pytorch/pytorch issues and pull requests

#165391 - [user-streams] Add stream support to inductor wrapper codegen

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

#165155 - [ATen] Add reduction tag to reduction operators

Pull Request - State: open - Opened by eellison about 2 months ago - 2 comments
Labels: release notes: inductor

#164494 - [inductor] thread hint_override in more kernel args

Pull Request - State: closed - Opened by pianpwk 2 months ago - 3 comments
Labels: Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#164427 - [inductor] Add flag to run post-grad graph

Pull Request - State: closed - Opened by angelayi 2 months ago - 3 comments
Labels: module: inductor, ciflow/inductor, release notes: inductor

#164242 - [inductor] add a runtime assert for triton shapes

Pull Request - State: closed - Opened by isuruf 2 months ago - 6 comments
Labels: open source, Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

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

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

#163986 - fix flex attention eager: dont round down scores to low-precision (closes #163588)

Pull Request - State: closed - Opened by v0i0 2 months ago - 11 comments
Labels: Merged, ciflow/trunk, topic: bug fixes, module: inductor, ciflow/inductor, release notes: inductor, merging

#163747 - [Inductor] Check if profiling before using record_function in CompiledFxGraph

Pull Request - State: closed - Opened by azahed98 2 months ago - 8 comments
Labels: Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#163589 - [Inductor] deterministic mode

Pull Request - State: closed - Opened by shunting314 2 months ago - 16 comments
Labels: Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

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

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

#163467 - GEMM-template Horizontal

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

#163364 - [Opitmus] fix fp8 activation quatization for duplicates forward output

Pull Request - State: closed - Opened by mengluy0125 3 months ago - 7 comments
Labels: fb-exported, Merged, ciflow/trunk, ciflow/inductor, release notes: inductor, merging, meta-exported

#163324 - [Inductor] support masked vectorization for the tail_loop for fp8 datatype

Pull Request - State: open - Opened by jiayisunx 3 months ago - 5 comments
Labels: open source, Merged, Reverted, ciflow/trunk, ciflow/periodic, module: inductor, ciflow/inductor, release notes: inductor, ci-no-td

#163316 - [Inductor] support masked vectorization for the tail_loop for float64 datatype

Pull Request - State: open - Opened by jiayisunx 3 months ago - 6 comments
Labels: open source, Merged, Reverted, ciflow/trunk, ciflow/periodic, module: inductor, ciflow/inductor, release notes: inductor, ci-no-td

#163090 - [multi-kernel] shape-similarity kernel selection

Pull Request - State: closed - Opened by pianpwk 3 months ago - 3 comments
Labels: Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#162873 - [Graph Partition] allow sharing default device context

Pull Request - State: closed - Opened by BoyuanFeng 3 months ago - 11 comments
Labels: module: cuda graphs, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging, ci-no-td

#162761 - Fix windows path escape characters

Pull Request - State: closed - Opened by ghostspiders 3 months ago - 9 comments
Labels: triaged, open source, Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#162759 - [Inductor] Fix ComboKernels failing due to missing helper functions

Pull Request - State: closed - Opened by karthickai 3 months ago - 5 comments
Labels: Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#162696 - [CUDAGraph][UX] warn many times for rerecording from dynamic shapes

Pull Request - State: closed - Opened by BoyuanFeng 3 months ago - 3 comments
Labels: Merged, module: cuda graphs, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging, ci-no-td

#162688 - [Triton] [Inductor] Restrict subprocess autotuning to just Triton

Pull Request - State: closed - Opened by njriasan 3 months ago - 23 comments
Labels: fb-exported, Merged, Reverted, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging, ci-no-td, ciflow/h100, ciflow/h100-cutlass-backend

#162663 - [CUDAGraph][UX] error instead of warn for excessive re-record

Pull Request - State: closed - Opened by BoyuanFeng 3 months ago - 5 comments
Labels: module: cuda graphs, ciflow/trunk, module: inductor, module: dynamo, ciflow/inductor, release notes: inductor, ci-no-td

#162432 - [standalone_compile] binary format write should be atomic

Pull Request - State: closed - Opened by zou3519 3 months ago - 5 comments
Labels: Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#162309 - [inductor][triton] support static cuda launcher after triton # 7866

Pull Request - State: closed - Opened by davidberard98 3 months ago - 3 comments
Labels: Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#162275 - [inductor] require shape in TritonCSEVariable

Pull Request - State: closed - Opened by isuruf 3 months ago - 16 comments
Labels: open source, Merged, Reverted, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging, ci-no-td

