* Implement ROCm detectors
* Cleanup tensor input
* Fixup image creation
* Add support for yolonas in onnx
* Get build working with onnx
* Update docs and simplify config
* Remove unused imports
* Initial support for Hailo-8L
Added file for Hailo-8L detector including dockerfile, h8l.mk, h8l.hcl, hailo8l.py, ci.yml and ssd_mobilenat_v1.hef as the inference network.
Added files to help with the installation of Hailo-8L dependences like generate_wheel_conf.py, requirements-wheel-h8l.txt and modified setup.py to try and work with any hardware.
Updated docs to reflect Initial Hailo-8L support including oject_detectors.md, hardware.md and installation.md.
* Update .github/workflows/ci.yml
typo h8l not arm64
Co-authored-by: Nicolas Mowen <nickmowen213@gmail.com>
* Update docs/docs/configuration/object_detectors.md
Clarity for the end user and correct uses of words
Co-authored-by: Nicolas Mowen <nickmowen213@gmail.com>
* Update docs/docs/frigate/installation.md
typo
Co-authored-by: Nicolas Mowen <nickmowen213@gmail.com>
* update Installation.md to clarify Hailo-8L installation process.
* Update docs/docs/frigate/hardware.md
Co-authored-by: Josh Hawkins <32435876+hawkeye217@users.noreply.github.com>
* Update hardware.md add Inference time.
* Oops no new line at the end of the file.
* Update docs/docs/frigate/hardware.md typo
Co-authored-by: Josh Hawkins <32435876+hawkeye217@users.noreply.github.com>
* Update dockerfile to download the ssd_modilenet_v1 model instead of having it in the repo.
* Updated dockerfile so it dose not download the model file.
add function to download it at runtime.
update model path.
* fix formatting according to ruff and removed unnecessary functions.
---------
Co-authored-by: Nicolas Mowen <nickmowen213@gmail.com>
Co-authored-by: Josh Hawkins <32435876+hawkeye217@users.noreply.github.com>
Setting cache-to=compression=zstd causes the resulting user-pulled image
to have zstd-compressed layers, which are not compatible with docker
prior to 23.0. Ubuntu 20.04 still ships with docker 20.10, which yields
`Error processing tar file` when pulling these images.
Renaming the jetpack cache images is my way of clearing the cache of the
prior zstd layers, and it clarifies the convention I used for the other
cache images in which there is one cache per base image/job, not per
target/step. We don't need to delete the non-jetson cache images because
they haven't been rebuilt since zstd was enabled.
* fixup! Split independent builds into parallel jobs
* Combine caches within steps of same job
* Remove Maintain Cache workflow
Now that we're caching to ghcr instead of gha, we don't have to worry
about gha's cache eviction after 7 days/10 GB.
* Factor out common setup steps
* Re-order
* Split independent builds into parallel jobs
* Cache jetson builds
* Use zstd compression
* Switch from gha cache to registry cache
A CI run (four images cached with mode-max) populates the cache with 295
cache entries totalling 23.44 GB. This exceeds gha's 10GB limit, causing
trashing. Try with a registry instead.
* Enable manual CI runs
* Non-Jetson changes
Required for later commits:
- Allow base image to be overridden (and don't assume its WORKDIR)
- Ensure python3.9
- Map hwaccel decode presets as strings instead of lists
Not required:
- Fix existing documentation
- Simplify hwaccel scale logic
* Prepare for multi-arch tensorrt build
* Add tensorrt images for Jetson boards
* Add Jetson ffmpeg hwaccel
* Update docs
* Add CODEOWNERS
* CI
* Change default model from yolov7-tiny-416 to yolov7-320
In my experience the tiny models perform markedly worse without being
much faster
* fixup! Update docs
* Make main frigate build non rpi specific and build rpi using base image
* Add boards to sidebar
* Fix docker build
* Fix docs build
* Update pr branch for testing
* remove target from rpi build
* Remove manual build
* Add push build for rpi
* fix typos, improve wording
* Add arm build for rpi
* Cleanup and add default github ref name
* Cleanup docker build file system
* Setup to use docker bake
* Add ci/cd for bake
* Fix path
* Fix devcontainer
* Set targets
* Fix build
* Fix syntax
* Add wheels target
* Move dev container to trt
* Update key and fix rpi local
* Move requirements files and set intermediate targets
* Add back --load
* Update docs for community board development
* Update installation docs to reflect different builds available
* Update docs with official and community supported headers
* Update codeowners docs
* Update docs
* Assemble main and standard builds
* Change order of pushes
* Remove community board after successful build
* Fix rpi bake file names
* Initial WIP dockerfile and scripts to add tensorrt support
* Add tensorRT detector
* WIP attempt to install TensorRT 8.5
* Updates to detector for cuda python library
* TensorRT Cuda library rework WIP
Does not run
* Fixes from rebase to detector factory
* Fix parsing output memory pointer
* Handle TensorRT logs with the python logger
* Use non-async interface and convert input data to float32. Detection runs without error.
* Make TensorRT a separate build from the base Frigate image.
* Add script and documentation for generating TRT Models
* Add support for TensorRT devcontainer
* Add labelmap to trt model script and docs. Cleanup of old scripts.
* Update detect to normalize input tensor using model input type
* Add config for selecting GPU. Fix Async inference. Update documentation.
* Update some CUDA libraries to clean up version warning
* Add CI stage to build TensorRT tag
* Add note in docs for image tag and model support
* fix makefile variable
* add branch for testing
* fix arm32 build
* use amd64 for web build
* install wheels in a separate layer for better parallel builds
* try build-push-action
* try using gh context
* use short sha
* cleanup
* Make it easier to run the devcontainer
* Some more improvements
* Tidy up few other things
* Better name stages
* Fix CI
* Setup everything with one click
* Allow to set IMAGE_OWNER
* Change IMAGE_OWNER to IMAGE_REPO
* Fix CI with IMAGE_REPO
* Fix nodejs installation
* Test devcontainer build as part of CI
* Build devcontainer in its own job
* Fix devcontainer cli installation
* Fix devcontainer build
* Fix devcontainer build in CI again
* Enable buildkit only
* Increase coverage of devcontainer test
* Fix devcontainer start in CI
* Ensure latest version of docker compose is used
* Fix install compose action
* Disable CI stuff which does not work until we fix them
* try a different approach for build_web
* add automatic image builds
* build web first
* try disabling log file
* chown dir
* use volume
* set cache path
* test a push
* limit to dev/master branch commits
