Armand du Parc Locmaria 52e182611d initial usbgpu support (#37906)
* zero ll patched big model

* probe in a subprocess so usbgpu lock gets released

* compiles

* runs

* num_jobs gets overwritten, use side effect

* poll tg devices

* make sure build crashes on missing gpu

* fine not to rely on Device.default

* seperate tg env for each model runner

* comment

* Revert "seperate tg env for each model runner"

This reverts commit f6470cc4258eaeb3e8e37907ef370871c9af5aa4.

* env is shared, gate on flag

* no fallback warp dev must be set

* build for current device only, unless pc/release

* comment

* list

* listen for plug in

* add icon to status bar, read params on every frame (?)

* log available devices

* try copy out when loading?

* Revert "log available devices"

This reverts commit e8c52a5d59456d4820ecb13b99a6c46ea1386a20.

* Revert "try copy out when loading?"

This reverts commit 518f403aa03faeda1950fe3dbce0d9e4c1584455.

* don't trigger device probe/caching on modeld prepare

* re-export with ll and road edges

* dont cache devices in manager process

* get USBGPU from params

* no usbgpu env

* missed one

* sconscript don't poll

* unconditional env

* always explicitely set devices on input tensors

* set DEV so amd uses right compiler and iface??

* fix flag

* bump tg

* rm xdg_cache_home

* tg don't bump all the way

* missing gmmu=0 at compile time

* dm set dev

* tg backend

* update gitignore

* missing import

* unused imports

* rely on Device.DEFAULT at compile time (already the case bc onnxrunner)

* comments

* dm warp needs DEV set too

* build both smol and big

* misc typos

* set dev at compile time

* don't need

* DEV=CPU when getting metadata, ensure we don't grab gpu lock

* this would also grab lock

* put bool

* warp compile always prepare only

* missed one

* poll ui

* missing here

* don't force usbgpu at build time

* tmp patch fetch_fw

* catch all, follow hardwared patterns

* simpler

* compile make input queues

* revert this

* group this more readable

* rm empty line

* make dummy frame using numpy

* revert compile make input queues

* no compiler at runtime

* cleanup

* fine to rebuild all on change to device node for now

* fix usbgpu_present

* fix sconscript

* no size in header stream decompress

* DEBUG=2

* minimal viable feedback

* egpu gray

* oops

* gotta do this actually

* modeld build only depends on modeld devices

* don't ship onnx to release? or chunk

* don't need

* can only set compiler on dev=

* none device works, will use default

* make linter happy

* chunk agnostic onnx input to compile_modeld

* chunk big onnx

* +x chunker

* fix #!

* and don't ship chunked onnx to release

* firmware now in correct location

* better err on missing onnx/chunk

* SConscript also need to accept chunked onnx

* metadata also need to load maybe chunked

* dedupe cmd

* this needs to be on cpu

* devices are set in the tgflags, we already depend on them

* rebuilding on changed order is fine

* read file chunked can already load either chunked or not

* chunk all big onnx

* less confusing

* unused import

* python device to load onnx bytes

* default device for runners, python for metadata

* why not

* chunked to shm
2026-05-19 22:41:57 -07:00
2026-05-10 17:29:04 -07:00
2026-05-19 22:41:57 -07:00
2026-05-07 18:50:52 -07:00
2026-04-24 17:56:31 -07:00
2026-05-10 17:29:04 -07:00
2026-04-22 15:05:07 -07:00
2026-03-20 02:44:14 +00:00
2026-05-19 22:41:57 -07:00
2026-05-10 17:29:04 -07:00
2026-05-19 22:41:57 -07:00
2026-05-14 23:29:26 -07:00
2025-09-25 21:02:26 -07:00
2025-05-14 14:34:11 +01:00
2026-05-10 17:29:04 -07:00
2025-11-13 17:08:14 -08:00
2022-01-10 23:14:30 -08:00
2026-05-11 18:55:18 -07:00
2026-02-23 16:56:58 -08:00
2026-05-10 18:27:36 -07:00
2026-05-18 19:52:06 -07:00
2020-01-17 10:22:00 -08:00
2024-06-09 17:44:34 -07:00
2024-07-31 23:17:40 -07:00
2026-05-13 15:12:18 -07:00
2026-05-01 09:57:07 -07:00
2020-05-14 15:36:56 -07:00
2026-05-19 19:40:48 -07:00
2022-08-13 23:28:44 -07:00
2026-05-13 15:12:18 -07:00

openpilot

openpilot is an operating system for robotics.
Currently, it upgrades the driver assistance system in 300+ supported cars.

