pytorch/pytorch issues and pull requests

#165151 - Custom operator registration doesn't work when libtorch.so changes

Issue - State: open - Opened by Flamefire about 1 month ago - 4 comments
Labels: module: custom-operators, module: pt2-dispatcher

#163588 - Flex Attention Silently Incorrect with bfloat16 and Masks

Issue - State: closed - Opened by zaptrem about 2 months ago - 7 comments
Labels: triaged, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#162228 - Backpropagation to flex_attention `score_mod` biases fails based on presence of graph breaks

Issue - State: closed - Opened by stefan-baumann 3 months ago - 11 comments
Labels: triaged, module: correctness (silent), oncall: pt2, module: pt2-dispatcher, pt2: ubn

#161473 - flex attention is slower than sdpa

Issue - State: closed - Opened by LinB203 3 months ago - 4 comments
Labels: triaged, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#161077 - DISABLED test_override_meta (main.TestCustomOp)

Issue - State: closed - Opened by pytorch-bot[bot] 3 months ago - 2 comments
Labels: triaged, module: flaky-tests, module: custom-operators, skipped, oncall: pt2, module: pt2-dispatcher

#160955 - NotImplementedError: 64-bit indexing is not yet implemented for triton templates

Issue - State: closed - Opened by sirluk 3 months ago - 1 comment
Labels: triaged, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#160909 - compile with PrivateUse1 see tensors on "meta"

Issue - State: closed - Opened by zzzwen 3 months ago - 23 comments
Labels: triaged, module: dispatch, oncall: pt2, module: pt2-dispatcher, module: PrivateUse1

#160901 - [dynamo] traced autograd.Function silently incorrect due to bad requires_grad propagation

Issue - State: closed - Opened by xmfan 3 months ago
Labels: high priority, triaged, oncall: pt2, module: dynamo, module: higher order operators, module: pt2-dispatcher

#160840 - `torch.nn.functional.interpolate` generates inconsistent computation results after `torch.compile`

Issue - State: closed - Opened by dlibk 3 months ago - 5 comments
Labels: high priority, triaged, oncall: pt2, module: pt2-dispatcher

#160409 - test_flex_attention fails for float16 & bfloat16 dtypes on GPU

Issue - State: closed - Opened by Flamefire 3 months ago - 1 comment
Labels: triaged, oncall: pt2, module: pt2-dispatcher, module: flex attention

#160385 - TTGIR error for FlexAttention on B200

Issue - State: closed - Opened by drisspg 3 months ago - 2 comments
Labels: module: cuda, triaged, oncall: pt2, upstream triton, module: higher order operators, module: pt2-dispatcher, module: flex attention, Blackwell

#160333 - Tensor subclasses don't work with torch.library.triton_op

Issue - State: open - Opened by zou3519 3 months ago
Labels: tensor subclass, module: pt2-dispatcher

#160018 - Triton kernel generated for torch.compile(create_block_mask) using Flex Attention throws CUDA Illegal memory access

Issue - State: open - Opened by yubofredwang 3 months ago - 2 comments
Labels: module: crash, triaged, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#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

#159247 - [Flex Attention] Accuracy issue with kv length not multiple of kv block size

Issue - State: closed - Opened by Valentine233 4 months ago - 3 comments
Labels: triaged, oncall: pt2, module: inductor, module: higher order operators, module: pt2-dispatcher, module: flex attention

#158414 - received a FakeScriptObject input when dispatching DispatchKey.AutocastCUDA. but no python implementation is found

Issue - State: closed - Opened by asglover 4 months ago - 6 comments
Labels: triaged, module: custom-operators, oncall: pt2, module: pt2-dispatcher

#158081 - `torch.compile` errors with inductor raised `GuardOnDataDependentSymNode` exception

Issue - State: open - Opened by StrongerXi 4 months ago - 3 comments
Labels: triaged, oncall: pt2, module: dynamic shapes, module: inductor, module: pt2-dispatcher

