103 Pull requests merged by 36 people

• Fix static attention non-HF RoPE implementation

  #11808 merged Jun 20, 2025

• Make the requant pass call the per_tensor overload

  #11789 merged Jun 19, 2025

• Arm backend: Remove hard coded TOSA profile in VGF backend

  #11818 merged Jun 19, 2025

• Arm backend: Fix bug of inserting unnecessary casts for aten.where.self

  #11816 merged Jun 19, 2025

• Arm backend: Test temp memory allocation return code in backend

  #11814 merged Jun 19, 2025

• Arm backend: Improve pooling args handling

  #11819 merged Jun 19, 2025

• Arm backend: Add support for aten.round

  #11813 merged Jun 19, 2025

• Arm backend: Add support for grouped convolution

  #11817 merged Jun 19, 2025

• Arm backend: Add unittest for MHA

  #11812 merged Jun 19, 2025

• Fix wheel build CI jobs by installing torchvision

  #11806 merged Jun 19, 2025

• Temporarily fix Moshi test

  #11799 merged Jun 19, 2025

• Static attention Python I/O manager

  #11763 merged Jun 19, 2025

• dtype selective build from model API in OSS

  #11760 merged Jun 18, 2025

• Value description

  #11798 merged Jun 18, 2025

• Tensor description

  #11792 merged Jun 18, 2025

• Arm backend: Change QAT weight observer

  #11787 merged Jun 18, 2025

• [ET-VK][Ops] enabling double support for quantization and dequantization ops

  #11795 merged Jun 18, 2025

• Add legacy mode test

  #11492 merged Jun 18, 2025

• Improve executor_runner cmake instructions

  #11773 merged Jun 18, 2025

• [ET-VK][Ops] enabling double support for quantization and dequantization ops

  #11553 merged Jun 18, 2025

• Propagate core_aten_exceptions to quantize_and_export_to_executorch and export_to_executorch.

  #11772 merged Jun 18, 2025

• Re-introduce the type-erasing tensor class.

  #11790 merged Jun 18, 2025

• Update CODEOWNERS

  #11785 merged Jun 18, 2025

• Set pyre-strict for passes unit tests.

  #11740 merged Jun 18, 2025

• Use MAP_SHARED to allow sharing memory between processing

  #11733 merged Jun 18, 2025

• Arm backend: Prevent illegal fusion in FuseEqualPlaceholdersPass

  #11781 merged Jun 18, 2025

• Introduce extension/llm/export_llm

  #11746 merged Jun 18, 2025

• Support Span to construct from a single element similar to ArrayRef

  #11778 merged Jun 18, 2025

• [llm] Update metadata max_seq_len based on the max range of dynamic shapes

  #11611 merged Jun 18, 2025

• Support Span to construct from a single element similar to ArrayRef

  #11757 merged Jun 18, 2025

• [llm] Fix start_pos not being updated in prefill_chunk()

  #11592 merged Jun 18, 2025

• [XNNPACK Quantizer] Select between TConvs and Convs

  #11732 merged Jun 18, 2025

• [Quantized DeConv Support] Dynamically Quantized Deconvolutions with groups ==1

  #11731 merged Jun 18, 2025

• [Quantized DeConv Support] Enable Quantized Transposed Convs with groups==1

  #11730 merged Jun 18, 2025

• skip et quantizer numeric debugging tests for infra update

  #11764 merged Jun 17, 2025

• [ET-VK][Ops] choose_qparams op shaders and impl

  #11769 merged Jun 17, 2025

• [ET-VK][Ops] dequantization op shaders and impl

  #11768 merged Jun 17, 2025

• [ET-VK][Ops] quantization op shaders and impl

  #11767 merged Jun 17, 2025

• [ET-VK][Ops] choose_qparams op shaders and impl

  #11557 merged Jun 17, 2025

• [ET-VK][Ops] dequantization op shaders and impl

  #11483 merged Jun 17, 2025

• [ET-VK][Ops] quantization op shaders and impl

  #11369 merged Jun 17, 2025

• Reapply D74208085: "Switch fbcode builds of ExecuTorch and PyTorch to fbsource sleef (#11261)"

