-
Notifications
You must be signed in to change notification settings - Fork 654
Insights: pytorch/executorch
Overview
Could not load contribution data
Please try again later
113 Pull requests merged by 42 people
-
[ET-VK] Introduce custom op correctness + speed testing suite & add vulkan operator testing to CI
#13835 merged
Aug 30, 2025 -
[ET-VK][ez] Add ability to check for dot product extension support + upgrade glslc
#13834 merged
Aug 30, 2025 -
[ET-VK] Introduce custom op correctness + speed testing suite & add vulkan operator testing to CI
#13815 merged
Aug 30, 2025 -
[ET-VK][ez] Add ability to check for dot product extension support + upgrade glslc
#13814 merged
Aug 30, 2025 -
export static llama with masked softmax
#13832 merged
Aug 30, 2025 -
Summary: Minor cleanup post quantized_add op
#13824 merged
Aug 30, 2025 -
Add initializer with file and data file paths
#13830 merged
Aug 29, 2025 -
Update executorch branch to swiftpm-0.8.0.20250829
#13828 merged
Aug 29, 2025 -
Clean up includes in text_decoder_runner.h
#13827 merged
Aug 29, 2025 -
Add tensor accessor and to umbrella includes
#13826 merged
Aug 29, 2025 -
Clean up optimized op_add & op_sub
#13764 merged
Aug 29, 2025 -
Use macos-14-xlarge instead of macos-latest-xlarge which has switched…
#13821 merged
Aug 29, 2025 -
Add backend-agnostic implementation for quantize_per_tensor
#13769 merged
Aug 29, 2025 -
Add custom cadence::linalg_svd operation with contiguous strides
#13718 merged
Aug 29, 2025 -
Clean up optimized op_mul
#13763 merged
Aug 29, 2025 -
Test Half/Bfloat16: div, lt, ge, gt, eq, ne
#13762 merged
Aug 29, 2025 -
enable_qnn_masked_softmax
#13788 merged
Aug 29, 2025 -
update custom annotations for QAT
#13747 merged
Aug 29, 2025 -
NXP Backend: Use --index-url for eiq.nxp.com/repository
#13759 merged
Aug 29, 2025 -
Fix rerunfailures to 15.1
#13807 merged
Aug 29, 2025 -
Remove test_aot_buffer_mutation
#13808 merged
Aug 29, 2025 -
Arm backend: Enable building and running with the pte on an address
#13761 merged
Aug 29, 2025 -
JNI replace llama runner with llm text runner
#13784 merged
Aug 29, 2025 -
remove convert in compiler_funcs
#13698 merged
Aug 29, 2025 -
Summary: Throw appropriate Executorch exception which ic mapped to ap…
#13771 merged
Aug 28, 2025 -
Qualcomm AI Engine Direct - GA Qwen 2.5 0.5B
#12333 merged
Aug 28, 2025 -
Install NXP and OpenVINO backends, top-level CMake targets
#13673 merged
Aug 28, 2025 -
Android LlamaDemo update gradle version
#13751 merged
Aug 28, 2025 -
Revert "Disable tokenizers on Windows (#13667)"
#13767 merged
Aug 28, 2025 -
Summary: Initial CMSS-NN Add Op
#13296 merged
Aug 28, 2025 -
[ET][Kernels] Increase Half/Bfloat16 support
#13719 merged
Aug 28, 2025 -
Arm backend: Refactor partitioning of int64
#13726 merged
Aug 28, 2025 -
Add 16A8W linear ops support and test
#13658 merged
Aug 28, 2025 -
Arm backend: Added TOSA workflow minimal example
#13727 merged
Aug 28, 2025 -
Refactor op_div: fix bug + enable Half/Bfloat16
#13740 merged
Aug 28, 2025 -
Build an AAR with QNN
#13687 merged
Aug 27, 2025 -
Summary: Add MCU model verification to CI
#13572 merged
Aug 27, 2025 -
Enable long audio for WhisperAudioProcessor
#13736 merged
Aug 27, 2025 -
Enable xnnpack weight sharing in ios and android
#13738 merged
Aug 27, 2025 -
Arm backend: Fix bug in decompose_linear_pass
#13725 merged
Aug 27, 2025 -
Add pendings ops to the to be supported list
#13688 merged
Aug 27, 2025 -
Qualcomm AI Engine Direct - Add&update README
#13504 merged
Aug 27, 2025 -
Fix remaining build issues on Windows, re-enable Windows pybind preset build CI job
#13702 merged
Aug 27, 2025 -
Fix portable library build on Windows
#13700 merged
Aug 27, 2025 -
Add 16A8W quantization configuration utility for ARM backend
#13641 merged
Aug 27, 2025 -
Revert "[EXIR] Register _clone_dim_order op and map aten.clone"
#13723 merged
Aug 27, 2025 -
Arm backend: Fix numerical accuracy error.
#13578 merged
Aug 27, 2025 -
Arm backend: Add back sigmoid_16bit_add xfail
#13724 merged
Aug 27, 2025 -
Disable tokenizers on Windows
#13667 merged
Aug 27, 2025 -
Enumerated shape
#13613 merged
Aug 27, 2025 -
Add 'kernels_torchao' to the Apple workflow
#13720 merged
Aug 27, 2025 -
Remove int32_t support for quant/dequant kernels
#13703 merged
Aug 27, 2025 -
[ET][Kernels] Increase Half/Bfloat16 support
#13646 merged
Aug 27, 2025 -
Move undelegated constants
#13606 merged
Aug 27, 2025 -
Qualcomm AI Engine Direct - Fixed the accuracy bug for conv with bias in block quant
#13676 merged
Aug 26, 2025 -
Re-enable Windows build CI
#13699 merged
Aug 26, 2025 -
[EXIR] Register _clone_dim_order op and map aten.clone
#12971 merged
Aug 26, 2025 -
Added documentation and typing to WhisperAudioProcessor
#13661 merged
Aug 26, 2025 -
Bump optimum ET pin
#13530 merged
Aug 26, 2025 -
add ability to compare intermedidate outputs
#13482 merged
Aug 26, 2025 -
[ET-VK][ez] Allow partitioner to filter by source nn.Module type name
#13693 merged
Aug 26, 2025 -
[ET-VK] Implement expand
#13692 merged
Aug 26, 2025 -
[ET-VK] Add high dim support for permute
#13691 merged
Aug 26, 2025 -
Revert "Add check_for_installed_private_headers_in_cmake_out"
#13685 merged
Aug 26, 2025 -
[ET-VK][ez] Allow partitioner to filter by source nn.Module type name
#13644 merged
Aug 26, 2025 -
[ET-VK] Implement expand
#13643 merged
Aug 26, 2025 -
[ET-VK] Add high dim support for permute
#13642 merged
Aug 26, 2025 -
Qualcomm AI Engine Direct - op enablement sign asin xor floor_divide binary
#13675 merged
Aug 26, 2025 -
Qualcomm AI Engine Direct - Fix mem_handel register twice issue
#13410 merged
Aug 26, 2025 -
Remove unused value tags from ObjC ExecuTorchValue
#13679 merged
Aug 26, 2025 -
QNN Whisper Runner transcribe documentation
#13665 merged
Aug 26, 2025 -
Clean up ET_Check in depthwise conv, replace with assert
#13655 merged
Aug 26, 2025 -
Revert #13526 and #13640
#13684 merged
Aug 26, 2025 -
Add 1MB temp allocator for executor
#13533 merged
Aug 26, 2025 -
Arm backend: Fixes for VGF models
#13681 merged
Aug 26, 2025 -
Arm backend: Make TOSA backend NCHW-compatible
#12994 merged
Aug 26, 2025 -
NXP backend: Add MobileNetV2 example model and test
#12892 merged
Aug 26, 2025 -
Arm backend: Use cmake for building in Ethos-U jupyter example
#13630 merged
Aug 26, 2025 -
Allow zero-element inputs for method.
#13623 merged
Aug 26, 2025 -
Fix bad optimized kernel for add.
#13633 merged
Aug 26, 2025 -
Disable mm + add -> addmm fusion if added tensor rank >2
#13632 merged
Aug 26, 2025 -
Override unload_method in training_module to erase the tensors pointing to the released memory
#13590 merged
Aug 26, 2025 -
Smollm targets
#13659 merged
Aug 26, 2025 -
Inline requantize kernels
#13592 merged
Aug 26, 2025 -
NXP backend: Remove IR optimization to remove dead branches.
