Add Arduino library support for ExecuTorch#20221
Open
psiddh wants to merge 2 commits into
Open
Conversation
Add tooling to package the ExecuTorch runtime as an Arduino library, enabling PyTorch model inference on Arduino microcontrollers. The library vendors ET runtime sources, CMSIS-NN kernels, and portable ops into a self-contained package that compiles under the Arduino build system. Key components: - `build_arduino_library.sh` assembles the distributable library from repository sources (no vendored copies checked in) - `ExecuTorchArduino.h` configures the build environment for Arduino (fixes for std::variant, cmake_macros.h stub, build defines) - `platform_stubs.c` provides C library stubs for the LLEXT environment - Example sketches using the native ExecuTorch C++ API (no wrapper layer) - Zephyr board config for Arduino Uno Q (STM32U585, Cortex-M33) Validated on Arduino Uno Q with DS-CNN keyword spotting model (int8, CMSIS-NN): 390+ source files compile, 106 KB flash (13%), 91 KB RAM. Authored with assistance from Claude.
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/20221
Note: Links to docs will display an error until the docs builds have been completed. ❌ 2 New Failures, 15 Pending, 1 Unrelated Failure, 7 Unclassified FailuresAs of commit 2b37444 with merge base 1bf982a ( NEW FAILURES - The following jobs have failed:
UNCLASSIFIED FAILURES - DrCI could not classify the following jobs because the workflow did not run on the merge base. The failures may be pre-existing on trunk or introduced by this PR:
FLAKY - The following job failed but was likely due to flakiness present on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
This PR needs a
|
- Add Meta copyright headers to all files - HelloExecuTorch: uses core ET runtime only (portable ops, no backend) - KeywordSpotting: uses Cortex-M backend with CMSIS-NN accelerated ops - Clarify the distinction in sketch comments
Contributor
There was a problem hiding this comment.
Pull request overview
This PR adds tooling and example code to package ExecuTorch as an installable Arduino library, with support for building/running on Arduino-class microcontrollers (including a Zephyr-based Arduino Uno Q configuration).
Changes:
- Add
build_arduino_library.shto assemble a self-contained Arduino library by copying ExecuTorch sources + required third-party headers. - Introduce Arduino library metadata/header glue (
library.properties,ExecuTorchArduino.h) plus C-library stubs for the target environment. - Add Arduino example sketches and a Zephyr board config tuned for Uno Q.
Reviewed changes
Copilot reviewed 9 out of 9 changed files in this pull request and generated 9 comments.
Show a summary per file
| File | Description |
|---|---|
| zephyr/samples/hello-executorch/boards/arduino_uno_q.conf | Adds Uno Q memory sizing defaults for ExecuTorch allocators/stack/heap. |
| examples/arduino/README.md | Documents the Arduino library packaging flow and usage. |
| examples/arduino/platform_stubs.c | Adds libc symbol stubs intended for Zephyr LLEXT-style environments. |
| examples/arduino/library.properties | Defines Arduino library metadata (name/version/architectures/includes). |
| examples/arduino/ExecuTorchArduino.h | Provides Arduino-specific compile configuration and exposes the native ExecuTorch C++ API. |
| examples/arduino/examples/KeywordSpotting/KeywordSpotting.ino | Adds a DS-CNN keyword spotting sketch using the native ExecuTorch API. |
| examples/arduino/examples/HelloExecuTorch/HelloExecuTorch.ino | Adds a minimal “runtime_init()” sanity-check sketch. |
| examples/arduino/build_arduino_library.sh | Adds the library assembly script (copies sources, generates stubs, applies patches). |
| examples/arduino/.gitignore | Ignores generated Arduino library output directory. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+9
to
+12
| void _Exit(int status) { | ||
| (void)status; | ||
| while (1) {} | ||
| } |
Comment on lines
+14
to
+18
| int fprintf(FILE* stream, const char* fmt, ...) { | ||
| (void)stream; | ||
| (void)fmt; | ||
| return 0; | ||
| } |
Comment on lines
+72
to
+80
| // Allocate planned buffers | ||
| size_t n_planned = meta->num_memory_planned_buffers(); | ||
| Span<uint8_t>* spans = static_cast<Span<uint8_t>*>( | ||
| method_allocator.allocate(n_planned * sizeof(Span<uint8_t>))); | ||
| for (size_t i = 0; i < n_planned; i++) { | ||
| size_t sz = static_cast<size_t>(meta->memory_planned_buffer_size(i).get()); | ||
| uint8_t* buf = static_cast<uint8_t*>(method_allocator.allocate(sz)); | ||
| spans[i] = {buf, sz}; | ||
| } |
Comment on lines
+103
to
+106
| // Read output | ||
| EValue output; | ||
| method->get_outputs(&output, 1); | ||
| if (output.isTensor()) { |
Comment on lines
+15
to
+16
| #include "model.h" | ||
| #include "test_inputs.h" |
Comment on lines
+3
to
+5
| // Runs a quantized DS-CNN model (MLPerf Tiny KWS benchmark) on hardcoded | ||
| // MFCC test inputs and prints the detected keyword. The model classifies | ||
| // 12 keywords: silence, unknown, yes, no, up, down, left, right, on, off, |
Comment on lines
+93
to
+103
| # Schema headers (generated — need a prior cmake build) | ||
| mkdir -p "$ET_SRC/schema" | ||
| cp "$ET_ROOT/schema/"*.h "$ET_SRC/schema/" 2>/dev/null || true | ||
| # Look for generated headers in common build dirs | ||
| for build_dir in "$ET_ROOT/cmake-out" "$ET_ROOT/cmake-out-mac" \ | ||
| "$ET_ROOT/outputs/build_uno_q"; do | ||
| if [ -d "$build_dir/schema/include/executorch/schema" ]; then | ||
| cp "$build_dir/schema/include/executorch/schema/"*.h "$ET_SRC/schema/" | ||
| break | ||
| fi | ||
| done |
Comment on lines
+165
to
+167
| 2. **`cmake_macros.h` stub** — c10/torch headers expect a cmake-generated | ||
| file. The stub defines `C10_USING_CUSTOM_GENERATED_MACROS`. | ||
|
|
Comment on lines
+200
to
+202
| - **390+ source files** compile with zero errors | ||
| - **Flash**: 106 KB used (13% of 786 KB) | ||
| - **RAM**: 91 KB used (69% of 131 KB) |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Add tooling to package the ExecuTorch runtime as an Arduino library, enabling PyTorch model inference on Arduino microcontrollers. The library vendors ET runtime sources, CMSIS-NN kernels, and portable ops into a self-contained package that compiles under the Arduino build system.
Key components:
build_arduino_library.shassembles the distributable library from repository sources (no vendored copies checked in)ExecuTorchArduino.hconfigures the build environment for Arduino (fixes for std::variant, cmake_macros.h stub, build defines)platform_stubs.cprovides C library stubs for the LLEXT environmentValidated on Arduino Uno Q with DS-CNN keyword spotting model (int8, CMSIS-NN): 390+ source files compile, 106 KB flash (13%), 91 KB RAM.
Authored with assistance from Claude.