  #11765 merged Jun 17, 2025

• Reapply D74208085: "Switch fbcode builds of ExecuTorch and PyTorch to fbsource sleef (#11261)"

  #11657 merged Jun 17, 2025

• Refactor XNNPACK tester to extract delegate-independent tester classes

  #11596 merged Jun 17, 2025

• Add MSE numerical comparator

  #11759 merged Jun 17, 2025

• Qualcomm AI Engine Direct - Deprecate convert_bmm_to_matmul pass

  #11392 merged Jun 17, 2025

• Qualcomm AI Engine Direct - GA Albert, Bert, Distilbert, Eurobert

  #11546 merged Jun 17, 2025

• Use wrappers from xnnpack.h for unary and binary ops (#11584)

  #11666 merged Jun 17, 2025

• Fix text_llm_runner unit test

  #11750 merged Jun 17, 2025

• Add numerical comparator base class and L1 comparator

  #11751 merged Jun 17, 2025

• Arm backend: Fix bug in decompose linear vector norm

  #11755 merged Jun 17, 2025

• Arm backend: Add support for per-channel quantization

  #11752 merged Jun 17, 2025

• Arm backend: Added decomposition for MaxPool2d with dilation > 0.

  #11724 merged Jun 17, 2025

• [llm] Add generate_from_pos API to LLM runner

  #11570 merged Jun 17, 2025

• Add function for input preprocessing in numerical comparator

  #11739 merged Jun 17, 2025

• Update documents for buffer allocator

  #11467 merged Jun 17, 2025

• Fix for overriding installed executorch when running optimum-et models

  #11745 merged Jun 17, 2025

• Fixed linter

  #11742 merged Jun 16, 2025

• Implement ReplaceMulTensorWithMulAndFullOpsPass.

  #11577 merged Jun 16, 2025

• Update CMakeLists.txt for extension/apple to strip debug symbols path

  #11659 merged Jun 16, 2025

• Revert vulkan changes from D76646172 fixup patch

  #11727 merged Jun 16, 2025

• Fix channels_last_tagged_reshape_pass to handle mixed memory format tuple outputs

  #11647 merged Jun 16, 2025

• Added evaluation script for qualcomm LlamaModel

  #11663 merged Jun 16, 2025

• Qualcomm AI Engine Direct - GA Model Enablement (Roberta)