#13574 merged
Aug 26, 2025 -
Increase binary size limit by 8 bytes
#13671 merged
Aug 26, 2025 -
Fix data loader build on Windows, re-enable pybind job
#13564 merged
Aug 26, 2025 -
Fix portable library build on Windows
#13260 merged
Aug 26, 2025 -
Fix XNNPACK build on Windows
#13259 merged
Aug 26, 2025 -
Temporarily disable windows preset build in CI
#13669 merged
Aug 25, 2025 -
Create Windows CMake preset
#13257 merged
Aug 25, 2025 -
Fix devtools CMake build failure on Windows
#13251 merged
Aug 25, 2025 -
Fix error reporting in Windows preset build job
#13247 merged
Aug 25, 2025 -
Qualcomm AI Engine Direct - Fix broken unpacking in T5 dataset loading
#13625 merged
Aug 25, 2025 -
Qualcomm AI Engine Direct - Improve GA Static Phi-4-mini accuracy
#13573 merged
Aug 25, 2025 -
Qualcomm AI Engine Direct - Scripts and accuracy improvement for Qwen3_0.6B/1.7B and Qwen 2.5_1.5B
#13544 merged
Aug 25, 2025 -
Fully enable the stale PR workflow
#13656 merged
Aug 25, 2025 -
Resurface low level runtime API page
#13651 merged
Aug 25, 2025 -
Allow none and string input types for Method
#13645 merged
Aug 25, 2025 -
Summary: Follow up fix to pr#13526
#13640 merged
Aug 25, 2025 -
Add support for data path in iOS
#13620 merged
Aug 25, 2025 -
Run all periodic models when ciflow/periodic label is present
#13634 merged
Aug 25, 2025 -
Added JS bindings for tokenizers library
#13566 merged
Aug 25, 2025 -
Update cpuinfo pin to latest
#13624 merged
Aug 25, 2025 -
NXP backend: Add support for conversion of Conv1D operator
#13549 merged
Aug 25, 2025 -
Add check_for_installed_private_headers_in_cmake_out
#13485 merged
Aug 25, 2025 -
Update coremltools to 9b1
#13614 merged
Aug 25, 2025 -
Fix aten.amax lowering issue
#13381 merged
Aug 25, 2025 -
NXP backend: Use zero point for quantized padding.
#13576 merged
Aug 25, 2025 -
Cortex_m backend: Loosen edge op check.
#13550 merged
Aug 25, 2025 -
NXP backend: Fix
tanhmerge conflict.#13626 merged
Aug 25, 2025 -
Arm backend: Dont try to fuse const for TOSA ops
#13575 merged
Aug 25, 2025
77 Pull requests opened by 31 people
-
ET windows build flags
#13636 opened
Aug 25, 2025 -
ET compiler specific macros for decorators
#13637 opened
Aug 25, 2025 -
ET use compiler macros in ops
#13638 opened
Aug 25, 2025 -
Force -O3 for executorch op_div.cpp in "clang 17" also
#13660 opened
Aug 25, 2025 -
Add audio to multimodal runner
#13662 opened
Aug 25, 2025 -
Add Voxtral runner
#13663 opened
Aug 25, 2025 -
Make token and stat callback optional in multimodal runner
#13664 opened
Aug 25, 2025 -
Start pos api llava
#13668 opened
Aug 25, 2025 -
Fix build on Windows, re-enable Windows build CI job
#13672 opened
Aug 25, 2025 -
Enable per-row/per-col grouping in CoreML LUT ops
#13674 opened
Aug 26, 2025 -
Exynos Backend for Executorch to bring up on Exynos SoC
#13677 opened
Aug 26, 2025 -
Arm Backend: Add support for ELU.default operator
#13683 opened
Aug 26, 2025 -
Add error logging to scripts/print_exported_headers
#13686 opened
Aug 26, 2025 -
Update demo app documentation for qualcomm
#13694 opened
Aug 26, 2025 -
Fix build_android_library.sh
#13695 opened
Aug 26, 2025 -
Fix batch norm partitioning with Conv3d
#13696 opened
Aug 26, 2025 -
[WIP] Add tests for multimodal runner
#13697 opened
Aug 26, 2025 -
Move oss only files to oss folder
#13715 opened
Aug 26, 2025 -
[Windows] Run Python unit test CI on Windows
#13716 opened
Aug 26, 2025 -
Use extension module for pybind
#13717 opened
Aug 26, 2025 -
Add support for checking more than one output from delegate in numerical comparator
#13722 opened
Aug 27, 2025 -
Add 16A8W quantization configuration utility for ARM backend
#13728 opened
Aug 27, 2025 -
Optimize index_out via fast path
#13731 opened
Aug 27, 2025 -
Add qnn backend to pip package
#13733 opened
Aug 27, 2025 -
[ET][Portable] Add op grid_sampler_2d
#13734 opened
Aug 27, 2025 -
[EXIR] Register _clone_dim_order op and map aten.clone
#13735 opened
Aug 27, 2025 -
[Draft] Parq 2-bit experiment
#13737 opened
Aug 27, 2025 -
Update selective build example + CI for top-level targets
#13741 opened
Aug 27, 2025 -
Remove last remnants of requantize kernels
#13749 opened
Aug 27, 2025 -
Add support for strongly typed softmax
#13750 opened
Aug 27, 2025 -
Include audio preprocessing for raw audio tensor
#13752 opened
Aug 28, 2025 -
Arm backend: Add 16A8W linear ops support and test
#13754 opened
Aug 28, 2025 -
Qualcomm AI Engine Direct - Static LLM Refactor & Qwen3 1.7B Improvement
#13755 opened
Aug 28, 2025 -
NXP backend: Disable training mode and replace deprecated call
#13756 opened
Aug 28, 2025 -
NXP backend: Add preprocessing pass to split multilayer `GRU`.
#13757 opened
Aug 28, 2025 -
NXP backend: Separable convolution decomposition in executorch
#13758 opened
Aug 28, 2025 -
NXP backend: Fix incorrect `linear` test.
#13760 opened
Aug 28, 2025 -
Clean up optimized op_le
#13765 opened
Aug 28, 2025 -
Fix qnn backend error message
#13768 opened
Aug 28, 2025 -
Summary: Initial LinearOps (Stateless Fully Connected) with cmsis-nn …
#13770 opened
Aug 28, 2025 -
Android LlamaDemo small refactor
#13772 opened
Aug 28, 2025 -
Batch audio in chunks
#13774 opened
Aug 28, 2025 -
Make lookahead decode to support early stop
#13776 opened
Aug 28, 2025 -
Add backend-agnostic implementation for dequantize_per_tensor
#13777 opened
Aug 28, 2025 -
Start pos api llava 2
#13780 opened
Aug 28, 2025 -
Remove prefill methods for image and text input
#13783 opened
Aug 29, 2025 -
Add 16A8W support and test for mul operation
#13785 opened
Aug 29, 2025 -
Arm backend: Add 16A8W support and test for add operation
#13789 opened
Aug 29, 2025 -
[Executorch][target recipes] Add target based recipes for lowering models to a target device
#13791 opened
Aug 29, 2025 -
Arm backend: Fix compilation errors in executor runner
#13792 opened
Aug 29, 2025 -
Add 16A8W support and test for mul operation
#13793 opened
Aug 29, 2025 -
Add qnn backend to pip packages
#13794 opened
Aug 29, 2025 -
Arm backend: Add 16A8W support and test for mul operation
#13795 opened
Aug 29, 2025 -
Arm backend: Add 16A8W support and test for sigmoid operation
#13796 opened
Aug 29, 2025 -
Arm backend: Add 16A8W support and test for tanh operation
#13797 opened
Aug 29, 2025 -
Arm backend: Add 16A8W support and test for slice operation
#13798 opened
Aug 29, 2025 -
Arm backend: Add 16A8W support for view and transpose operations
#13799 opened
Aug 29, 2025 -
Arm backend: Add 16A8W support and test for cat operation
#13800 opened
Aug 29, 2025 -
Arm backend: Add 16A8W FCNode support with BMM dependency fix
#13801 opened
Aug 29, 2025 -
Arm backend: Add INT16 support to rescale operation
#13802 opened
Aug 29, 2025 -
Arm backend: Add passes to handle int64 const and int64 output ops
#13803 opened
Aug 29, 2025 -
Arm backend: Add requirements.txt for arm backend
#13804 opened
Aug 29, 2025 -
model : support LiquidAI LFM2 hybrid family
#13805 opened
Aug 29, 2025 -
Adding Vulkan support for executorch on IG4A
#13806 opened
Aug 29, 2025 -
Add create_mutable_buffer util
#13813 opened
Aug 29, 2025 -
[ET-VK] Quantized Int8 Linear
#13816 opened
Aug 29, 2025 -
[ET-VK] Quantized Int8 Convolution
#13817 opened
Aug 29, 2025 -
[ET-VK][AOT] Enable exporting Q8 Quantized Linear + Convolution
#13818 opened
Aug 29, 2025 -
Arm backend: Add NGP unit test scripts
#13819 opened
Aug 29, 2025 -
D81187339: [Cadence] Add backend-agnostic implementation for quantized_add
#13820 opened
Aug 29, 2025 -
Refactor stack utils
#13822 opened
Aug 29, 2025 -
Support Half/BFloat16 in native_group_norm
#13823 opened
Aug 29, 2025 -
Jni layer use multimodal runner
#13825 opened
Aug 29, 2025 -
Fix bug in elu
#13829 opened
Aug 29, 2025 -
Fix reduction over dim list for empty input
#13833 opened
Aug 30, 2025 -
[Windows] Add model run CI
#13836 opened
Aug 31, 2025 -
[Windows] Add CI wheel build jobs for Windows
#13837 opened
Aug 31, 2025
17 Issues closed by 9 people
-
Is the process to convert .safetensors to .pte documented?