#157833 - torch.compile with flex_attention: 'ShapeAsConstantBuffer' object has no attribute 'dtype'

Issue - State: closed - Opened by s22chan 4 months ago - 3 comments
Labels: triaged, oncall: pt2, module: inductor, module: higher order operators, module: pt2-dispatcher, module: flex attention

#157675 - Cannot create a mask for each sequence in a batch with Flex Attention

Issue - State: closed - Opened by luispl77 5 months ago - 3 comments
Labels: oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#157543 - Vmap error raised by mask_mod of FlexAttention

Issue - State: closed - Opened by NiuMa-1234 5 months ago - 2 comments
Labels: triaged, module: vmap, oncall: pt2, module: functorch, module: higher order operators, module: pt2-dispatcher, module: flex attention

#157276 - several `transformers` tests fail with `torch 2.8 RC` but pass with `torch 2.7.1` on `T4` (but both pass on `A10`)

Issue - State: open - Opened by ydshieh 5 months ago - 3 comments
Labels: high priority, triage review, module: cuda, oncall: pt2, module: inductor, module: higher order operators, module: pt2-dispatcher, module: flex attention

#157104 - Document AOTAutograd autocast divergence from eager behavior and provide a workaround

Issue - State: closed - Opened by zou3519 5 months ago - 1 comment
Labels: triaged, oncall: pt2, module: pt2-dispatcher

#156688 - [CPU][flex attention] Llama 3 failed on CPU with PyTorch 2025-06-22 nightly wheel

Issue - State: closed - Opened by LifengWang 5 months ago
Labels: needs reproduction, triaged, module: regression, oncall: pt2, module: higher order operators, oncall: cpu inductor, module: pt2-dispatcher, module: flex attention

#156322 - Provide a way to allow dynamo to trace into an operator defined with `torch.library.custom_op`

Issue - State: closed - Opened by fxmarty-amd 5 months ago - 3 comments
Labels: triaged, module: custom-operators, oncall: pt2, module: pt2-dispatcher

#155721 - Add runtime profiler info for AOTDispatcher prologue

Issue - State: closed - Opened by bdhirsh 5 months ago
Labels: triaged, oncall: pt2, module: pt2-dispatcher

#155386 - OpOverloads should have annotations

Issue - State: open - Opened by zou3519 6 months ago - 1 comment
Labels: triaged, module: custom-operators, module: dispatch, oncall: pt2, module: pt2-dispatcher

#155385 - [custom_op] Custom ops created by @custom_op should get type hints propagated to their OpOverload

Issue - State: open - Opened by zou3519 6 months ago - 1 comment
Labels: triaged, module: custom-operators, oncall: pt2, module: pt2-dispatcher, internal ramp-up task

#154820 - In-place operations are reordered across the forward-backward in autograd function

Issue - State: closed - Opened by zxfseserer 6 months ago - 6 comments
Labels: triaged, oncall: pt2, module: dynamo, module: pt2-dispatcher, ubn

#154554 - flex_attention + NJT output inconsistent with non-NJT results

Issue - State: open - Opened by mauriceweiler 6 months ago - 2 comments
Labels: triaged, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#154111 - DTensor + torch.compile on CPU: compiled matmul fails with multiple shape inputs

Issue - State: closed - Opened by kberat-synnada 6 months ago - 5 comments
Labels: oncall: distributed, triaged, oncall: pt2, module: aotdispatch, module: dtensor, module: pt2-dispatcher

#153993 - torch.compile custom autograd.Function + dynamic shapes + torch.tensor constructor in backward can cause crash

Issue - State: closed - Opened by bdhirsh 6 months ago - 2 comments
Labels: high priority, oncall: pt2, module: pt2-dispatcher

#153763 - DISABLED test_impl_device_cpu (main.TestCustomOp)

Issue - State: closed - Opened by pytorch-bot[bot] 6 months ago - 1 comment
Labels: triaged, module: flaky-tests, module: custom-operators, skipped, oncall: pt2, module: pt2-dispatcher

