pytorch/pytorch issues and pull requests

#165891 - [2/N] Change C-style casts to static_cast or reinterpret_cast

Pull Request - State: open - Opened by cyyever about 1 month ago - 1 comment
Labels: module: cpu, open source, release notes: quantization, release notes: linalg_frontend

#165477 - Add pyrefly suppressions 11/n (using ghstack)

Pull Request - State: open - Opened by maggiemoss about 1 month ago - 1 comment
Labels: oncall: distributed, oncall: jit, module: cpu, release notes: quantization, fx, module: inductor, module: dynamo, ciflow/inductor, release notes: distributed (checkpoint), release notes: export

#165317 - [Code Clean] Clean asserts in torch/ao/quantization/experimental/* and torch/ao/quantization/pt2e/*

Pull Request - State: open - Opened by zhudada0120 about 1 month ago - 6 comments
Labels: oncall: distributed, module: cpu, triaged, open source, release notes: quantization, topic: not user facing, module: inductor, module: dynamo, release notes: distributed (checkpoint), release notes: inductor (aoti)

#165309 - [2/N] Don't return values in void functions

Pull Request - State: open - Opened by cyyever about 1 month ago - 1 comment
Labels: oncall: distributed, oncall: jit, module: cpu, triaged, open source, NNC, release notes: sparse, ciflow/inductor, release notes: inductor (aoti), ciflow/h100-symm-mem

#165273 - [Pytorch] Add NEON Vectorized<int> family of translation layers

Pull Request - State: open - Opened by Nicoshev about 1 month ago - 4 comments
Labels: module: cpu, fb-exported, ciflow/trunk, topic: not user facing, ciflow/linux-aarch64, meta-exported

#164877 - Pyrefly suppressions 6/n

Pull Request - State: closed - Opened by maggiemoss about 1 month ago - 3 comments
Labels: oncall: jit, module: cpu, Merged, ciflow/trunk, release notes: quantization, fx, module: dynamo, ciflow/inductor, merging, release notes: export

#164809 - Don't return values in void functions

Issue - State: closed - Opened by cyyever about 1 month ago - 3 comments
Labels: module: cpu, triaged, open source, Merged, ciflow/trunk, release notes: sparse, merging

#164692 - [3/N] Add clang-tidy readability checks

Pull Request - State: open - Opened by cyyever about 2 months ago - 3 comments
Labels: oncall: distributed, oncall: jit, module: cpu, open source, module: amp (automated mixed precision), NNC, ciflow/trunk, release notes: quantization, topic: not user facing, ciflow/inductor, release notes: inductor (aoti), ciflow/h100-symm-mem

#164652 - [2/N] Fix clang-tidy readability checks

Pull Request - State: closed - Opened by cyyever about 2 months ago - 11 comments
Labels: oncall: distributed, oncall: jit, module: cpu, open source, Merged, NNC, Reverted, ciflow/trunk, release notes: quantization, module: inductor, module: dynamo, ciflow/inductor, release notes: distributed (checkpoint), merging, module: compiled autograd, ci-no-td, release notes: inductor (aoti), ciflow/h100-symm-mem

#164645 - Enable all SIM rules except disabled ones

Pull Request - State: open - Opened by cyyever about 2 months ago - 15 comments
Labels: oncall: distributed, module: cpu, open source, Merged, Reverted, ciflow/trunk, release notes: quantization, release notes: distributed (sharded), fx, module: inductor, module: dynamo, ciflow/inductor, release notes: distributed (checkpoint), ci-no-td

#164645 - Enable all SIM rules except disabled ones

Pull Request - State: open - Opened by cyyever about 2 months ago - 15 comments
Labels: oncall: distributed, module: cpu, open source, Merged, Reverted, ciflow/trunk, release notes: quantization, release notes: distributed (sharded), fx, module: inductor, module: dynamo, ciflow/inductor, release notes: distributed (checkpoint), ci-no-td

#164450 - Remove workaround to old CUDA bug

Pull Request - State: closed - Opened by pearu about 2 months ago - 2 comments
Labels: module: cpu

#164354 - Remove workaround to old CUDA bug

Pull Request - State: open - Opened by pearu about 2 months ago - 4 comments
Labels: module: cuda, module: cpu, open source, release notes: cpp, topic: not user facing