#162192 - Add contiguous subgraph transformation threshold

Pull Request - State: closed - Opened by exclamaforte 3 months ago - 8 comments
Labels: Merged, ciflow/trunk, topic: new features, module: inductor, ciflow/inductor, release notes: inductor, merging, ciflow/rocm-mi300

#162106 - Add torch.compile support for triton.constexpr_function

Pull Request - State: closed - Opened by oulgen 3 months ago - 3 comments
Labels: Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#162017 - [inductor] performance model interface

Pull Request - State: closed - Opened by coconutruben 3 months ago - 9 comments
Labels: ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor

#161862 - [CUDAGraph] add config to error on skipping cudagraph

Pull Request - State: closed - Opened by BoyuanFeng 3 months ago - 3 comments
Labels: Merged, module: cuda graphs, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#161829 - [Inductor] update exp codegen for better precision

Pull Request - State: open - Opened by BoyuanFeng 3 months ago
Labels: ciflow/trunk, release notes: inductor

#161486 - [Quant][Inductor][CPU] add qlinear int8-mixed-bf16 patterns

Pull Request - State: closed - Opened by jiayisunx 3 months ago - 3 comments
Labels: open source, Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#161447 - [Inductor] Fix cross-device scalar lowering - cpu scalar with cuda tensor fails in torch.compile

Pull Request - State: closed - Opened by karthickai 3 months ago - 7 comments
Labels: Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#161439 - [Inductor] Fix cross-device scalar lowering (cpu scalar with cuda tensor fails in torch.compile)

Pull Request - State: closed - Opened by karthickai 3 months ago - 3 comments
Labels: module: inductor, ciflow/inductor, release notes: inductor

#161423 - [Pattern Matcher] improve error msg

Pull Request - State: closed - Opened by mlazos 3 months ago - 3 comments
Labels: Merged, ciflow/trunk, topic: not user facing, module: inductor, ciflow/inductor, release notes: inductor, merging

#160987 - [Optimus][decompose_mm] Fix BooleanAtom corner case

Pull Request - State: closed - Opened by mengluy0125 4 months ago - 5 comments
Labels: fb-exported, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#160933 - [dynamic shapes] use prims_common contiguity in create_example_tensors

Pull Request - State: open - Opened by pianpwk 4 months ago - 10 comments
Labels: fb-exported, Merged, Reverted, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, ci-no-td

#160894 - Fix meta function for aten.complex

Pull Request - State: closed - Opened by isuruf 4 months ago - 3 comments
Labels: open source, Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#160885 - [Triton] [Inductor] Add a Blackwell specific Template for persistent matmul

Pull Request - State: closed - Opened by njriasan 4 months ago - 3 comments
Labels: fb-exported, module: inductor, ciflow/inductor, release notes: inductor, Blackwell, ciflow/h100

#160792 - [Optimus] Add layernorm replacement pass

Pull Request - State: open - Opened by mengluy0125 4 months ago - 5 comments
Labels: fb-exported, Stale, module: inductor, ciflow/inductor, release notes: dynamo, release notes: inductor

#160711 - [BE][inductor] tl.dot(..., allow_tf32=...) -> tl.dot(..., input_precision=...)

Pull Request - State: closed - Opened by davidberard98 4 months ago - 6 comments
Labels: Merged, ciflow/trunk, topic: not user facing, module: inductor, ciflow/inductor, release notes: inductor, merging

#160677 - [Inductor] Add DeviceAssert op to enable device-side assertion in torch.compile

Pull Request - State: open - Opened by karthickai 4 months ago - 17 comments
Labels: Merged, ciflow/trunk, module: decompositions, module: inductor, ciflow/inductor, release notes: inductor

#160658 - [inductor][triton] Update triton_builtin handling after triton # 7239

Pull Request - State: closed - Opened by davidberard98 4 months ago - 3 comments
Labels: Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#160578 - [dynamic shapes] handle Max(*,1) for inductor layout contiguity

Pull Request - State: closed - Opened by pianpwk 4 months ago - 8 comments
Labels: fb-exported, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#160552 - [cutlass] fix dictionary iteration error

Pull Request - State: closed - Opened by mlazos 4 months ago - 10 comments
Labels: Merged, Reverted, ciflow/trunk, topic: not user facing, module: inductor, ciflow/inductor, release notes: inductor, ci-no-td