#152593 - Flex Attention doesn't scale with custom bias

Issue - State: open - Opened by danjenson 7 months ago - 2 comments
Labels: triaged, oncall: pt2, module: pt2-dispatcher, module: flex attention

#152309 - Softmax Decomp Causes Incorrect Gradients when Using `torch.compile` with `F.multi_head_attention_forward`

Issue - State: closed - Opened by defaultd661 7 months ago - 14 comments
Labels: high priority, triaged, module: correctness (silent), oncall: pt2, module: decompositions, module: aotdispatch, module: pt2-dispatcher, module: sdpa, ubn

#152297 - Flex attention: batch-index-dependent block mask causes error with changing batch size

Issue - State: closed - Opened by zhihanyang2022 7 months ago - 1 comment
Labels: triaged, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#151893 - DISABLED test_builtin_score_mods_different_block_size_float32_score_mod0_BLOCK_SIZE3_cuda_float32 (main.TestFlexAttentionCUDA)

Issue - State: closed - Opened by pytorch-bot[bot] 7 months ago - 3 comments
Labels: module: flaky-tests, skipped, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#151892 - DISABLED test_builtin_score_mods_different_block_size_float32_score_mod3_BLOCK_SIZE2_cuda_float32 (main.TestFlexAttentionCUDA)

Issue - State: closed - Opened by pytorch-bot[bot] 7 months ago - 2 comments
Labels: module: flaky-tests, skipped, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#151891 - DISABLED test_skip_odd_keys_bfloat16_cuda_bfloat16 (main.TestFlexAttentionCUDA)

Issue - State: closed - Opened by pytorch-bot[bot] 7 months ago - 1 comment
Labels: module: flaky-tests, skipped, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#151889 - DISABLED test_builtin_score_mods_dynamic_float16_score_mask_mod7_cuda_float16 (main.TestFlexAttentionCUDA)

Issue - State: closed - Opened by pytorch-bot[bot] 7 months ago - 3 comments
Labels: module: flaky-tests, skipped, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#151365 - flex_attention error in torch.compile

Issue - State: closed - Opened by jjh42 7 months ago - 2 comments
Labels: oncall: pt2, module: pt2-dispatcher, module: flex attention

#151301 - DISABLED test_fake_registration (main.TestOpProfiles)

Issue - State: closed - Opened by pytorch-bot[bot] 7 months ago - 14 comments
Labels: triaged, module: flaky-tests, module: custom-operators, skipped, oncall: pt2, module: pt2-dispatcher

#151281 - DISABLED test_duplicate_registration_impl (main.TestOpProfiles)

Issue - State: open - Opened by pytorch-bot[bot] 7 months ago - 13 comments
Labels: triaged, module: flaky-tests, module: custom-operators, skipped, oncall: pt2, module: pt2-dispatcher

#151188 - Make auto_functionalize HOPs cacheable

Issue - State: closed - Opened by zou3519 7 months ago
Labels: oncall: pt2, module: higher order operators, module: pt2-dispatcher, dynamo-triage-jan2025

#150056 - FlexAttention inductor tensor has no attribute `get_dtype`

Issue - State: closed - Opened by tsengalb99 8 months ago - 2 comments
Labels: needs reproduction, triaged, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#149960 - attn_implementation="eager" Buggy on Blackwell

Issue - State: closed - Opened by Oseltamivir 8 months ago - 4 comments
Labels: high priority, triage review, needs reproduction, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#149801 - `INTERNAL ASSERT FAILED` in `torch.func.vmap` and `torch.scatter_add`

Issue - State: closed - Opened by vwrewsge 8 months ago - 4 comments
Labels: triaged, module: vmap, oncall: pt2, module: functorch, module: pt2-dispatcher

#149789 - flex_attention create_block_mask() + inductor: integer division or modulo by zero

Issue - State: closed - Opened by rmmr 8 months ago - 1 comment
Labels: triaged, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#149707 - [aot] mark dynamic activations as maybe dynamic