#7958 closed
Aug 28, 2025 -
How to run Qwen using Executorch?
#7467 closed
Aug 28, 2025 -
FVP (CMSIS-NN Integ) : Add support to Quantized Arithmetic Operations: cortex_m::quantized_add.out
#13704 closed
Aug 28, 2025 -
Support enumerated shapes for models with multiple inputs
#9632 closed
Aug 28, 2025 -
Installable top-level CMake targets
#12949 closed
Aug 28, 2025 -
Abstract Backend testing logic
#10100 closed
Aug 28, 2025 -
[Cortex-M] Quantized Add with CMSIS-Integ
#12492 closed
Aug 28, 2025 -
--skip-fvp-setup flag is ignored
#12447 closed
Aug 28, 2025 -
Set up Windows build CI
#12962 closed
Aug 27, 2025 -
Support build on native Windows
#4661 closed
Aug 26, 2025 -
Is it possible to use Qualcomm Backend & ExecuTorch (experimental) training?
#5417 closed
Aug 26, 2025 -
"Using Executorch with C++" Documentation out of date
#13631 closed
Aug 26, 2025 -
[Delegate Testing] Add pybinding support for QNN
#12896 closed
Aug 26, 2025 -
Add dim order variant clone operator
#12645 closed
Aug 26, 2025 -
Android GA: Add Exceptions in JNI layer, Establish the foundational pieces
#13583 closed
Aug 26, 2025 -
[Vulkan] Add (limited) support for >4-dim tensors
#13082 closed
Aug 25, 2025
31 Issues opened by 9 people
-
How to enable XNN_ENABLE_SPARSE in Executorch
#13787 opened
Aug 29, 2025 -
Check Windows perf on XNNPACK
#13786 opened
Aug 29, 2025 -
Official Windows Support
#13781 opened
Aug 28, 2025 -
Build QNN AOT Pybind11 extensions in setup.py
#13778 opened
Aug 28, 2025 -
Some CI jobs are failing on Apple
#13773 opened
Aug 28, 2025 -
Simplify qualcomm qnn backend setup
#13766 opened
Aug 28, 2025 -
Set up Windows unit test CI
#13753 opened
Aug 28, 2025 -
Extend broadcast_args_pass.py to also handle CMSIS-NN kernel selection:
#13748 opened
Aug 27, 2025 -
Create dedicated test_suite for for cortex_m ops
#13739 opened
Aug 27, 2025 -
[RFC] Multi-backend recipes for easy target focused model deployment
#13732 opened
Aug 27, 2025 -
[Arm] Support INT16 ADD ops with TOSA reference model run
#13730 opened
Aug 27, 2025 -
[Arm] Support Linear INT16 TOSA reference model run
#13729 opened
Aug 27, 2025 -
[Phase 2] : Perform / Analyze E2E inference of mv2/mv3 on FVP
#13713 opened
Aug 26, 2025 -
FVP(CMSIS-NN Integ) : Add support to Activation Functions - cortex_m::quantized_hardtanh.out
#13712 opened
Aug 26, 2025 -
FVP(CMSIS-NN Integ) : Add support to Activation Functions - cortex_m::quantized_relu.out
#13711 opened
Aug 26, 2025 -
FVP (CMSIS-NN Integ): Add support to Neural Network Layers - cortex_m::quantized_max_pool2d.out
#13710 opened
Aug 26, 2025 -
FVP (CMSIS-NN Integ): Add support to Neural Network Layers - cortex_m::quantized_avg_pool2d.out
#13709 opened
Aug 26, 2025 -
FVP (CMSIS-NN Integ): Add support to Neural Network Layers - cortex_m::quantized_linear_per_tensor_out
#13708 opened
Aug 26, 2025 -
FVP (CMSIS-NN Integ): Add support to Neural Network Layers - cortex_m::quantized_conv2d.out
#13707 opened
Aug 26, 2025 -
FVP (CMSIS-NN Integ) : Add support to Quantized Arithmetic Operations: cortex_m::quantized_sub.out
#13706 opened
Aug 26, 2025 -
FVP (CMSIS-NN Integ) : Add support to Quantized Arithmetic Operations: cortex_m::quantized_mul.out
#13705 opened
Aug 26, 2025 -
Fully support mv2/ mv3 with fused ops + cmsis-integration on FVP
#13701 opened
Aug 26, 2025 -
Conv3d + Batch Norm fails to lower on XNNPACK
#13690 opened
Aug 26, 2025 -
Enable Pico2 support for ExecuTorch
#13680 opened
Aug 26, 2025 -
Partition only user-annotated subgraphs
#13657 opened
Aug 25, 2025 -
Update docs for low-level C++ APIs.
#13654 opened
Aug 25, 2025 -
[Backend Tester] Update README and in-code documentation
#13652 opened
Aug 25, 2025 -
Add checksum to ET to validate file size/contents
#13639 opened
Aug 25, 2025 -
Add 16A8W linear ops support and test
#13635 opened
Aug 25, 2025 -
where can i do the quantize or de-quantize operation in demo
#13628 opened
Aug 25, 2025 -
Refactor/Simplify the quantized_op_fusion_pass.py with subgraph_rewriter
#13627 opened
Aug 25, 2025
392 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.