  #11354 merged Jun 16, 2025

• Qualcomm AI Engine Direct - Enable Lookahead Decoding

  #11437 merged Jun 16, 2025

• Fix lint

  #11728 merged Jun 16, 2025

• Install examples and domain libraries as last (optional) step

  #11653 merged Jun 16, 2025

• Define ET_HAS_EXCEPTIONS macro

  #11664 merged Jun 16, 2025

• lintrunner init can fail, don't suppress stderr

  #11667 merged Jun 16, 2025

• Support for cmdline flag to toggle Debug vs Release build types in cmake

  #11662 merged Jun 16, 2025

• Remove buck-out on clean ET install

  #11641 merged Jun 16, 2025

• Add et_dump related API to module

  #9284 merged Jun 14, 2025

• [ET-VK][Ops] choose_qparams_per_token_asymmetric.default test setup

  #11556 merged Jun 14, 2025

• [ET-VK][Ops] choose_qparams.tensor test setup

  #11555 merged Jun 14, 2025

• [ET-VK][Ops] choose_qparams ops skeleton test framework

  #11554 merged Jun 14, 2025

• [ET-VK][Ops] dequantize_per_token.default test setup

  #11482 merged Jun 14, 2025

• [ET-VK][Ops] dequantize_per_tensor.default test setup

  #11481 merged Jun 14, 2025

• [ET-VK][Ops] dequantize ops skeleton test framework

  #11480 merged Jun 14, 2025

• [ET-VK][Ops] quantize_per_tensor.default test setup

  #11368 merged Jun 14, 2025

• [ET-VK][Ops] quantize_per_token.default test setup

  #11367 merged Jun 14, 2025

• [ET-VK][Ops] quantize ops skeleton test framework

  #11366 merged Jun 14, 2025

• [ET-VK][Ops] common test utils for converting aten types to vulkan types

  #11575 merged Jun 14, 2025

• [ET-VK] double, short, and uint16 dtype runtime support

  #11365 merged Jun 14, 2025

• [ET] enabling half dtype output for dequantization and making logic consistent

  #11552 merged Jun 14, 2025

• [ET] enabling half dtype input for quantization

  #11479 merged Jun 14, 2025

• Reapply "Add vectorized_math.h (#11204)", "Add optimized_portable_kernels test (#11205)", and "Add vectorization in elementwise_util (#9432)"

  #11604 merged Jun 14, 2025

• Define ET_HAS_EXCEPTIONS macro

  #11603 merged Jun 14, 2025

• [cadence][aot]Implement mul.Tensor to quant fusion.

  #11580 merged Jun 13, 2025

• [ET-VK][ez] Allow Vulkan operator tests to run on devserver and Mac

  #11652 merged Jun 13, 2025

• [ET-VK][ez] Allow Vulkan operator tests to run on devserver and Mac

  #11648 merged Jun 13, 2025

• Qualcomm AI Engine Direct - Enable custom operator

  #8726 merged Jun 13, 2025

• Mapping between runtime and aot intermediate outputs

  #11624 merged Jun 13, 2025

• Fix flags in apple_common.cmake

  #11643 merged Jun 13, 2025

• [ET-VK][ez] Fix type ambiguity relating to TextureLimits class

  #11646 merged Jun 13, 2025

• [ET-VK][ez] Fix type ambiguity relating to TextureLimits class

  #11642 merged Jun 13, 2025

• [Exutorch] Add broadcast support for le op

  #11635 merged Jun 13, 2025

• Use GraphBuilder in test_replace_ops_passes. #3

  #11501 merged Jun 13, 2025

• revert NXP backend: Enable initial unit tests workflow

  #11612 merged Jun 13, 2025

• Arm backend: Fix warnings from cppcheck in runtime

  #11639 merged Jun 13, 2025

• Arm backend: Test and track resnet18 and resnet50 models in CI

  #11638 merged Jun 13, 2025

• Arm backend: Add embedding.default + index_select

  #11636 merged Jun 13, 2025

• Arm backend: Add validate_valid_dtype

  #11472 merged Jun 13, 2025

• fix platform inconsistencies

  #11547 merged Jun 13, 2025

46 Pull requests opened by 25 people

• [draft] Change macos runner

  #11644 opened Jun 13, 2025

• [llm] Add sentencepiece tokenizer support to llm runner

  #11645 opened Jun 13, 2025

• Implement load_into for Mmap Data Loader

  #11654 opened Jun 13, 2025

• Enable kleidiAI in llama runner

  #11658 opened Jun 13, 2025

• Use elementwise_util instead of functional_util in op_neg

  #11660 opened Jun 13, 2025

• Revert D76405701

  #11661 opened Jun 13, 2025

• Define ET_HAS_EXCEPTIONS macro

  #11681 opened Jun 14, 2025

• Reapply "Add vectorized_math.h (#11204)", "Add optimized_portable_kernels test (#11205)", and "Add vectorization in elementwise_util (#9432)"

  #11682 opened Jun 14, 2025

• Only support int8 dtype for quant operators

  #11685 opened Jun 14, 2025

• [WIP] Add unit tests for getting the delegated payload in exir/backend.

  #11699 opened Jun 15, 2025

• Implement load_into for SharedPtrDataLoader and add test (#11562)

  #11707 opened Jun 15, 2025

• Add script to fetch benchmark results for execuTorch

  #11734 opened Jun 16, 2025

• deduplicate torch ao debugger tests between pytorch/ao and ExecuTorch

  #11735 opened Jun 16, 2025

• Revert D76405701

  #11736 opened Jun 16, 2025

• [ExecuTorch][#10447] Extend `PyBundledModule` with `extension.BundledModule`

  #11737 opened Jun 16, 2025

• Introduce apply_torch_ops_aten_passes to test mul fusion e2e

  #11741 opened Jun 16, 2025

• [BE] Add selected custom ops to CI

  #11744 opened Jun 16, 2025

• Update PAL override docs for new APIs

  #11747 opened Jun 17, 2025

• Refactor XNN workspace sharing to allow runtime gating

  #11748 opened Jun 17, 2025

• NXP backend: Enable initial unit tests workflow (#11396)