#164202 - [Inductor] support masked vectorization for the tail_loop for bool datatype

Pull Request - State: open - Opened by jiayisunx about 2 months ago - 1 comment
Labels: module: cpu, open source, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor

#164177 - Add initial suppressions for pyrefly

Pull Request - State: closed - Opened by maggiemoss about 2 months ago - 6 comments
Labels: module: cpu, module: amp (automated mixed precision), Merged, ciflow/trunk, release notes: quantization, release notes: linalg_frontend, merging

#164054 - [2/N] Apply ruff UP035 check in torch files

Issue - State: closed - Opened by cyyever about 2 months ago - 9 comments
Labels: oncall: distributed, module: cpu, open source, Merged, ciflow/trunk, release notes: foreach_frontend, merging

#164044 - Speed up FP precision lookup

Pull Request - State: closed - Opened by lakshayg about 2 months ago - 15 comments
Labels: module: cpu, triaged, open source, Merged, Reverted, ciflow/trunk, topic: not user facing, module: dynamo, ciflow/inductor, merging, ci-no-td

#164042 - [AMP] Add deprecated decorator for torch.xxx.amp.GradScaler class

Pull Request - State: closed - Opened by FFFrog about 2 months ago - 4 comments
Labels: module: cpu, open source, topic: not user facing

#164025 - [wip] aoti_torch_cpu_embedding_bag_byte_unpack

Pull Request - State: closed - Opened by hl475 about 2 months ago - 4 comments
Labels: module: cpu, fb-exported, release notes: quantization, ciflow/inductor, release notes: inductor (aoti), meta-exported

#164023 - [cuDNN][conv][64-bit] Disable cuDNN for 64-bit depthwise convs again

Pull Request - State: closed - Opened by pytorchbot about 2 months ago - 1 comment
Labels: module: cpu, open source

#163691 - Add attention benchmarking to operator microbenchmarks

Pull Request - State: closed - Opened by jainapurva about 2 months ago - 1 comment
Labels: oncall: distributed, module: cpu, release notes: quantization, release notes: releng, ciflow/mps, module: inductor, module: dynamo, ciflow/inductor, release notes: distributed (checkpoint), release notes: inductor (aoti), ciflow/h100-symm-mem, ciflow/vllm

#163601 - Change python grid calc for MTIA back to python mode

Pull Request - State: closed - Opened by nandesuka about 2 months ago - 15 comments
Labels: module: cpu, fb-exported, Merged, ciflow/trunk, topic: not user facing, module: inductor, ciflow/inductor, merging, meta-exported

#163591 - [PyTorch] Add SVE128 ISA (#158932)

Pull Request - State: closed - Opened by Nicoshev about 2 months ago - 1 comment
Labels: module: cpu, release notes: sparse, module: inductor, ciflow/inductor

#163581 - [cuDNN][Convolution] Disable cuDNN for 3D convolutions with kernel size != 1 for cuDNN 9.8+

Pull Request - State: closed - Opened by eqy about 2 months ago - 7 comments
Labels: module: cudnn, module: cuda, module: cpu, module: convolution, triaged, open source, ciflow/trunk, topic: bug fixes, release notes: cudnn, module: inductor, ciflow/inductor, merging

#163490 - [Inductor] support masked vectorization for the tail_loop for integer datatypes

Pull Request - State: open - Opened by jiayisunx about 2 months ago - 1 comment
Labels: module: cpu, open source, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor

#163268 - [PyTorch] Add SVE128 ISA (#158932)

Pull Request - State: closed - Opened by Nicoshev 2 months ago - 1 comment
Labels: module: cpu, release notes: sparse, module: inductor, ciflow/inductor

#163233 - [CPU] Support transpose and packing fusion for bit8

Pull Request - State: closed - Opened by Valentine233 2 months ago - 4 comments
Labels: module: cpu, open source, Merged, ciflow/trunk, topic: not user facing, merging

#163171 - [cuDNN][conv][64-bit] Disable cuDNN for 64-bit depthwise convs again

Pull Request - State: closed - Opened by eqy 2 months ago - 15 comments
Labels: module: cudnn, module: cpu, module: convolution, triaged, module: 64-bit, open source, Merged, ciflow/trunk, topic: not user facing, merging