-
Summary: Add Initial support to pico2 (Arm Cortex M)
#12518 commented on
Aug 26, 2025 • 7 new comments -
Arm backend: Annotate ADD/SUB with indepenedent observers
#13516 commented on
Aug 29, 2025 • 3 new comments -
[6/N] Add update in module
#11533 commented on
Aug 27, 2025 • 3 new comments -
Add logical and op
#13342 commented on
Aug 28, 2025 • 2 new comments -
Qualcomm AI Engine Direct - Avoid duplicated external ION memory registration
#13421 commented on
Aug 26, 2025 • 2 new comments -
[5/N] Add update in method
#11463 commented on
Aug 27, 2025 • 1 new comment -
guard_or_false in dim_order == 0 check
#12651 commented on
Aug 29, 2025 • 1 new comment -
NXP backend: Update to Neutron Converter SDK_25.06
#13128 commented on
Aug 29, 2025 • 1 new comment -
[devtool] make ETDumpGen use bufferdatasink
#8740 commented on
Aug 30, 2025 • 0 new comments -
Adding new memory plannig algorithm heap_optimized_greedy
#8773 commented on
Aug 30, 2025 • 0 new comments -
[devtool] make datasink as a sepreate directory
#8804 commented on
Aug 30, 2025 • 0 new comments -
update demo scripts to use .ptd
#8834 commented on
Aug 30, 2025 • 0 new comments -
Make the quantized path the main compilation path (#7939)
#8852 commented on
Aug 30, 2025 • 0 new comments -
Cleanup XNN delegate post removal of upsample decomposition
#8910 commented on
Aug 30, 2025 • 0 new comments -
Enable SDPA without kv cache
#8950 commented on
Aug 30, 2025 • 0 new comments -
Update xnnpack_README.md
#8951 commented on
Aug 31, 2025 • 0 new comments -
Dev 4bit embedding
#8962 commented on
Aug 31, 2025 • 0 new comments -
Check for null method in module execute
#9713 commented on
Aug 31, 2025 • 0 new comments -
[devtool] introduce sanity check when creating bufferdatasink
#8708 commented on
Aug 30, 2025 • 0 new comments -
[devtool] introduce sanity check when creating bufferdatasink
#8707 commented on
Aug 30, 2025 • 0 new comments -
[devtool] introduce sanity check when creating bufferdatasink
#8706 commented on
Aug 30, 2025 • 0 new comments -
[cadence] create sdk CI actions for CPU and DSP
#8639 commented on
Aug 30, 2025 • 0 new comments -
[devtool] update copy_tensor_to_debug_buffer
#8623 commented on
Aug 30, 2025 • 0 new comments -
[devtool] create stream_data_sink
#8604 commented on
Aug 29, 2025 • 0 new comments -
Fix XNN backend in pybindings by not linking backend runtime targets in AOT lib
#8565 commented on
Aug 30, 2025 • 0 new comments -
Allow querying all backends registered with the executorch runtime
#8521 commented on
Aug 30, 2025 • 0 new comments -
Enable dtype selective build by default
#8507 commented on
Aug 30, 2025 • 0 new comments -
[TEST] BUILD
#8484 commented on
Aug 30, 2025 • 0 new comments -
[executorch][flat tensor] Store number of external tensors in flatbuffer
#8483 commented on
Aug 30, 2025 • 0 new comments -
Add performance benchmark config: MPS 8da4w
#8461 commented on
Aug 30, 2025 • 0 new comments -
Fix HiFi relu for int8
#8423 commented on
Aug 30, 2025 • 0 new comments -
Support multi-module exports in Inspector and ETRecord
#8336 commented on
Aug 30, 2025 • 0 new comments -
decode to prefill
#9662 commented on
Aug 31, 2025 • 0 new comments -
Make a real buck file for Apple extension to run tests
#9566 commented on
Aug 31, 2025 • 0 new comments -
Refactor LLMEdgeManager's to_dtype
#9542 commented on
Aug 31, 2025 • 0 new comments -
Added support for bias in optimized linear operation
#9527 commented on
Aug 31, 2025 • 0 new comments -
[etLLM][Config, Part1] Convert Args to DictConfig
#9450 commented on
Aug 31, 2025 • 0 new comments -
Reset IO for the multiple iters output accurancy
#9412 commented on
Aug 31, 2025 • 0 new comments -
Arm backend: Enable mypy lintrunner for backends/arm/test
#9357 commented on
Aug 31, 2025 • 0 new comments -
[Executorch][kv cache] Make quantized cache return only the updated cache portion
#9351 commented on
Aug 31, 2025 • 0 new comments -
[Executorch][llama] Change runner to decouple prompt length from sequence length
#9350 commented on
Aug 31, 2025 • 0 new comments -
Remove ops with 0 bytes memory allocation in memory profiler. (#6545)
#9338 commented on
Aug 31, 2025 • 0 new comments -
[ExecuTorch] Remove xnn_executor_runner
#9292 commented on
Aug 31, 2025 • 0 new comments -
[Do not land] export Mimi HF version
#9255 commented on
Aug 31, 2025 • 0 new comments -
Add arm64 size_test
#9252 commented on
Aug 31, 2025 • 0 new comments -
Make oss Coreml Llama supports both list and tensor KV cache inputs
#9225 commented on
Aug 31, 2025 • 0 new comments -
bump pytorch version
#9214 commented on
Aug 31, 2025 • 0 new comments -
Test Sync
#9204 commented on
Aug 31, 2025 • 0 new comments -
Arm backend: Add small check when input type is a list
#9186 commented on
Aug 31, 2025 • 0 new comments -
Bump Android SDK version in docker image
#9156 commented on
Aug 31, 2025 • 0 new comments -
[cadence][g3] enable fp32 permute
#9053 commented on
Aug 31, 2025 • 0 new comments -
Remove explicit descriptor where not needed
#9046 commented on
Aug 31, 2025 • 0 new comments -
Remove explicit descriptor
#9001 commented on
Aug 31, 2025 • 0 new comments -
[minibench] Add RSS stats
#8982 commented on
Aug 31, 2025 • 0 new comments -
Use etdump gen instead of event tracer
#8968 commented on
Aug 31, 2025 • 0 new comments -
Add XNN U8 op support via quantization
#8330 commented on
Aug 30, 2025 • 0 new comments -
Add test case to export, quantize and lower vision encoder model for ET
#7532 commented on
Aug 30, 2025 • 0 new comments -
Add utils to replace torchtune SDPA with ET Custom SDPA
#7531 commented on
Aug 30, 2025 • 0 new comments -
Changes to SDPA to support no kv cache export
#7530 commented on
Aug 30, 2025 • 0 new comments -
support ET dump for llama3 runner
#7507 commented on
Aug 30, 2025 • 0 new comments -
Remove duplicated utilities in cadence backend
#7462 commented on
Aug 30, 2025 • 0 new comments -
Handle model state in benchmark CI jobs
#7432 commented on
Aug 30, 2025 • 0 new comments -
executorch combined diff
#7409 commented on
Aug 30, 2025 • 0 new comments -
Add additional default decompositions for upsample operators
#7402 commented on
Aug 30, 2025 • 0 new comments -
executorch/exir/tests
#7396 commented on
Aug 30, 2025 • 0 new comments -
executorch/exir/verification/test
#7395 commented on
Aug 30, 2025 • 0 new comments -
executorch/extension/android_test
#7394 commented on
Aug 29, 2025 • 0 new comments -
executorch/extension/llm/modules/test
#7393 commented on
Aug 29, 2025 • 0 new comments -
executorch/extension/pybindings/test
#7392 commented on
Aug 29, 2025 • 0 new comments -
executorch/extension/training/examples/XOR
#7391 commented on
Aug 29, 2025 • 0 new comments -
executorch/extension/training/pybindings/test
#7390 commented on
Aug 29, 2025 • 0 new comments -
executorch/profiler/test
#7389 commented on
Aug 29, 2025 • 0 new comments -
executorch/test/end2end
#7388 commented on
Aug 29, 2025 • 0 new comments -
executorch/test/models
#7387 commented on
Aug 29, 2025 • 0 new comments -
executorch/build/packaging
#7386 commented on
Aug 29, 2025 • 0 new comments -
executorch/backends/xnnpack/test/tester
#7385 commented on
Aug 29, 2025 • 0 new comments -
executorch/backends/xnnpack/partition/graphs
#7384 commented on
Aug 29, 2025 • 0 new comments -
executorch/backends/vulkan/test
#7383 commented on
Aug 29, 2025 • 0 new comments -
executorch/backends/qualcomm/utils
#7382 commented on
Aug 29, 2025 • 0 new comments -
Warn if wrong type is given for Llama export for XNNPACK
#8195 commented on
Aug 30, 2025 • 0 new comments -
[Do Not Merge] multi partitioned graph in one qnn delegate
#8175 commented on
Aug 30, 2025 • 0 new comments -
create MetaETDumpGen class to abstract copy_tensor_to_debug_buffer function
#8157 commented on
Aug 30, 2025 • 0 new comments -
.gitignores for ios demo artifacts
#8153 commented on
Aug 30, 2025 • 0 new comments -
support input_pos > 0 for prefill model
#8127 commented on
Aug 30, 2025 • 0 new comments -
Static llama kv
#8096 commented on
Aug 30, 2025 • 0 new comments -
Adjust tolerance for quantized XNN conv1d tests
#8093 commented on
Aug 29, 2025 • 0 new comments -
Revert to use mean_out than mean_dim_out
#8021 commented on
Aug 29, 2025 • 0 new comments -
remove the exec_aten namespace
#8018 commented on
Aug 29, 2025 • 0 new comments -
Fix comment in memory_planning.py
#8010 commented on
Aug 29, 2025 • 0 new comments -
fix spec_prop_pass
#7974 commented on
Aug 29, 2025 • 0 new comments -
Support (un)squeeze in XNNPACK delegate via conversion to view
#7961 commented on
Aug 30, 2025 • 0 new comments -
Conditionally support expand_copy in XNNPACK delegate
#7960 commented on
Aug 30, 2025 • 0 new comments -
Add view_copy/static_reshape support to XNNPACK delegate
#7959 commented on
Aug 30, 2025 • 0 new comments -
use dim order in all backend examples
#7953 commented on
Aug 30, 2025 • 0 new comments -
forward fix
#7871 commented on
Aug 30, 2025 • 0 new comments -
Fix export script to use pt2e if not selected
#7854 commented on
Aug 30, 2025 • 0 new comments -
Support Half/BFloat16 in native_group_norm (needs accuracy fix)
#7846 commented on
Aug 30, 2025 • 0 new comments -
[ET-VK] Using shared memory to save position in conv2d dw output op.