#160480 - [Triton] [Inductor] Enable TMA store for TMA mm templates

Pull Request - State: closed - Opened by njriasan 4 months ago - 70 comments
Labels: fb-exported, Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging, ciflow/h100, meta-exported

#160422 - [Inductor][Triton] Pass GPUTarget param to updated make_ir function

Pull Request - State: closed - Opened by NikhilAPatel 4 months ago - 4 comments
Labels: fb-exported, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#160408 - [FakeTensor] Supplement the relevant logic for converting conv1d to conv2d in meta_conv

Pull Request - State: closed - Opened by thenumberouscode 4 months ago - 22 comments
Labels: triaged, open source, Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#160227 - [FlexAttention] Swap from and to & for new triton

Pull Request - State: closed - Opened by drisspg 4 months ago - 3 comments
Labels: Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging, module: flex attention

#160181 - [do not merge][inductor] optimize welford reduction

Pull Request - State: closed - Opened by jiayisunx 4 months ago - 1 comment
Labels: open source, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor

#160087 - [cpp][inductor] Fix crash on bmm when input is used twice.

Pull Request - State: closed - Opened by frost-intel 4 months ago - 8 comments
Labels: open source, Merged, ciflow/trunk, topic: not user facing, module: inductor, ciflow/inductor, release notes: inductor, merging

#160000 - [triton_heuristics] Optimize the triton launcher in pt2

Pull Request - State: closed - Opened by davidberard98 4 months ago - 15 comments
Labels: Merged, Reverted, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging, ci-no-td

#159878 - Fix meta for constant_pad_nd

Pull Request - State: closed - Opened by isuruf 4 months ago - 8 comments
Labels: open source, Merged, ciflow/trunk, topic: bug fixes, module: inductor, ciflow/inductor, release notes: inductor, merging

#159810 - [inductor] remove no_x_dim

Pull Request - State: closed - Opened by davidberard98 4 months ago - 17 comments
Labels: Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#159760 - [Cutlass] Fix wrapper code generation breakage

Pull Request - State: closed - Opened by mlazos 4 months ago - 7 comments
Labels: Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#159450 - [Graph Partition] add graph partition doc

Pull Request - State: closed - Opened by BoyuanFeng 4 months ago - 3 comments
Labels: module: docs, Merged, module: cuda graphs, ciflow/trunk, topic: docs, topic: not user facing, module: inductor, release notes: inductor, merging, compile-docs

#159404 - Enable output padding when only outermost dim is dynamic

Pull Request - State: open - Opened by nandesuka 4 months ago - 42 comments
Labels: fb-exported, Merged, Reverted, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, ci-no-td

#159134 - [inductor] respect layout tags for ops with registered lowerings

Pull Request - State: closed - Opened by xmfan 4 months ago - 8 comments
Labels: Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#158998 - [inductor] Fix mm decomposition evaluating symints

Pull Request - State: open - Opened by azahed98 4 months ago
Labels: module: inductor, release notes: inductor

#158983 - [inductor] move all cpu scalars using pinned memory for graph partition (#155360)

Pull Request - State: open - Opened by v0i0 4 months ago - 8 comments
Labels: Merged, Reverted, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, ci-no-td

#158821 - [Optimus] Support decompose mm with dynamic shapes

Pull Request - State: closed - Opened by mengluy0125 5 months ago - 8 comments
Labels: fb-exported, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#158696 - [Optimus][Unit test] clean up the unit test

Pull Request - State: closed - Opened by mengluy0125 5 months ago - 5 comments
Labels: fb-exported, Merged, ciflow/trunk, topic: not user facing, module: inductor, ciflow/inductor, release notes: dynamo, release notes: inductor, merging

#158576 - Add stack trace to Inductor IR nodes if `inductor.config.trace.provenance_tracing=True`

Pull Request - State: closed - Opened by yushangdi 5 months ago - 6 comments
Labels: fb-exported, Merged, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

#158383 - Add better typing to avaialbe kernel options for flex attention

Pull Request - State: closed - Opened by drisspg 5 months ago - 6 comments
Labels: Merged, ciflow/trunk, topic: docs, release notes: inductor, merging, suppress-bc-linter, module: flex attention

#158231 - [inductor][cpu] Stop lowering div to reciprocal multiplication to preserve precision when the divisor is a scalar and device is on cpu

Pull Request - State: closed - Opened by thenumberouscode 5 months ago - 11 comments
Labels: triaged, open source, ciflow/trunk, module: inductor, ciflow/inductor, release notes: inductor, merging

GitHub / pytorch/pytorch issues and pull requests