#163039 - [CPU] Adding missing brackets in native MaxUnpool log

Pull Request - State: closed - Opened by can-gaa-hou 2 months ago - 8 comments
Labels: module: cpu, open source, Merged, ciflow/trunk, topic: not user facing, merging

#162776 - [MPS] Fix conv layout handling

Pull Request - State: closed - Opened by malfet 2 months ago - 3 comments
Labels: module: cpu, Merged, topic: bug fixes, release notes: mps, ciflow/mps, module: inductor, ciflow/inductor, merging

#162718 - Fix overflow in slow_conv3d when kernel size is too large.

Pull Request - State: closed - Opened by mansiag05 2 months ago - 6 comments
Labels: oncall: distributed, module: cpu, triaged, module: mkldnn, open source, module: amp (automated mixed precision), Merged, ciflow/trunk, release notes: quantization, topic: not user facing, module: inductor, module: dynamo, release notes: distributed (checkpoint), merging, module: compiled autograd, release notes: inductor (aoti)

#162683 - torch.matmul regression on CPU

Issue - State: closed - Opened by jiqing-feng 2 months ago - 4 comments
Labels: module: performance, module: cpu, triaged, module: linear algebra

#162603 - [release/2.4] add triton name

Pull Request - State: closed - Opened by ethanwee1 2 months ago - 1 comment
Labels: oncall: distributed, module: cpu, module: amp (automated mixed precision), NNC, release notes: releng, module: inductor, module: dynamo, release notes: distributed (checkpoint)

#162524 - Export d82049754

Pull Request - State: closed - Opened by Nicoshev 2 months ago - 2 comments
Labels: module: cpu, release notes: sparse, module: inductor, ciflow/inductor

#162363 - [Conv1d] Check overflow before we compute padding size.

Pull Request - State: open - Opened by thenumberouscode 2 months ago - 21 comments
Labels: module: nn, module: cpu, triaged, open source, release notes: nn

#162203 - Update round size with 1 division behavior

Pull Request - State: open - Opened by morrison-turnansky 3 months ago - 5 comments
Labels: oncall: distributed, module: cuda, module: cpu, triaged, open source, NNC, release notes: quantization, topic: not user facing, module: inductor, module: dynamo, release notes: distributed (checkpoint), module: compiled autograd, release notes: inductor (aoti)

#161785 - Inductor: dispatch GELU to oneDNN in CPU compile path and re-enable AArch64 fusions

Pull Request - State: closed - Opened by puneetmatharu 3 months ago - 3 comments
Labels: module: cpu, open source, module: inductor, release notes: inductor (aoti)

#161094 - Fix usage of forwarding references

Pull Request - State: open - Opened by lakshayg 3 months ago - 15 comments
Labels: oncall: jit, module: cpu, triaged, open source, Merged, NNC, Reverted, ciflow/trunk, release notes: quantization, ci-no-td

#161052 - [Intel GPU] Enable backward for SDPA XPU [Don't merge, test only]

Pull Request - State: closed - Opened by LuFinch 3 months ago - 2 comments
Labels: module: cpu, module: mkldnn, open source, ciflow/binaries_wheel, release notes: inductor (aoti)

#161049 - [feat]: add optimized exp_u20 implementation from Arm Optimized Routi…

Pull Request - State: open - Opened by Anallear 3 months ago - 16 comments
Labels: module: cpu, triaged, open source, module: arm, topic: not user facing

#161041 - Use to() instead of contiguous() to generate channels last tensor for Intel XPU

Pull Request - State: open - Opened by NeoZhangJianyu 3 months ago - 2 comments
Labels: module: cpu, open source

#161031 - Replace _device_t with torch.types.Device in torch/cpu/init.py

Pull Request - State: closed - Opened by grantinator 3 months ago - 4 comments
Labels: module: cpu, triaged, open source, Merged, ciflow/trunk, topic: not user facing, merging

#160959 - Using std::vector or c10::SmallVector instead of CArray

Pull Request - State: closed - Opened by FFFrog 3 months ago - 3 comments
Labels: oncall: distributed, module: cpu, open source, Merged, ciflow/trunk, release notes: quantization, topic: not user facing, merging