Docs · Roadmap · Contribute · Community · Try it on a comma four

Quick start: bash <(curl -fsSL openpilot.comma.ai)

openpilot tests License: MIT X Follow Discord

Using openpilot in a car

To use openpilot in a car, you need four things:

  1. Supported Device: a comma four, available at comma.ai/shop/comma-four.
  2. Software: The setup procedure for the comma four allows users to enter a URL for custom software. Use the URL openpilot.comma.ai to install the release version.
  3. Supported Car: Ensure that you have one of the 300+ supported cars.
  4. Car Harness: You will also need a car harness to connect your comma four to your car.

We have detailed instructions for how to install the harness and device in a car. Note that it's possible to run openpilot on other hardware, although it's not plug-and-play.

Branches

Running master and other branches directly is supported, but it's recommended to run one of the following prebuilt branches:

comma four branch comma 3X branch URL description
release-mici release-tizi openpilot.comma.ai This is openpilot's release branch.
release-mici-staging release-tizi-staging openpilot-test.comma.ai This is the staging branch for releases. Use it to get new releases slightly early.
nightly nightly openpilot-nightly.comma.ai This is the bleeding edge development branch. Do not expect this to be stable.
nightly-dev nightly-dev installer.comma.ai/commaai/nightly-dev Same as nightly, but includes experimental development features for some cars.

To start developing openpilot

openpilot is developed by comma and by users like you. We welcome both pull requests and issues on GitHub.

Want to get paid to work on openpilot? comma is hiring and offers lots of bounties for external contributors.

Safety and Testing

  • openpilot observes ISO26262 guidelines, see SAFETY.md for more details.
  • openpilot has software-in-the-loop tests that run on every commit.
  • The code enforcing the safety model lives in panda and is written in C, see code rigor for more details.
  • panda has software-in-the-loop safety tests.
  • Internally, we have a hardware-in-the-loop Jenkins test suite that builds and unit tests the various processes.
  • panda has additional hardware-in-the-loop tests.
  • We run the latest openpilot in a testing closet containing 10 comma devices continuously replaying routes.
MIT Licensed

openpilot is released under the MIT license. Some parts of the software are released under other licenses as specified.

Any user of this software shall indemnify and hold harmless Comma.ai, Inc. and its directors, officers, employees, agents, stockholders, affiliates, subcontractors and customers from and against all allegations, claims, actions, suits, demands, damages, liabilities, obligations, losses, settlements, judgments, costs and expenses (including without limitation attorneys’ fees and costs) which arise out of, relate to or result from any use of this software by user.

THIS IS ALPHA QUALITY SOFTWARE FOR RESEARCH PURPOSES ONLY. THIS IS NOT A PRODUCT. YOU ARE RESPONSIBLE FOR COMPLYING WITH LOCAL LAWS AND REGULATIONS. NO WARRANTY EXPRESSED OR IMPLIED.

User Data and comma Account

By default, openpilot uploads driving data to our servers. You can also access your data through comma connect. We use your data to train better models and improve openpilot for everyone.

openpilot is open source software, and users can disable data collection if they wish.

openpilot logs the road-facing cameras, CAN, GPS, IMU, magnetometer, thermal sensors, crashes, and operating system logs. The driver-facing camera and microphone are only logged if you explicitly opt-in in settings.

By using openpilot, you agree to our Privacy Policy. You understand that use of this software or its related services will generate certain types of user data, which may be logged and stored at the sole discretion of comma. By accepting this agreement, you grant an irrevocable, perpetual, worldwide right to comma for the use of this data.

S
Description
No description provided
Readme MIT 1.1 GiB
Languages
Python 68.6%
C++ 26.3%
Cap'n Proto 1.7%
C 1.4%
Shell 1%
Other 1%