Pull Request - State: closed - Opened by xmfan 8 months ago - 7 comments
Labels: Merged, topic: not user facing, module: inductor, module: dynamo, ciflow/inductor, module: pt2-dispatcher

#149707 - [aot] mark dynamic activations as maybe dynamic

Pull Request - State: open - Opened by xmfan 8 months ago - 5 comments
Labels: topic: not user facing, module: inductor, module: dynamo, ciflow/inductor, module: pt2-dispatcher

#149617 - Torch nightly `torch-2.8.0.dev20250320` breaks torchcodec

Issue - State: closed - Opened by NicolasHug 8 months ago - 1 comment
Labels: module: custom-operators, bug, oncall: pt2, module: pt2-dispatcher

#149468 - torch.library.opcheck doesn't check strides for CPU Tensors

Issue - State: open - Opened by zou3519 8 months ago - 2 comments
Labels: high priority, module: cpp-extensions, triaged, module: custom-operators, module: library, module: pt2-dispatcher

#149105 - [inductor] post grad graph with scatter_upon_const_tensor lowering is not runnable

Issue - State: closed - Opened by xmfan 8 months ago - 2 comments
Labels: triaged, oncall: pt2, module: aotdispatch, module: pt2-dispatcher

#149006 - Missing Additional documentation in autograd.py

Issue - State: closed - Opened by Jason1ien 8 months ago - 3 comments
Labels: module: docs, triaged, module: library, oncall: pt2, module: pt2-dispatcher

#148977 - module.cuda() doesn't work under FakeTensorMode

Issue - State: open - Opened by bdhirsh 8 months ago - 4 comments
Labels: module: nn, triaged, oncall: pt2, module: fakeTensor, module: pt2-dispatcher

#148889 - DISABLED test_make_closure_dynamic_shapes (main.DynamicShapesHigherOrderOpTests)

Issue - State: closed - Opened by pytorch-bot[bot] 8 months ago - 3 comments
Labels: triaged, module: flaky-tests, skipped, module: higher order operators, module: pt2-dispatcher

#148835 - [Inductor] Error detected in ReluBackward0

Issue - State: closed - Opened by Cookiee235 9 months ago - 1 comment
Labels: oncall: pt2, module: aotdispatch, module: pt2-dispatcher

#148827 - Inductor may permute inputs to flex attention, leading to assertion error

Issue - State: closed - Opened by Aleko2286 9 months ago - 6 comments
Labels: triaged, oncall: pt2, module: inductor, module: higher order operators, module: pt2-dispatcher, module: flex attention

#148711 - Flex attention significantly slower than SDPA

Issue - State: closed - Opened by nikonikolov 9 months ago - 5 comments
Labels: triaged, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention, module: sdpa

#148701 - aot_eager produces wrong output with all_gather_tensor_autograd

Issue - State: open - Opened by eellison 9 months ago - 2 comments
Labels: high priority, oncall: distributed, triaged, actionable, module: correctness (silent), oncall: pt2, module: pt2-dispatcher

#148356 - Inductor layout constraints for custom operators changed from 2.5->2.6, breaking BC

Issue - State: closed - Opened by zou3519 9 months ago
Labels: module: custom-operators, oncall: pt2, module: pt2-dispatcher

#148244 - [inductor] `nn.Upsample-torch.linalg.lu_factor` outputs inconsistent results with eager

Issue - State: closed - Opened by shaoyuyoung 9 months ago - 4 comments
Labels: high priority, triaged, module: correctness (silent), oncall: pt2, module: aotdispatch, module: pt2-dispatcher

#148094 - DISABLED test_split_dynamic (main.AutoFunctionalizeTests)

Issue - State: closed - Opened by pytorch-bot[bot] 9 months ago - 6 comments
Labels: triaged, module: flaky-tests, skipped, module: functionalization, oncall: pt2, module: pt2-dispatcher

#148080 - DISABLED test_split (main.AutoFunctionalizeTests)