  #11756 opened Jun 17, 2025

• [5/N] Add get_option/set_option APIs

  #11758 opened Jun 17, 2025

• ET_HAS_EXCEPTIONS: require defined(_MSC_VER) to conclude _HAS_EXCEPTIONS implies exceptions are on

  #11770 opened Jun 17, 2025

• [Quantized DeConv Support] Enable Quantized Transposed Convs with groups==1

  #11774 opened Jun 18, 2025

• [Quantized DeConv Support] Dynamically Quantized Deconvolutions with groups ==1

  #11775 opened Jun 18, 2025

• [XNNPACK Quantizer] Select between TConvs and Convs

  #11776 opened Jun 18, 2025

• Fix bug in add op to ignore alpha != 1 (fixes: #11683)

  #11777 opened Jun 18, 2025

• Implement python deserialize for flat_tensor

  #11779 opened Jun 18, 2025

• Qualcomm AI Engine Direct - Delegate mutable buffer and fix the mutable buffer issue

  #11782 opened Jun 18, 2025

• Arm backend: Replace Tensor_Scalar with Tensor_Tensor in sqrt op

  #11783 opened Jun 18, 2025

• NXP backend: Add quantization of aten.view

  #11784 opened Jun 18, 2025

• Rename CMake target 'executorch' to 'prim_ops_lib' for clarity

  #11786 opened Jun 18, 2025

• Qualcomm AI Engine Direct - support cli

  #11788 opened Jun 18, 2025

• Bump pinned PT commit to 20250616 and mirror torch/standalone to ExecuTorch

  #11791 opened Jun 18, 2025

• Add last_token_pos in llama_transformer

  #11793 opened Jun 18, 2025

• downgrade flatbuffers in oss

  #11794 opened Jun 18, 2025

• Fix bug in sub op to ignore alpha != 1 (fixes: #11684)

  #11796 opened Jun 18, 2025

• Pull c10 headers directly from PyTorch internally, not from the c10 mirror

  #11800 opened Jun 18, 2025

• Enable sleef in xplat_exported_deps of aten_headers_for_executorch

  #11801 opened Jun 18, 2025

• Re-reapply "Add vectorized_math.h (#11204)", "Add optimized_portable_kernels test (#11205)", and "Add vectorization in elementwise_util (#9432)"

  #11802 opened Jun 18, 2025

• Support exp op in XNNPACK backend

  #11803 opened Jun 18, 2025

• Add tanh op to XNNPACK backend

  #11804 opened Jun 18, 2025

• [XNNPACK] Add support for Linear fused BatchNorm

  #11805 opened Jun 19, 2025

• Ability to specify full file configs from anywhere

  #11809 opened Jun 19, 2025

• Get rid of llm_config enums

  #11810 opened Jun 19, 2025

• Update export_llama in READMEs to use export_llm

  #11811 opened Jun 19, 2025

• Arm backend: Build c/c++ code with -Wall -Werror

  #11815 opened Jun 19, 2025

4 Issues closed by 4 people

• Model Load Time: SIMD Packing Kernels for Blockwise Quantization

  #9007 closed Jun 18, 2025

• Move to binary/unary subgraph

  #11584 closed Jun 17, 2025

• Update MediaTek backend docs using backend template

  #8532 closed Jun 17, 2025

• Error in Main Activity.java and ModelRunner.java

  #11725 closed Jun 17, 2025

56 Issues opened by 15 people

• Issue with dynamic_shapes when there are atleast one input argument whose dimensions are not dynamic

  #11820 opened Jun 19, 2025

• Qwen vocabulary size mismatch when using c++ tokenizer

  #11807 opened Jun 19, 2025

• Bug: edsr qnn export is broken

  #11797 opened Jun 18, 2025

• Is weight sharing in Vulkan backend supported?

