
• 0 new comments Open #31828 Bug in MultiStepLR lr scheduler Apr 12, 2020
• 0 new comments Open #27176 Issue with Exporting Interpolate to ONNX Apr 13, 2020
• 0 new comments Open #34028 'LocalFileSystem' object has no attribute 'makedirs' Apr 13, 2020
• 0 new comments Open #26921 [DO NOT DELETE OR MERGE] Nightly release Apr 13, 2020
• 0 new comments Open #27167 [DO NOT DELETE OR MERGE] Postnightly release Apr 13, 2020
• 0 new comments Open #27804 enhance _embedding_bag_backward_cpu path by adding sum fast path for DLRM Apr 9, 2020
• 0 new comments Open #27957 Vectorize linspace on CPU. Apr 8, 2020
• 0 new comments Open #31301 [tensorboard] Fix function input parameter for add_hparams Apr 13, 2020
• 0 new comments Open #31544 [tensorboard] Let hparam render values correctly Apr 13, 2020
• 0 new comments Open #32587 [WIP] clamp with max and min as tensors Apr 13, 2020
• 0 new comments Open #33338 [WIP] initial implementation of symbolic shapes Apr 10, 2020
• 0 new comments Open #33905 Refactor jit::Operator Apr 10, 2020
• 0 new comments Open #34005 Migrate gamma from CUDA_tensor_apply2 to TensorIterator Apr 8, 2020
• 0 new comments Open #34006 Migrate bernoulli from CUDA_tensor_apply2 to TensorIterator Apr 8, 2020
• 0 new comments Open #34140 [WIP][quantization] Use torchbind for Linear and RNN PackedParams Apr 7, 2020
• 0 new comments Open #34250 Standardize expanded TensorOptions representation in native_functions.yaml Apr 6, 2020
• 0 new comments Open #34251 Change default value of MemoryFormat? to None Apr 6, 2020
• 0 new comments Open #34252 Add MemoryFormat to TensorOptions in codegen. Apr 6, 2020
• 0 new comments Open #34258 Use both absolute and relative tolerance in testing Apr 8, 2020
• 0 new comments Open #34396 add quantized GELU activation Apr 8, 2020
• 0 new comments Open #34767 [ONNX] Support for torch.logical_not Apr 12, 2020
• 0 new comments Open #34793 [DO NOT MERGE]Decompose MultiheadAttention module Apr 6, 2020
• 0 new comments Open #34827 Migrate inverse short-time Fourier transform from torchaudio Apr 7, 2020
• 0 new comments Open #35083 [WIP] Support delay allreduce in DDP Apr 8, 2020
• 0 new comments Open #35286 autograd: move scalar input to a different device when needed Apr 13, 2020
• 0 new comments Open #35327 DO NOT MERGE, getting signal for release/1.5 Apr 9, 2020
• 0 new comments Open #35371 Quantized ConvTranspose2d Apr 9, 2020
• 0 new comments Open #35460 Quantized ConvTranspose1d Apr 9, 2020
• 0 new comments Open #35492 Fix linter warning: replace f-strings with str.format for Py2 compat Apr 13, 2020
• 0 new comments Open #35537 Port `renorm` cpu from TH to ATen Apr 10, 2020
• 0 new comments Open #35563 [WIP]: Add polar method for complex tensors. Apr 7, 2020
• 0 new comments Open #35577 Reenable quantized RNN test Apr 7, 2020
• 0 new comments Open #35682 [TensorBoard] fixed issue when key is an int, and typo Apr 11, 2020
• 0 new comments Open #35702 WIP [release/1.5] Fix PyTorch separate compilation (Reland) Apr 8, 2020
• 0 new comments Open #35762 CUDA implementation of Sparse Adagrad Fusion for GPUs Apr 7, 2020
• 0 new comments Open #35787 [WIP] Bind in CellParams for RNN Apr 7, 2020
• 0 new comments Open #35836 [PyTorch BC] Enhance model shrink logic Apr 9, 2020
• 0 new comments Open #35934 Split torch.nn docs Apr 7, 2020
• 0 new comments Open #35962 Strip relocatable cubins during CUDA separable compilation Apr 7, 2020
• 0 new comments Open #36043 D20860370 Apr 7, 2020
• 0 new comments Open #36050 [Lite Interpreter] Remove register_mobile_ops.cpp Apr 9, 2020
• 0 new comments Open #36051 [Lite Interpreter] Put mobile-used prim operators in a new file Apr 9, 2020
• 0 new comments Open #36070 [DO NOT MERGE] 1.5 release branch Apr 10, 2020