Issue - State: closed - Opened by pytorch-bot[bot] 9 months ago - 7 comments
Labels: triaged, module: flaky-tests, skipped, module: functionalization, oncall: pt2, module: pt2-dispatcher

#148067 - DISABLED test_slice_dynamic (main.AutoFunctionalizeTests)

Issue - State: closed - Opened by pytorch-bot[bot] 9 months ago - 5 comments
Labels: high priority, triaged, module: flaky-tests, skipped, module: functionalization, oncall: pt2, module: pt2-dispatcher

#148035 - DISABLED test_slice (main.AutoFunctionalizeTests)

Issue - State: closed - Opened by pytorch-bot[bot] 9 months ago - 6 comments
Labels: triaged, module: flaky-tests, skipped, module: functionalization, oncall: pt2, module: pt2-dispatcher

#147822 - `AssertionError: Mixing fake modes NYI` in FakeTensorMode context

Issue - State: closed - Opened by default1360 9 months ago - 2 comments
Labels: triaged, oncall: pt2, module: fakeTensor, module: pt2-dispatcher

#147701 - Compiled `flex_attention` assuming wrong output tensor shape

Issue - State: closed - Opened by mauriceweiler 9 months ago - 5 comments
Labels: triaged, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#147565 - [dynamo][checkpoint] non-reentrant checkpoint + ambient saved tensor hooks is silently incorrect

Issue - State: closed - Opened by xmfan 9 months ago - 1 comment
Labels: module: activation checkpointing, triaged, oncall: pt2, module: dynamo, module: higher order operators, module: pt2-dispatcher

#147551 - FlexAttention compiled has illegal memory access or device-side assert even though all tensors are contiguous

Issue - State: open - Opened by leijurv 9 months ago - 6 comments
Labels: triaged, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#147468 - [Triton] [Upstream] FlexDecoding Test Failures opToStageAndCluster[&op].first < numStages && "Op with invalid stage!

Issue - State: closed - Opened by drisspg 9 months ago - 6 comments
Labels: triaged, oncall: pt2, upstream triton, module: higher order operators, module: pt2-dispatcher, module: flex attention

#147463 - Performance regression on modded-nanogpt torch-2.7.0.dev20250208→torch-2.7.0.dev20250209

Issue - State: closed - Opened by YouJiacheng 9 months ago - 16 comments
Labels: triaged, oncall: pt2, module: inductor, module: higher order operators, module: pt2-dispatcher, module: flex attention

#147405 - dynamo should recompile with constant tensors that use ambient device guards

Issue - State: closed - Opened by jamesjwu 9 months ago - 3 comments
Labels: triaged, actionable, oncall: pt2, module: aotdispatch, module: pt2-dispatcher, dynamo-must-fix

#147336 - Investigate FlexAttention performance degradation on low precision inputs

Issue - State: closed - Opened by danielvegamyhre 9 months ago - 4 comments
Labels: triaged, oncall: pt2, upstream triton, module: higher order operators, module: pt2-dispatcher, module: flex attention

#147268 - flex_attention throws `CUDA error: an illegal memory access was encountered`

Issue - State: closed - Opened by mauriceweiler 9 months ago - 1 comment
Labels: triaged, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#146686 - Unexpected specialization during estimate_runtime

Issue - State: closed - Opened by laithsakka 10 months ago - 3 comments
Labels: triaged, oncall: pt2, module: dynamic shapes, module: aotdispatch, module: pt2-dispatcher

#146624 - [Flex Attention] Cannot determine truth value of Relational

Issue - State: closed - Opened by alexdremov 10 months ago - 2 comments
Labels: triaged, oncall: pt2, module: pt2-dispatcher, module: flex attention

#146212 - [aot_eager] retain_grad is ignored

Issue - State: closed - Opened by anijain2305 10 months ago - 2 comments
Labels: high priority, triage review, triaged, oncall: pt2, module: pt2-dispatcher

#145481 - [custom ops] [2.7 nightly] custom ops with typing.List breaks when importing annotations from future