#7818 commented on
Aug 30, 2025 • 0 new comments -
Create a forked KV IO transformer for exporting coreML Llama
#7778 commented on
Aug 30, 2025 • 0 new comments -
Add build-time config for threadpool size
#7773 commented on
Aug 30, 2025 • 0 new comments -
[ET-VK] Tuning conv2d dw op batching size for improved performance.
#7597 commented on
Aug 30, 2025 • 0 new comments -
[ET-VK] Fixing conv2d dw incorrect output when stride != dilation issue.
#7571 commented on
Aug 30, 2025 • 0 new comments -
Switch from print_exported_headers to CMake's installed headers in scripts/build_apple_frameworks.sh
#13559 commented on
Aug 25, 2025 • 0 new comments -
fix and link
#11267 commented on
Sep 1, 2025 • 0 new comments -
Don't depend on `aten_headers_for_executorch` when running with HTP simulator since it already gets this dep from xplat
#11260 commented on
Sep 1, 2025 • 0 new comments -
Remove outdated editable mode warning from building from source docs
#11245 commented on
Sep 1, 2025 • 0 new comments -
[ET-VK] De vectorizing sum and moving bias application to the end in conv 2d op to improve performance.
#11242 commented on
Sep 1, 2025 • 0 new comments -
[ET-VK] De vectorizing sum and moving bias application to the end in conv 2d op to improve performance.
#11241 commented on
Sep 1, 2025 • 0 new comments -
[ET-VK] De vectorizing sum and moving bias application to the end in conv 2d op to improve performance.
#11236 commented on
Sep 1, 2025 • 0 new comments -
[ET-VK] De vectorizing sum and moving bias application to the end in conv 2d op to improve performance.
#11235 commented on
Sep 1, 2025 • 0 new comments -
Expose gen_oplist_lib to clients
#11231 commented on
Sep 1, 2025 • 0 new comments -
Export recipes integration in export_llama
#11228 commented on
Sep 1, 2025 • 0 new comments -
Explicitly Add 'DEXECUTORCH_BUILD_EXTENSION_TENSOR' arg when 'xnnpack…
#11173 commented on
Sep 1, 2025 • 0 new comments -
Enable all platform tests in CMake
#11124 commented on
Sep 1, 2025 • 0 new comments -
Use GraphBuilder in reorder unit tests.
#11120 commented on
Sep 1, 2025 • 0 new comments -
Android input and output tag JNI
#11057 commented on
Sep 1, 2025 • 0 new comments -
Move get_quantizer_and_quant_params to quantizer_lib
#11056 commented on
Sep 1, 2025 • 0 new comments -
add extension training to android builds
#11003 commented on
Sep 1, 2025 • 0 new comments -
Pupolate METADATA.bzl with developer responses
#10976 commented on
Sep 1, 2025 • 0 new comments -
[pytorch hash update] update the pinned pytorch hash
#10955 commented on
Aug 31, 2025 • 0 new comments -
Fix lint
#10916 commented on
Sep 1, 2025 • 0 new comments -
[ET-VK] Adding width packing support for Conv2d pw
#10909 commented on
Sep 1, 2025 • 0 new comments -
solve broken devtool workflow
#10885 commented on
Sep 1, 2025 • 0 new comments -
Forward fix on NXP backend
#10829 commented on
Sep 1, 2025 • 0 new comments -
[ET-VK] Moving device capabilities check to DispatchNode and PrepackNode ctor.
#10785 commented on
Sep 1, 2025 • 0 new comments -
[ET-VK] Reducing memory wastage by tightening DescriptorPoolConfig values.
#10784 commented on
Sep 1, 2025 • 0 new comments -
install headers from runtime/backend
#13558 commented on
Aug 25, 2025 • 0 new comments -
Let's not expose the underlying Method.
#13543 commented on
Aug 26, 2025 • 0 new comments -
Update deprecated pkg_resources api in edge._ops
#13513 commented on
Aug 29, 2025 • 0 new comments -
NXP backend: Refactor executorch_pipeline.py, allow different sources for calibration inputs
#13494 commented on
Aug 29, 2025 • 0 new comments -
Arm backend: Add decomposition and test for fmod
#13456 commented on
Aug 26, 2025 • 0 new comments -
Arm backend: Preserve output order
#13454 commented on
Aug 29, 2025 • 0 new comments -
im2row HiFi optimizations
#13431 commented on
Aug 29, 2025 • 0 new comments -
NXP backend: Remove optimization in remove_unused_tensors_and_buffers.py
#13415 commented on
Aug 29, 2025 • 0 new comments -
NXP backend: Remove optimization in prune_quantize_operators.py
#13409 commented on
Aug 29, 2025 • 0 new comments -
Qualcomm AI Engine Direct - Test QNN2.37 uplevel
#13352 commented on
Aug 31, 2025 • 0 new comments -
Bump qnn to 2.37
#13333 commented on
Aug 29, 2025 • 0 new comments -
Arm backend: Bump Arm baremetal compilers to gcc 14.3
#13317 commented on
Aug 28, 2025 • 0 new comments -
Try to fix arm CI jobs
#13292 commented on
Aug 25, 2025 • 0 new comments -
Fix MediaTek CI job
#13291 commented on
Aug 25, 2025 • 0 new comments -
RPATH Fix for portable_lib Python Extension
#13290 commented on
Aug 25, 2025 • 0 new comments -
Arm backend: Update VelaIO handling
#13282 commented on
Aug 27, 2025 • 0 new comments -
Experimental GGUF-2-PTE Converter
#13266 commented on
Aug 27, 2025 • 0 new comments -
Add AOTI backend skeleton code
#13123 commented on
Aug 28, 2025 • 0 new comments -
Add initial backends/cadence/vision module scaffold with optimized softmax kernel (no iDMA)
#12480 commented on
Aug 28, 2025 • 0 new comments -
PAL File for Arm BareMetal Executor Runner
#12399 commented on
Aug 29, 2025 • 0 new comments -
[Draft] Qualcomm AI Engine Direct - Unexpected graph for mutable buffer after export during Quantization
#11309 commented on
Sep 1, 2025 • 0 new comments -
Device telemetry for benchmark
#11301 commented on
Sep 1, 2025 • 0 new comments -
Add where layer to ops registry
#11277 commented on
Sep 1, 2025 • 0 new comments -
[ET-VK] Removing descriptor pool intialization from DescriptorPool ctor.
#10777 commented on
Sep 1, 2025 • 0 new comments -
fix tabular output
#10271 commented on
Sep 1, 2025 • 0 new comments -
fix tabular output
#10246 commented on
Aug 31, 2025 • 0 new comments -
qnn runner: add memory consumption logging
#10237 commented on
Aug 31, 2025 • 0 new comments -
Move benchmarking workflow cli from testinfra to executorch
#10162 commented on
Aug 31, 2025 • 0 new comments -
[#9971] Gracefully error out in ETDump part 3 for *profiling_delegate
#10147 commented on
Aug 31, 2025 • 0 new comments -
[TEST] Try to build Android C++ one pass
#10124 commented on
Aug 31, 2025 • 0 new comments -
[ET-VK] Minor improvement to permute op.
#10117 commented on
Aug 31, 2025 • 0 new comments -
[ET-VK] Modify quantized linear naive shader to linearly dispatch work to improve performance.
#10116 commented on
Aug 31, 2025 • 0 new comments -
[NOT FOR LAND] Build AAR with QNN
#10081 commented on
Aug 31, 2025 • 0 new comments -
[WIP] Mimi: export with streaming
#10044 commented on
Aug 31, 2025 • 0 new comments -
Add coreml recipes
#10037 commented on
Aug 31, 2025 • 0 new comments -
Add some basic xnnpack recipes
#10035 commented on
Aug 31, 2025 • 0 new comments -
[ET-VK] Tuning native layer norm local workgroup size to improve thread occupancy during reduce.
#9984 commented on
Aug 31, 2025 • 0 new comments -
Removing custom RMSNorm support
#9966 commented on
Aug 31, 2025 • 0 new comments -
[WIP] Devtool end-to-end tests
#9925 commented on
Aug 31, 2025 • 0 new comments -
[ET-VK] Minor performance improvements to native layer norm.