  #11780 opened Jun 18, 2025

• llama_runner doesn't record latency on forward

  #11771 opened Jun 17, 2025

• Vulkan missing partition constraints for unsupported dtypes

  #11766 opened Jun 17, 2025

• Dtype not selected from exported pte via gen_oplist.py

  #11762 opened Jun 17, 2025

• Rename `executorch` cmake target to `prim_ops_lib`

  #11761 opened Jun 17, 2025

• Android native crash when running DeepLabV3Demo with vulkan backend in Android 16

  #11754 opened Jun 17, 2025

• Failed build of apple-ios example

  #11753 opened Jun 17, 2025

• Gradle Error while opening the project

  #11749 opened Jun 17, 2025

• Shared memory for multiple entry points with XNNPACK delegate

  #11738 opened Jun 16, 2025

• IF i want to build my own executorch lib to deploymodel on cpu, and the api is designed by myself

  #11726 opened Jun 16, 2025

• torch.split fails in to_edge

  #11723 opened Jun 16, 2025

• CoreML randn_like fails to lower with internal error

  #11722 opened Jun 16, 2025

• CoreML randint occasionally generates out of distribution values

  #11721 opened Jun 16, 2025

• CoreML fails to lower torch.rand

  #11720 opened Jun 16, 2025

• CoreML model with no inputs fails to load at runtime

  #11719 opened Jun 16, 2025

• CoreML gives wrong outputs or crashes for torch.diagonal

  #11718 opened Jun 16, 2025

• CoreML errors out handling torch.linalg.norm with ord="nuc"

  #11717 opened Jun 16, 2025

• CoreML doesn't handle negative dim values correctly in cumsum

  #11716 opened Jun 16, 2025

• CoreML argmin/argmax intermittently crashes process

  #11715 opened Jun 16, 2025

• CoreML floor_divide crashes process

  #11714 opened Jun 16, 2025

• CoreML fails to lower asinh/acosh

  #11712 opened Jun 16, 2025

• CoreML missing partitioner constraint for PixelUnshuffle on older targets

  #11711 opened Jun 16, 2025

• CoreML CircularPad1/2/3d fails to lower

  #11710 opened Jun 15, 2025

• CoreML ReplicationPad3d fails to load at runtime

  #11709 opened Jun 15, 2025

• CoreML ReflectionPad3d fails to load at runtime

  #11708 opened Jun 15, 2025

• CoreML fails to lower maxpool with indices

  #11706 opened Jun 15, 2025

• CoreML fails to lower ConvTranspose with output padding

  #11705 opened Jun 15, 2025

• Conv3d with circular padding on portable fails at runtime

  #11704 opened Jun 15, 2025

• CoreML conv fails to lower with circular padding mode

  #11703 opened Jun 15, 2025

• CoreML InstanceNorm3d fails to load

  #11702 opened Jun 15, 2025

• CoreML BatchNorm3d crashes process

  #11701 opened Jun 15, 2025

• In-place activations alter graph output after lowering

  #11700 opened Jun 15, 2025

• CoreML missing partitioner constraint for topk with sorted=False

  #11698 opened Jun 14, 2025

• CoreML missing partitioner constraint for max_pool with dilation

  #11697 opened Jun 14, 2025

• CoreML fails to lower pooling with single-element padding list

  #11696 opened Jun 14, 2025

• CoreML missing partitioner constraint for avg_pool2d with a divisor_override

  #11695 opened Jun 14, 2025

• CoreML missing partition constraint for tensors with rank greater than 5

  #11694 opened Jun 14, 2025

• CoreML fails to lower integer ReLU

  #11693 opened Jun 14, 2025

• BF16 relu fails to lower (no delegation)

  #11692 opened Jun 14, 2025

• CoreML int32 mm with fixed arg fails to load at runtime

  #11691 opened Jun 14, 2025

• CoreML fails to lower int16 mm with param arg

  #11690 opened Jun 14, 2025

• CoreML fails to lower addmm with integer alpha/beta parameters

  #11689 opened Jun 14, 2025

• CoreML crashes when running convolution

  #11688 opened Jun 14, 2025

• CoreML ignores add/sub alpha parameter

  #11687 opened Jun 14, 2025

• CoreML partitioner errors out with 8-bit integer types

  #11686 opened Jun 14, 2025

• XNNPACK sub ignores alpha value

  #11684 opened Jun 14, 2025

• XNNPACK add ignores alpha value

  #11683 opened Jun 14, 2025

• Can we use ExecuTorch from cmake-out to export PTEs/PTDs for tests?