Issue - State: closed - Opened by dabeschte 10 months ago - 3 comments
Labels: high priority, triaged, module: custom-operators, oncall: pt2, module: pt2-dispatcher

#145460 - Flex Attention not support score_mod with gradients

Issue - State: closed - Opened by LoserCheems 10 months ago - 2 comments
Labels: triaged, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#145081 - partitioner hangs for some long chains of ops with many users

Issue - State: closed - Opened by bdhirsh 10 months ago
Labels: high priority, triaged, oncall: pt2, module: aotdispatch, module: pt2-dispatcher

#144912 - DISABLED test_flex_attention (main.TestCompiledAutograd)

Issue - State: closed - Opened by jeffdaily 10 months ago - 1 comment
Labels: module: rocm, triaged, skipped, oncall: pt2, module: higher order operators, module: compiled autograd, module: pt2-dispatcher, module: flex attention

#144376 - torch.compile post_accumulate_grad_hook ordering is wrong for tiebreakers

Issue - State: open - Opened by xmfan 11 months ago
Labels: triaged, oncall: pt2, module: aotdispatch, module: pt2-dispatcher

#144142 - [torch.compile] Errors on autograd.Function forward returns non-Tensor

Issue - State: closed - Opened by yanboliang 11 months ago
Labels: triaged, oncall: pt2, module: higher order operators, module: pt2-dispatcher, dynamo-autograd-function

#144095 - activation memory budget partitioner can fail with unbacked symints

Issue - State: closed - Opened by bdhirsh 11 months ago - 2 comments
Labels: high priority, triaged, oncall: pt2, module: dynamic shapes, module: aotdispatch, module: pt2-dispatcher

#144072 - Compile error for custom op with optional mutable tensor list argument

Issue - State: closed - Opened by jerrychenhf 11 months ago - 3 comments
Labels: triaged, module: custom-operators, module: functionalization, oncall: pt2, module: aotdispatch, module: pt2-dispatcher

#143832 - flex_attention: OutOfResources

Issue - State: closed - Opened by rmmr 11 months ago - 1 comment
Labels: triaged, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#143788 - Flex attention with nested tensors, bug in `create_nested_block_mask`

Issue - State: closed - Opened by VivekPanyam 11 months ago - 2 comments
Labels: triaged, module: nestedtensor, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#143192 - onnx export with dynamo using torch.cond for dynamic models throws error

Issue - State: closed - Opened by ionymikler 11 months ago - 3 comments
Labels: module: onnx, triaged, oncall: pt2, module: higher order operators, export-triage-review, oncall: export, module: pt2-dispatcher

#143180 - Dynamo graph break on FlexAttentionBackwardHOP

Issue - State: closed - Opened by xmfan 11 months ago - 2 comments
Labels: triaged, oncall: pt2, module: higher order operators, module: compiled autograd, module: pt2-dispatcher, module: flex attention

#143163 - Cannot compile a block that contains Flex attention without graph breaks

Pull Request - State: closed - Opened by nicolas-dufour 11 months ago - 2 comments
Labels: triaged, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#143053 - test_custom_scan_op regression

Issue - State: closed - Opened by eellison 11 months ago - 5 comments
Labels: triaged, bug, oncall: pt2, module: dynamic shapes, module: higher order operators, module: pt2-dispatcher

#142817 - Significant Backward Pass Scaling Bottleneck in Flex Attention: 12x Slower than Forward Pass at 32K Sequence Length

Issue - State: closed - Opened by cora-codes 11 months ago - 6 comments
Labels: module: performance, triaged, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

#142367 - [Break XPU][Inductor UT] The cpu test case `TestFlexAttention::test_GQA_float32_score_mod6` failed on XPU server.

Issue - State: closed - Opened by etaf 12 months ago - 2 comments
Labels: module: cpu, triaged, oncall: pt2, module: higher order operators, module: pt2-dispatcher, module: flex attention

GitHub / pytorch/pytorch issues and pull requests