#160953 - Port 4 dynamo test files for the intel XPU

Pull Request - State: closed - Opened by shangerxin 3 months ago - 19 comments
Labels: oncall: distributed, module: cpu, open source, Merged, ciflow/trunk, release notes: quantization, topic: not user facing, module: inductor, module: dynamo, release notes: distributed (checkpoint), keep-going, merging, ciflow/xpu, module: xpu, release notes: inductor (aoti)

#160938 - [WOQ] Add XPU kernel for _weight_int8pack_mm

Pull Request - State: closed - Opened by xiaowangintel 3 months ago - 12 comments
Labels: module: cpu, open source, Merged, ciflow/trunk, merging, ciflow/xpu, release notes: inductor (aoti)

#160867 - [Intel GPU] Xpu matmul implementation for complex dtype

Pull Request - State: open - Opened by PawelSwider2000 3 months ago - 23 comments
Labels: module: cpu, open source, topic: not user facing, module: inductor, keep-going, ciflow/xpu

#160723 - Using std::make_unique<T>() instead of unique<T>(new T())

Pull Request - State: closed - Opened by FFFrog 3 months ago - 6 comments
Labels: oncall: distributed, oncall: jit, module: cpu, open source, Merged, ciflow/trunk, release notes: quantization, topic: not user facing, module: inductor, ciflow/inductor, merging

#160722 - Using std::make_unique<T>() instead of unique<T>(new T())

Pull Request - State: closed - Opened by FFFrog 3 months ago - 1 comment
Labels: oncall: distributed, oncall: jit, module: cpu, open source, module: amp (automated mixed precision), release notes: quantization, release notes: mps, ciflow/mps, module: inductor, module: dynamo, ciflow/inductor, release notes: distributed (checkpoint), release notes: inductor (aoti), ciflow/h100-symm-mem

#160606 - Use to() instead of contiguous() to generate channels last tensor for Intel XPU

Pull Request - State: closed - Opened by NeoZhangJianyu 3 months ago - 5 comments
Labels: module: cpu, open source, topic: not user facing, ciflow/xpu, module: xpu

#160524 - [Rocm7.1_internal_testing] Remove triton git commit from triton repo

Pull Request - State: closed - Opened by ethanwee1 3 months ago - 1 comment
Labels: oncall: distributed, module: rocm, module: cpu, release notes: releng, module: inductor, module: dynamo

#160496 - [Intel GPU] Remove SDPA FP32 implicit causal mask fallback [Don't merge]

Pull Request - State: closed - Opened by LuFinch 3 months ago - 3 comments
Labels: module: cpu, module: mkldnn, open source, ciflow/binaries_wheel

#160335 - Fix typo: 'complext'

Pull Request - State: closed - Opened by Raman-RH 3 months ago - 9 comments
Labels: module: cpu, open source, Merged, ciflow/trunk, topic: not user facing, merging

#160007 - Update export schema.py

Pull Request - State: closed - Opened by SherlockNoMad 3 months ago - 3 comments
Labels: module: cpu, fb-exported, release notes: quantization, ciflow/inductor

#159941 - [Intel GPU PT2E] Infer runtime out dtype based on dequant node in pattern

Pull Request - State: closed - Opened by ZhiweiYan-96 4 months ago - 5 comments
Labels: module: cpu, open source, release notes: quantization, module: inductor, ciflow/inductor, ciflow/xpu

#159859 - Added PyTorch LUT optimisation for GELU bf16 operators

Pull Request - State: open - Opened by leoken01 4 months ago - 4 comments
Labels: module: cpu, open source

#159778 - [Caffe2] Add float batch box cox SVE128 implementation

Pull Request - State: open - Opened by Nicoshev 4 months ago - 5 comments
Labels: module: cpu, fb-exported, release notes: sparse, module: inductor, ciflow/inductor

#159650 - [Release/2.7] pin requirements

Pull Request - State: closed - Opened by ethanwee1 4 months ago - 2 comments
Labels: oncall: distributed, oncall: jit, module: cpu, release notes: releng, module: inductor, module: dynamo

#159491 - [Dynamo][Better Engineering] Add typing annotations to guard and source (#158397)