  #11656 opened Jun 13, 2025

• MPS delegate crashes on iOS 26

  #11655 opened Jun 13, 2025

• Support getting the delegated payload

  #11651 opened Jun 13, 2025

• Debug discrepancy of `avg_inference_latency` reported from optimum-et models

  #11650 opened Jun 13, 2025

• Clean up AWS devices

  #11649 opened Jun 13, 2025

• Facing an issue with weird model sizes for *.pte model when compared to corresponding torch script model (*.pt)

  #11637 opened Jun 13, 2025

61 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.

• Phi3 runner uses TextLLMRunner

  #11551 commented on Jun 16, 2025 • 7 new comments

• Update install script and building from source docs

  #10652 commented on Jun 17, 2025 • 6 new comments

• Device telemetry for benchmark

  #11301 commented on Jun 17, 2025 • 3 new comments

• Milestone2.2: Optimize transposes in XNNPACK partition by removing redundant to_copy ops

  #11316 commented on Jun 13, 2025 • 3 new comments

• Use ExternalProject_Add to build tokenizers

  #11550 commented on Jun 13, 2025 • 2 new comments

• [3/N] Add get_option/set_option function in backend interface

  #11391 commented on Jun 18, 2025 • 2 new comments

• [4/N] Add backend options map

  #11462 commented on Jun 18, 2025 • 2 new comments

• [1/N] Add BackendOptions class

  #11389 commented on Jun 18, 2025 • 1 new comment

• Milestone2.1: Partition to_dim_order_copy op in XNN delegate

  #11286 commented on Jun 13, 2025 • 0 new comments

• Improve torch.ops typing

  #11276 commented on Jun 13, 2025 • 0 new comments

• NXP backend: Add support for depthwise and separable convolution.

  #11215 commented on Jun 18, 2025 • 0 new comments

• Enable all platform tests in CMake

  #11124 commented on Jun 13, 2025 • 0 new comments

• Add torchao kernels to xcframework

  #10963 commented on Jun 19, 2025 • 0 new comments

• [pytorch hash update] update the pinned pytorch hash

  #10955 commented on Jun 20, 2025 • 0 new comments

• Add aot example with Neutron Backend

  #10871 commented on Jun 18, 2025 • 0 new comments

• Clean up eager quant in llm_export

  #10684 commented on Jun 17, 2025 • 0 new comments

• [devtool] create stream_data_sink

  #8604 commented on Jun 18, 2025 • 0 new comments

• Adjust tolerance for quantized XNN conv1d tests

  #8093 commented on Jun 18, 2025 • 0 new comments

• install_executorch.sh --clean should remove buck-out/

  #11564 commented on Jun 13, 2025 • 0 new comments

• [Draft] Qualcomm AI Engine Direct - Unexpected graph for mutable buffer after export during Quantization