#9892 commented on
Aug 31, 2025 • 0 new comments -
[WIP] Mimi 4-bit quant on transformer and 8-bit on conv
#9882 commented on
Aug 31, 2025 • 0 new comments -
to benchmark chunked long context prefill
#9779 commented on
Aug 31, 2025 • 0 new comments -
Validate that method inputs have the expected dim order
#9757 commented on
Aug 31, 2025 • 0 new comments -
use kUnsetDebugHandle for unset debug handle in etdump delegation log
#9754 commented on
Aug 31, 2025 • 0 new comments -
smart_ptr multi-turn demo
#9743 commented on
Aug 31, 2025 • 0 new comments -
Remove unecessary tests from pull
#9738 commented on
Aug 31, 2025 • 0 new comments -
Back out "Add method variants in dynamic shim to use core tensor/evalue/result types"
#9719 commented on
Aug 31, 2025 • 0 new comments -
Prim ops move 2
#10763 commented on
Sep 1, 2025 • 0 new comments -
move pattern
#10756 commented on
Sep 1, 2025 • 0 new comments -
[TEST] Split prim ops into its own
#10741 commented on
Sep 1, 2025 • 0 new comments -
openvino_backend doesn't need to be static only
#10732 commented on
Sep 1, 2025 • 0 new comments -
Fix preq embedding dtype check
#10699 commented on
Sep 1, 2025 • 0 new comments -
Clean up eager quant in llm_export
#10684 commented on
Sep 1, 2025 • 0 new comments -
[executorch][android] Add Runtime.java to centralize native library l…
#10672 commented on
Sep 1, 2025 • 0 new comments -
Enable do_quant_fusion_and_const_prop by default
#10633 commented on
Sep 1, 2025 • 0 new comments -
Add a pass to fuse scalar mul with quant ops
#10630 commented on
Sep 1, 2025 • 0 new comments -
Hack vulkan so
#10565 commented on
Sep 1, 2025 • 0 new comments -
[MPS] Add portable grid_sampler_2d implementation + tests
#10561 commented on
Aug 27, 2025 • 0 new comments -
Allow removing permute pairs in addition to transpose pairs
#10501 commented on
Sep 1, 2025 • 0 new comments -
Increase max try in llm benchmark
#10500 commented on
Sep 1, 2025 • 0 new comments -
Experiment so
#10498 commented on
Sep 1, 2025 • 0 new comments -
Fix `numel()` downcast in dper_lib/silvertorch/core/legacy/tools/eval/tests/TestUtil.cpp +2
#10420 commented on
Sep 1, 2025 • 0 new comments -
Fix `numel()` downcast in executorch/backends/vulkan/test/utils/test_utils.cpp +2
#10419 commented on
Sep 1, 2025 • 0 new comments -
always turn on dynamo for map (#150962)
#10409 commented on
Sep 1, 2025 • 0 new comments -
Add test_qnn_delegates.py to oss ci
#10377 commented on
Sep 1, 2025 • 0 new comments -
Rename some "jarvis" instances into "falcon" or "cadence"
#10354 commented on
Sep 1, 2025 • 0 new comments -
Add CI for conv_former and fastvit for QNN
#10282 commented on
Sep 1, 2025 • 0 new comments -
Fix undefined fht_float in Apple OS
#10280 commented on
Sep 1, 2025 • 0 new comments -
Move default Vela/Regor configurations to Sram_Only
#10279 commented on
Sep 1, 2025 • 0 new comments -
[exir] Refactor EdgeProgramManager.transform
#10275 commented on
Sep 1, 2025 • 0 new comments -
Use a symmetric quantization with no clipping error to improve llama perplexity
#5163 commented on
Aug 27, 2025 • 0 new comments -
Add necessary third party libs
#3768 commented on
Aug 27, 2025 • 0 new comments -
Update deps paths
#3767 commented on
Aug 27, 2025 • 0 new comments -
Support bare-metal Windows
#3735 commented on
Aug 27, 2025 • 0 new comments -
let coreml leverage all backends
#3683 commented on
Aug 27, 2025 • 0 new comments -
migrate from capture_pre_autograd_graph to export(pre_dispatch=True)
#3671 commented on
Aug 27, 2025 • 0 new comments -
checking "transposed convolution" at beginning
#3631 commented on
Aug 27, 2025 • 0 new comments -
list all ops
#3630 commented on
Aug 27, 2025 • 0 new comments -
debug dim order op
#3579 commented on
Aug 27, 2025 • 0 new comments -
Add permute operator to vulkan_partitioner.py
#3547 commented on
Aug 26, 2025 • 0 new comments -
Mps qlinear
#3527 commented on
Aug 26, 2025 • 0 new comments -
Delete deprecated QNNPack Backend
#3508 commented on
Aug 26, 2025 • 0 new comments -
remove exir.capture from test_pass_infra
#3505 commented on
Aug 26, 2025 • 0 new comments -
reconcile Dim4D and NchwDim
#3428 commented on
Aug 26, 2025 • 0 new comments -
`bmm` and `baddbmm`
#3403 commented on
Aug 26, 2025 • 0 new comments -
Remove unused variables in eki/builder/offloading/BlockAllocator.cpp
#3017 commented on
Aug 26, 2025 • 0 new comments -
Add a Python Wrapper that parses logits from llama_runner eval_mode
#2861 commented on
Aug 26, 2025 • 0 new comments -
provide an option to not using sdpa
#2810 commented on
Aug 26, 2025 • 0 new comments -
Add evaluation mode for llama_runner
#2804 commented on
Aug 26, 2025 • 0 new comments -
NMS op implementation for ExecuTorch in portable mode
#2787 commented on
Aug 26, 2025 • 0 new comments -
Print error message text for a specified error number
#2780 commented on
Aug 26, 2025 • 0 new comments -
make install requirements work with older versions of bash
#2762 commented on
Aug 26, 2025 • 0 new comments -
Add etdump generation to llama runner
#2759 commented on
Aug 26, 2025 • 0 new comments -
[Executorch][llama] Fix android app build
#2747 commented on
Aug 26, 2025 • 0 new comments -
revert mask definition to avoid full_like
#2684 commented on
Aug 26, 2025 • 0 new comments -
Revert "Revert "Add proper pt2e calibration""
#5148 commented on
Aug 27, 2025 • 0 new comments -
unblock hifi build
#5115 commented on
Aug 27, 2025 • 0 new comments -
Try migrating llama2 to the new subclass based API from old API (Int8DynActInt4WeightQuantizer)
#5106 commented on
Aug 27, 2025 • 0 new comments -
Create a MTK Runner to be able to run with a mobile app
#5067 commented on
Aug 27, 2025 • 0 new comments -
Training Module
#4974 commented on
Aug 27, 2025 • 0 new comments -
[Build] fix file_data_loader.cpp build issues for windows
#4899 commented on
Aug 27, 2025 • 0 new comments -
Pure pytorch llava
#4895 commented on
Aug 27, 2025 • 0 new comments -
Adding sdpa_with_attn_bias
#4861 commented on
Aug 27, 2025 • 0 new comments -
temp PR
#4691 commented on
Aug 27, 2025 • 0 new comments -
prefill model
#4682 commented on
Aug 27, 2025 • 0 new comments -
[coreml] Remove references to build_apple_frameworks.sh "--Release" flag
#4638 commented on
Aug 27, 2025 • 0 new comments -
Replace runners prefix amz2023.
#4420 commented on
Aug 27, 2025 • 0 new comments -
Implement aten.meshgrid.default with 2 input tensors
#4341 commented on
Aug 27, 2025 • 0 new comments -
Retire deep copy of edge_program in to_backend method
#4311 commented on
Aug 27, 2025 • 0 new comments -
Use Helios' decomposition for SDPA before quantizing
#4194 commented on
Aug 27, 2025 • 0 new comments -
Propagate dump_graphs to print pre-quantization graphs.