Pull Request - State: closed - Opened by Lucaskabela 4 months ago - 5 comments
Labels: module: cpu, fb-exported, Merged, ciflow/trunk, topic: not user facing, fx, module: inductor, module: dynamo, ciflow/inductor, merging, release notes: export

#159462 - Inductor Fails on Conv1D After Permute with Stride Mismatch Error

Issue - State: closed - Opened by LiSsHhUuAaIi 4 months ago - 3 comments
Labels: module: cpu, triaged, oncall: pt2, module: fakeTensor, module: pt2-dispatcher

#159341 - [not for merging] Try out a fix

Pull Request - State: closed - Opened by janeyx99 4 months ago - 3 comments
Labels: module: cpu, module: inductor, ciflow/inductor

#159274 - [Caffe2] Build perfkernels targeting SVE128

Pull Request - State: open - Opened by Nicoshev 4 months ago - 6 comments
Labels: oncall: jit, module: cpu, fb-exported, release notes: quantization, module: inductor, ciflow/inductor

#158976 - Move some of vec into headeronly in preparation for Half.h

Pull Request - State: open - Opened by janeyx99 4 months ago - 6 comments
Labels: module: cpu, Merged, Reverted, ciflow/trunk, release notes: build, ciflow/inductor, ci-no-td

#158932 - Add SVE128 ISA

Pull Request - State: closed - Opened by Nicoshev 4 months ago - 95 comments
Labels: module: cpu, fb-exported, module: arm, Merged, ciflow/trunk, release notes: sparse, topic: improvements, module: inductor, ciflow/inductor, release notes: cpu (aarch64), meta-exported

#158919 - [ROCm][Windows] Fix finding ROCm/HIP version

Pull Request - State: closed - Opened by tvukovic-amd 4 months ago - 1 comment
Labels: oncall: distributed, module: rocm, module: cpu, module: mkldnn, release notes: releng, fx, module: inductor, module: dynamo, release notes: inductor (aoti)

#158842 - Skip slow tests for aarch64-inductor-benchmarks

Pull Request - State: closed - Opened by aditew01 4 months ago - 6 comments
Labels: module: cpu, open source, module: arm, Merged, ciflow/trunk, topic: not user facing, module: dynamo, ciflow/inductor, merging

#158730 - [BE] Fix extra-semi warnings

Pull Request - State: closed - Opened by malfet 4 months ago - 8 comments
Labels: oncall: distributed, module: cpu, Merged, NNC, release notes: quantization, topic: not user facing, merging

#158535 - [AOTI][CPU] Consider bias=None case for fbgemm_linear_fp16_weight

Pull Request - State: closed - Opened by hl475 4 months ago - 10 comments
Labels: module: cpu, fb-exported, Merged, ciflow/trunk, release notes: quantization, module: inductor, ciflow/inductor, merging, release notes: inductor (aoti)

#158445 - [cherry-pick] Fix AArch64 segfaults by disabling strict-aliasing in GridSamplerKernel for GCC 12 and above

Pull Request - State: closed - Opened by robert-hardwick 4 months ago - 4 comments
Labels: module: cpu, open source, module: arm, topic: not user facing, ciflow/linux-aarch64, arm priority

#158397 - [Dynamo][Better Engineering] Add typing annotations to guard and source

Pull Request - State: open - Opened by Lucaskabela 4 months ago - 13 comments
Labels: module: cpu, Merged, ciflow/trunk, topic: not user facing, module: inductor, module: dynamo, ciflow/inductor

#158280 - [release/2.5] upgrade3.13

Pull Request - State: closed - Opened by ethanwee1 4 months ago - 1 comment
Labels: oncall: distributed, module: cpu, release notes: releng, module: inductor, module: dynamo, release notes: distributed (checkpoint), release notes: inductor (aoti)

#158221 - [DO NOT MERGE] Enable MI355X ROCm CI testing.