  #11309 commented on Jun 18, 2025 • 0 new comments

• Qualcomm AI Engine Direct - CI for Non-LLM GA model

  #11357 commented on Jun 19, 2025 • 0 new comments

• [2/N] Add option context

  #11390 commented on Jun 18, 2025 • 0 new comments

• [5/N] Add update in method

  #11463 commented on Jun 14, 2025 • 0 new comments

• NXP backend: Add docker image for nxp backend tests

  #11473 commented on Jun 18, 2025 • 0 new comments

• [xnn update prep] deprecate sdpa

  #11506 commented on Jun 13, 2025 • 0 new comments

• [xnn update prep] Update ET Pthreadpool third-party to google

  #11507 commented on Jun 19, 2025 • 0 new comments

• Replace debug handle with `from_node` to trace operator transformation

  #11532 commented on Jun 13, 2025 • 0 new comments

• [6/N] Add update in module

  #11533 commented on Jun 13, 2025 • 0 new comments

• [7/N] Add sugar syntax for module.update

  #11534 commented on Jun 14, 2025 • 0 new comments

• Add coreml eager model compare

  #11574 commented on Jun 13, 2025 • 0 new comments

• [do not land] selective build size test

  #11582 commented on Jun 16, 2025 • 0 new comments

• Runtime crash in PreTokenizer when running optimum-et generated gemma3 via llama runner

  #11618 commented on Jun 13, 2025 • 0 new comments

• Android Benchmark App Reports Success Despite Empty Results

  #11620 commented on Jun 13, 2025 • 0 new comments

• Update Partitioner Dtype Logic

  #11581 commented on Jun 13, 2025 • 0 new comments

• Fail to allocate temp memory by exporting torch.topk

  #8700 commented on Jun 14, 2025 • 0 new comments

• Optimized add fails fails at runtime with rank error when broadcasting

  #10959 commented on Jun 14, 2025 • 0 new comments

• XNN int8 out of bounds write / invalid outputs

  #10960 commented on Jun 14, 2025 • 0 new comments

• XNN Int8 ReLU gives incorrect outputs

  #10961 commented on Jun 14, 2025 • 0 new comments

• XNN Conv2d won't lower with single stride/dilation/etc.

  #10965 commented on Jun 14, 2025 • 0 new comments

• XNN AvgPool2d fails to lower with single element kernel_size

  #10968 commented on Jun 14, 2025 • 0 new comments

• XNN MaxPool2d fails to lower with single element kernel_size

  #10969 commented on Jun 14, 2025 • 0 new comments

• Support dim order in Java Tensor

  #11593 commented on Jun 15, 2025 • 0 new comments

• Implement load_into for the file descriptor data loader

  #11563 commented on Jun 16, 2025 • 0 new comments

• Add pybindings for program.h/cpp and method.h/cpp

  #11567 commented on Jun 16, 2025 • 0 new comments

• Successfully registered all kernels from shared library: NOT_SUPPORTED

  #8870 commented on Jun 16, 2025 • 0 new comments

• Simple indexing prevents ios18.gather from being schedule-able on the ANE

  #11615 commented on Jun 16, 2025 • 0 new comments

• Support Standalone Batch Norm

  #11586 commented on Jun 16, 2025 • 0 new comments

• Multiple Entrypoints and Shared State implementation; Request for comments/feedback.

  #8030 commented on Jun 16, 2025 • 0 new comments

• Introduce export_llama to extension/llm as export_llm

  #11527 commented on Jun 16, 2025 • 0 new comments

• Implement load_into for mmap data loader

  #11561 commented on Jun 16, 2025 • 0 new comments

• Export a model with multiple entry points

  #7458 commented on Jun 17, 2025 • 0 new comments

• executorch model Inference time is higher than the torch model

  #10297 commented on Jun 17, 2025 • 0 new comments

• [v0.7.0] Release Tracker

  #11075 commented on Jun 18, 2025 • 0 new comments

• Deprecate tag implicit qdq pass

  #11588 commented on Jun 18, 2025 • 0 new comments

• Support higher-order while_loop and scan torch ops

  #8769 commented on Jun 20, 2025 • 0 new comments

• Support Linear Fused Batchnorm

  #11587 commented on Jun 20, 2025 • 0 new comments

• fix spec_prop_pass

  #7974 commented on Jun 18, 2025 • 0 new comments

• Fix comment in memory_planning.py

  #8010 commented on Jun 18, 2025 • 0 new comments

• remove the exec_aten namespace

  #8018 commented on Jun 18, 2025 • 0 new comments

• Revert to use mean_out than mean_dim_out

  #8021 commented on Jun 18, 2025 • 0 new comments

• [docs][ez] Fix doc build workflow

  #8079 commented on Jun 18, 2025 • 0 new comments