#4167 commented on
Aug 27, 2025 • 0 new comments -
Remove the if-else branch for dynamic shape
#4151 commented on
Aug 27, 2025 • 0 new comments -
Pyre cleanups in Cadence backends folder
#4138 commented on
Aug 27, 2025 • 0 new comments -
Port ArchProbe as GPUInfo - Initial Skeleton
#4133 commented on
Aug 27, 2025 • 0 new comments -
Expose AIBench logs
#4104 commented on
Aug 27, 2025 • 0 new comments -
backend compatibility api
#3913 commented on
Aug 27, 2025 • 0 new comments -
Llava: add image preprocessing into the model and export it
#3912 commented on
Aug 27, 2025 • 0 new comments -
Complete revamp of float/promotion sympy handling
#3911 commented on
Aug 27, 2025 • 0 new comments -
rename op lib to be clearer on its usage
#3776 commented on
Aug 27, 2025 • 0 new comments -
[executorch][PR] Enable blas and add eigen blas
#2655 commented on
Aug 26, 2025 • 0 new comments -
[pybind] install ATen mode pybindings
#1554 commented on
Aug 26, 2025 • 0 new comments -
[Executorch][portable] Fix portable kernel utils for aten mode
#1553 commented on
Aug 26, 2025 • 0 new comments -
Remove explicit supports_static_listing key
#1361 commented on
Aug 26, 2025 • 0 new comments -
Example kernel for '[RFC] Scope for Custom Operator extension'
#1358 commented on
Aug 26, 2025 • 0 new comments -
[RFC] Scope for Custom Operator extension
#1357 commented on
Aug 26, 2025 • 0 new comments -
[MPS] Add support for conv1D (fixes w2l)
#1316 commented on
Aug 26, 2025 • 0 new comments -
Fail to allocate temp memory by exporting torch.topk
#8700 commented on
Aug 28, 2025 • 0 new comments -
This document（https://pytorch.org/executorch/stable/demo-apps-android.html#running-the-app） is out of date. Where is examples/demo-apps/android/ExecuTorchDemo?
#10314 commented on
Aug 28, 2025 • 0 new comments -
[CortexM] Stateful Ops (Linear) with CMSIS-NN integration with unit Testcases
#13507 commented on
Aug 28, 2025 • 0 new comments -
run mtk_build_examples.sh error
#11999 commented on
Aug 28, 2025 • 0 new comments -
CoreML model is crashing on iPhone GPU, but not on iPhone CPU or macOS GPU
#11541 commented on
Aug 28, 2025 • 0 new comments -
Issues with deloyment on RP2040
#7177 commented on
Aug 28, 2025 • 0 new comments -
[RFC] Canonicalize python file structure
#8699 commented on
Aug 28, 2025 • 0 new comments -
Label tracking meta-issue (edit me to get automatically CC'ed on issues! cc bot)
#7679 commented on
Aug 26, 2025 • 0 new comments -
Incorrect Observer Sharing/Derivation at Conv-ReLU+ Residual with Arm Ethos Quantizer
#12959 commented on
Aug 26, 2025 • 0 new comments -
QNN Transposed Conv2d with Dilation appears Incorrect
#13611 commented on
Aug 26, 2025 • 0 new comments -
Unsupported node types fail to lower with “KeyError”
#13604 commented on
Aug 26, 2025 • 0 new comments -
AOTI backend support
#13188 commented on
Aug 26, 2025 • 0 new comments -
NXP: Enable mypy python type checker
#13518 commented on
Aug 26, 2025 • 0 new comments -
QNN Avg and Max pool with ceil_mode=True appear to have incorrect outputs
#13612 commented on
Aug 26, 2025 • 0 new comments -
All-dim reduction operators fail with “IndexError: tuple index out of range”
#13608 commented on
Aug 26, 2025 • 0 new comments -
"How to Support a Custom Model in HTP Backend" example code is out of date
#13607 commented on
Aug 26, 2025 • 0 new comments -
CMake build fails with test errors when tests are not enabled
#13338 commented on
Aug 26, 2025 • 0 new comments -
Missing grid_sample_2d portable kernel
#13509 commented on
Aug 25, 2025 • 0 new comments -
Move Quant functions to a new file quant.py: Part 2
#2646 commented on
Aug 26, 2025 • 0 new comments -
Move Quant functions out to a new file quant.py: Part 1
#2644 commented on
Aug 26, 2025 • 0 new comments -
temp2
#2639 commented on
Aug 26, 2025 • 0 new comments -
temp1
#2638 commented on
Aug 26, 2025 • 0 new comments -
Enable blas and add eigen blas
#2627 commented on
Aug 26, 2025 • 0 new comments -
estimate model parameters
#2612 commented on
Aug 26, 2025 • 0 new comments -
Replace view-like ops with view ops
#2592 commented on
Aug 26, 2025 • 0 new comments -
fbcode
#2591 commented on
Aug 26, 2025 • 0 new comments -
Only add copy_ if necessary for correctness
#2556 commented on
Aug 26, 2025 • 0 new comments -
export coreml kv-cache without dynamic shape
#2543 commented on
Aug 26, 2025 • 0 new comments -
export mps kv-cache without dynamic shape
#2542 commented on
Aug 26, 2025 • 0 new comments -
kv cache as buffers in llama2
#2527 commented on
Aug 26, 2025 • 0 new comments -
Eval_llama clean up
#2523 commented on
Aug 26, 2025 • 0 new comments -
pass to remove auto_functionalized HOP
#2522 commented on
Aug 26, 2025 • 0 new comments -
quantize.py - useregister_parameter in lieu of register_buffer
#2494 commented on
Aug 26, 2025 • 0 new comments -
Fix llama quantize_per_token numerics
#2167 commented on
Aug 26, 2025 • 0 new comments -
Ban tensor constants in tracing.
#1844 commented on
Aug 26, 2025 • 0 new comments -
Fix use_static_deps
#1834 commented on
Aug 26, 2025 • 0 new comments -
Define meta function for embedding_byte
#1787 commented on
Aug 26, 2025 • 0 new comments -
Mutable Buffer Support in ExecuTorch
#1786 commented on
Aug 26, 2025 • 0 new comments -
[Executorch][llama2] Remove redundant xnnpack path
#1780 commented on
Aug 26, 2025 • 0 new comments -
Fix #pyre-strict lint error in two files and add type error comments
#1744 commented on
Aug 26, 2025 • 0 new comments -
Replace `constraints` with `dynamic_shapes` in executorch
#1676 commented on
Aug 26, 2025 • 0 new comments -
Replace `constraints` with `dynamic_shapes` in export-to-executorch tutorial
#1637 commented on
Aug 26, 2025 • 0 new comments -
executorch/backends/qualcomm/tests
#7381 commented on
Aug 29, 2025 • 0 new comments -
Bump torchfix from 0.5.0 to 0.7.0
#6943 commented on
Aug 29, 2025 • 0 new comments -
Export Stories110M model to compare on-device numerics
#6939 commented on
Aug 29, 2025 • 0 new comments -
[EE/BE] Adding legacy partitioner tests
#6920 commented on
Aug 29, 2025 • 0 new comments -
[EE/BE] Adding the legacy flow tests to abs op
#6916 commented on
Aug 29, 2025 • 0 new comments -
[EE/BE] T206498038 Adding the legacy flow tests to add op
#6907 commented on
Aug 29, 2025 • 0 new comments -
Coreml prefill
#6895 commented on
Aug 29, 2025 • 0 new comments -
[pybind] Do not set output data pointer if the output is memory planned
#6841 commented on
Aug 29, 2025 • 0 new comments -
Explicitly pass buffer sizes during memory planning when control flow submodule are around
#6840 commented on
Aug 29, 2025 • 0 new comments -
Move the rest qnn model jobs to the new qnn sdk docker
#6818 commented on
Aug 29, 2025 • 0 new comments -
share embeddding and output
#6800 commented on
Aug 29, 2025 • 0 new comments -
Skip embedding
#6777 commented on
Aug 28, 2025 • 0 new comments -
noop
#6739 commented on
Aug 28, 2025 • 0 new comments -
[Executorch][Optimized] Use portable's impl for optimized op_add's fallback
#6737 commented on
Aug 28, 2025 • 0 new comments -
[Exeuctorch][Portale] Refactor op_sub to be reused by optimized fallback
#6736 commented on
Aug 28, 2025 • 0 new comments -
[Executorch][Optimized] Use portable's impl for optimized op_add's fallback
#6735 commented on
Aug 28, 2025 • 0 new comments -
[Exeuctorch][Portale] Refactor op_add to be reused by optimized fallback
#6734 commented on
Aug 28, 2025 • 0 new comments -
[Exeuctorch][Optimized] Use portable path for fallback in op_mul
#6733 commented on
Aug 28, 2025 • 0 new comments -
[Executorch][portable] Split op_mul in impl plus op
#6732 commented on
Aug 28, 2025 • 0 new comments -
[Executorch] Renable operator optimization flags
#6715 commented on
Aug 28, 2025 • 0 new comments -
[Executorch][optimized] Fix op_div impl to use portable for fallback path
#6714 commented on
Aug 28, 2025 • 0 new comments -
Refactor op_div for shreability with optimized op_div
#6713 commented on
Aug 28, 2025 • 0 new comments -
export_for_training
#6698 commented on
Aug 28, 2025 • 0 new comments -
[do not land] xnnpack io
#6652 commented on
Aug 28, 2025 • 0 new comments -
adding suppression tags to improve autodeps noise
#6645 commented on
Aug 28, 2025 • 0 new comments -
executorch/backends/example
#7380 commented on
Aug 29, 2025 • 0 new comments -
executorch/backends/apple/mps/test
#7379 commented on
Aug 29, 2025 • 0 new comments -
executorch/backends/cadence/aot
#7378 commented on
Aug 29, 2025 • 0 new comments -
executorch/backends/apple/coreml/test
#7377 commented on
Aug 29, 2025 • 0 new comments -
executorch/backends/apple/coreml/runtime/test
#7376 commented on
Aug 29, 2025 • 0 new comments -
PATCH
#7370 commented on
Aug 29, 2025 • 0 new comments -
Changes to sdpa and attention module to support vision encoder attent…
#7350 commented on
Aug 29, 2025 • 0 new comments -
Update target path for CI macros 1/N
#7348 commented on
Aug 29, 2025 • 0 new comments -
[ET-VK] Adding MemoryPool class to allocate chunks of memory and use it for smaller allocations.