Pull Request - State: closed - Opened by saienduri 4 months ago - 2 comments
Labels: oncall: distributed, module: rocm, module: cpu, open source, NNC, release notes: quantization, topic: not user facing, module: inductor, module: dynamo, release notes: distributed (checkpoint), keep-going, ci-no-td, release notes: inductor (aoti)

#158117 - Fix AArch64 segfaults by disabling strict-aliasing in GridSamplerKernel for GCC 12 and above

Pull Request - State: closed - Opened by robert-hardwick 4 months ago - 11 comments
Labels: module: cpu, open source, module: arm, Merged, ciflow/trunk, topic: not user facing, merging, ciflow/linux-aarch64, arm priority

#157984 - [caffe2] Enable auto vectorization

Pull Request - State: closed - Opened by Nicoshev 4 months ago - 6 comments
Labels: module: cpu, fb-exported, Merged, ciflow/trunk, topic: not user facing, merging

#157955 - [Typing] Improve typing in PyTorch

Pull Request - State: closed - Opened by migeed-z 4 months ago - 1 comment
Labels: oncall: distributed, oncall: jit, module: cpu, module: amp (automated mixed precision), fx, module: inductor, module: dynamo, ciflow/inductor, release notes: export

#157893 - [CPU] Support GQA for flash attention

Pull Request - State: closed - Opened by Valentine233 4 months ago - 3 comments
Labels: module: cpu, open source, Merged, ciflow/trunk, topic: not user facing

#157827 - [WIP] [Inductor][Intel GPU] Always use channel last for only freezing mode.Channel last

Pull Request - State: closed - Opened by etaf 4 months ago - 1 comment
Labels: module: rocm, module: cpu, module: mkldnn, release notes: onnx, ciflow/mps, module: inductor, module: dynamo, ciflow/inductor, ciflow/linux-aarch64, release notes: inductor (aoti)

#157686 - [BE][1/4] format shell scripts with `SHFMT`

Pull Request - State: open - Opened by XuehaiPan 5 months ago - 1 comment
Labels: module: cpu, module: ci, open source, better-engineering, release notes: quantization, topic: not user facing, module: dynamo, ciflow/inductor, module: devx, release notes: devx

#157685 - [BE] add `SHFMT` linter to format shell scripts

Pull Request - State: open - Opened by XuehaiPan 5 months ago - 5 comments
Labels: module: rocm, module: cpu, module: ci, open source, ciflow/binaries, ciflow/trunk, release notes: quantization, release notes: releng, skip-pr-sanity-checks, module: dynamo, ciflow/inductor, module: devx, release notes: devx

#157636 - [BE][2/6] fix typos in test/ (test/test_*.py)

Pull Request - State: open - Opened by XuehaiPan 5 months ago - 1 comment
Labels: module: cpu, module: mkldnn, open source, topic: not user facing, ciflow/mps, ciflow/linux-aarch64

#157554 - [BE][5/5] fix typos in aten/ (aten/src/ATen/)

Pull Request - State: open - Opened by XuehaiPan 5 months ago - 1 comment
Labels: module: cpu, open source, better-engineering, topic: not user facing

#157553 - [BE][4/5] fix typos in aten/ (aten/src/ATen/native/)

Pull Request - State: open - Opened by XuehaiPan 5 months ago - 1 comment
Labels: module: cpu, open source, better-engineering, release notes: foreach_frontend, topic: not user facing, ciflow/mps

#157552 - [BE][3/5] fix typos in aten/ (aten/src/ATen/native/)

Pull Request - State: open - Opened by XuehaiPan 5 months ago - 1 comment
Labels: module: cpu, open source, better-engineering, release notes: quantization, topic: not user facing

#157520 - Enable TF32 as fp32 internal precision for matmul/linear/conv

Pull Request - State: closed - Opened by yanbing-j 5 months ago - 6 comments
Labels: module: cpu, module: mkldnn, open source, Merged, ciflow/trunk, release notes: linalg_frontend, module: inductor, ciflow/inductor, merging, ciflow/rocm, ci-no-td, ciflow/linux-aarch64, ciflow/inductor-rocm, ciflow/rocm-mi300

#157520 - Enable TF32 as fp32 internal precision for matmul/linear/conv

Pull Request - State: open - Opened by yanbing-j 5 months ago - 1 comment
Labels: module: cpu, module: mkldnn, open source, ciflow/trunk, release notes: linalg_frontend, module: inductor, ciflow/inductor, ciflow/rocm, ci-no-td, ciflow/linux-aarch64, ciflow/inductor-rocm, ciflow/rocm-mi300

GitHub / pytorch/pytorch issues and pull requests