#7338 commented on
Aug 29, 2025 • 0 new comments -
Add type error suppressions for upcoming upgrade
#7323 commented on
Aug 29, 2025 • 0 new comments -
Use macro `XT_KERNEL_CHECK` to handle errors returned by nnlib.
#7312 commented on
Aug 29, 2025 • 0 new comments -
Xtensa ISS PAL layer for logging/timing.
#7311 commented on
Aug 29, 2025 • 0 new comments -
Allow backends/cadence to use TestUtil.h.
#7304 commented on
Aug 29, 2025 • 0 new comments -
[Not for landing] coreml llama model
#7295 commented on
Aug 29, 2025 • 0 new comments -
Add legacy flow tests rebase
#7286 commented on
Aug 29, 2025 • 0 new comments -
test only
#7283 commented on
Aug 29, 2025 • 0 new comments -
Remove unused pyre ignores
#7253 commented on
Aug 29, 2025 • 0 new comments -
coreml attention tests
#7146 commented on
Aug 29, 2025 • 0 new comments -
add support to evalulate the model with attention sink
#7033 commented on
Aug 29, 2025 • 0 new comments -
Trigger wheel builds when modifying setup.py or pyproject.toml
#7017 commented on
Aug 29, 2025 • 0 new comments -
Update nightly pin
#7010 commented on
Aug 29, 2025 • 0 new comments -
[cadence][7/X] add remove ops to oss and callsites
#6982 commented on
Aug 29, 2025 • 0 new comments -
[cadence][7/X] add remove ops to oss and callsites
#6981 commented on
Aug 29, 2025 • 0 new comments -
[cadence][7/X] add remove ops to oss and callsites
#6980 commented on
Aug 29, 2025 • 0 new comments -
[TEMP] Enable training IR
#6633 commented on
Aug 28, 2025 • 0 new comments -
Copy the val tensor to the newly inserted getitem nodes in `_fixup_output_node`
#5744 commented on
Aug 28, 2025 • 0 new comments -
Rename CMake EXECUTORCH_BUILD_KERNELS_CUSTOM to EXECUTORCH_BUILD_EXTENSION_LLM
#5743 commented on
Aug 28, 2025 • 0 new comments -
Add a script to help user enable ccache
#5624 commented on
Aug 28, 2025 • 0 new comments -
`executorch::cpublas::internal::reduce` should return float
#5518 commented on
Aug 28, 2025 • 0 new comments -
We want to ensure that all python targets are autodeps compliant
#5465 commented on
Aug 28, 2025 • 0 new comments -
Fix various using namespace issues
#5463 commented on
Aug 28, 2025 • 0 new comments -
Remove _preserve_ops from export
#5455 commented on
Aug 28, 2025 • 0 new comments -
maybe fix conv.backward for joint graph export
#5450 commented on
Aug 28, 2025 • 0 new comments -
fix joint graph metadata (#136011)
#5447 commented on
Aug 28, 2025 • 0 new comments -
Update executorch pin to the newest nightly
#5428 commented on
Aug 28, 2025 • 0 new comments -
Adding executorch_prim::mod.Scalar
#5389 commented on
Aug 27, 2025 • 0 new comments -
Fix fbcode//executorch/backends/arm/quantizer:arm_quantizer type checking test
#5378 commented on
Aug 27, 2025 • 0 new comments -
Add type error suppressions for upcoming upgrade
#5352 commented on
Aug 27, 2025 • 0 new comments -
Add type error suppressions for upcoming upgrade
#5321 commented on
Aug 27, 2025 • 0 new comments -
Add type error suppressions for upcoming upgrade
#5314 commented on
Aug 27, 2025 • 0 new comments -
Add type error suppressions for upcoming upgrade
#5300 commented on
Aug 27, 2025 • 0 new comments -
Remove preserve_ops and consolidate decomp_table
#5295 commented on
Aug 27, 2025 • 0 new comments -
Remve _preserve_ops and consolidate with decomp_table
#5281 commented on
Aug 27, 2025 • 0 new comments -
Revert D59990127
#5276 commented on
Aug 27, 2025 • 0 new comments -
E2E decoding with Seamless
#5262 commented on
Aug 27, 2025 • 0 new comments -
updated hifi to use the hifi kernel
#5198 commented on
Aug 27, 2025 • 0 new comments -
Add type error suppressions for upcoming upgrade
#5179 commented on
Aug 27, 2025 • 0 new comments -
Add mtk_llama_executor_runner instructions and MTK Backend Readme Clarifications
#5173 commented on
Aug 27, 2025 • 0 new comments -
[Windows Build] Implement MMAP for mmap_data_loader.cpp
#5164 commented on
Aug 27, 2025 • 0 new comments -
linked sub mul div tanh sigmoid from oss
#6619 commented on
Aug 28, 2025 • 0 new comments -
Add UInt16 support to scalar_type.h
#6614 commented on
Aug 28, 2025 • 0 new comments -
coreml io
#6582 commented on
Aug 28, 2025 • 0 new comments -
Update cpuinfo to 8df44962
#6562 commented on
Aug 28, 2025 • 0 new comments -
[not for landing] perf-repro
#6551 commented on
Aug 28, 2025 • 0 new comments -
Cleanup executorch::aten::... for cadence operators.
#6481 commented on
Aug 28, 2025 • 0 new comments -
Use curl in bug-report.yml
#6476 commented on
Aug 28, 2025 • 0 new comments -
Try to trigger android test only if necessary
#6448 commented on
Aug 28, 2025 • 0 new comments -
add all users for now
#6405 commented on
Aug 28, 2025 • 0 new comments -
Redirect stdout and stderr from C++ stream to Python stream
#6390 commented on
Aug 28, 2025 • 0 new comments -
Remove custom implementation of optional
#6366 commented on
Aug 28, 2025 • 0 new comments -
executorch/backends/cadence/reference/operators: fix llvm-17-exposed format mismatches
#6332 commented on
Aug 28, 2025 • 0 new comments -
Add tests to Cadence examples
#6233 commented on
Aug 28, 2025 • 0 new comments -
[no commit] coreml test
#6172 commented on
Aug 28, 2025 • 0 new comments -
introduce ambiguous dim order check
#6147 commented on
Aug 28, 2025 • 0 new comments -
TEMP way to run export_for_training
#6136 commented on
Aug 28, 2025 • 0 new comments -
Add allocate tensor util that uses temp allocator
#6090 commented on
Aug 28, 2025 • 0 new comments -
[AOSP] Android.bp initial commit
#6077 commented on
Aug 28, 2025 • 0 new comments -
Enable export_for_training for all executorch/backends
#6035 commented on
Aug 28, 2025 • 0 new comments -
Try to use macos-m1-stable
#6003 commented on
Aug 28, 2025 • 0 new comments -
TEMP
#5973 commented on
Aug 28, 2025 • 0 new comments -
Enable Half/BF16: abs, full, gelu, hardtanh, logit, neg, sign
#5856 commented on
Aug 28, 2025 • 0 new comments -
Extend preservable op list to custom ops
#5850 commented on
Aug 28, 2025 • 0 new comments -
Integrating Axis Mapping into Quantized Linear Texture Implementation
#5758 commented on
Aug 28, 2025 • 0 new comments