Merge branch 'blakeblackshear:dev' into dev

.github/pull_request_template.md

@@ -13,6 +13,7 @@
 - [ ] New feature
 - [ ] Breaking change (fix/feature causing existing functionality to break)
 - [ ] Code quality improvements to existing code
+- [ ] Documentation Update
 
 ## Additional information
 

docker/tensorrt/Dockerfile.arm64

@@ -10,8 +10,8 @@ ARG DEBIAN_FRONTEND
 # Use a separate container to build wheels to prevent build dependencies in final image
 RUN apt-get -qq update \
     && apt-get -qq install -y --no-install-recommends \
-       python3.9 python3.9-dev \
-       wget build-essential cmake git \
+    python3.9 python3.9-dev \
+    wget build-essential cmake git \
     && rm -rf /var/lib/apt/lists/*
 
 # Ensure python3 defaults to python3.9
@@ -41,7 +41,8 @@ RUN --mount=type=bind,source=docker/tensorrt/detector/build_python_tensorrt.sh,t
     && TENSORRT_VER=$(cat /etc/TENSORRT_VER) /deps/build_python_tensorrt.sh
 
 COPY docker/tensorrt/requirements-arm64.txt /requirements-tensorrt.txt
-RUN pip3 wheel --wheel-dir=/trt-wheels -r /requirements-tensorrt.txt
+RUN pip3 uninstall -y onnxruntime \
+    && pip3 wheel --wheel-dir=/trt-wheels -r /requirements-tensorrt.txt
 
 FROM build-wheels AS trt-model-wheels
 ARG DEBIAN_FRONTEND

docker/tensorrt/requirements-arm64.txt

@@ -1 +1,2 @@
 cuda-python == 11.7; platform_machine == 'aarch64'
+onnxruntime @ https://nvidia.box.com/shared/static/9aemm4grzbbkfaesg5l7fplgjtmswhj8.whl; platform_machine == 'aarch64'

docs/docs/configuration/object_detectors.md

@@ -22,14 +22,14 @@ Frigate supports multiple different detectors that work on different types of ha
 - [ONNX](#onnx): OpenVINO will automatically be detected and used as a detector in the default Frigate image when a supported ONNX model is configured.
 
 **Nvidia**
-- [TensortRT](#nvidia-tensorrt-detector): TensorRT can run on Nvidia GPUs, using one of many default models.
-- [ONNX](#onnx): TensorRT will automatically be detected and used as a detector in the `-tensorrt` Frigate image when a supported ONNX model is configured.
+- [TensortRT](#nvidia-tensorrt-detector): TensorRT can run on Nvidia GPUs and Jetson devices, using one of many default models.
+- [ONNX](#onnx): TensorRT will automatically be detected and used as a detector in the `-tensorrt` or `-tensorrt-jp(4/5)` Frigate images when a supported ONNX model is configured.
 
 **Rockchip**
 - [RKNN](#rockchip-platform): RKNN models can run on Rockchip devices with included NPUs.
 
 **For Testing**
-- [CPU Detector (not recommended for actual use](#cpu-detector-not-recommended): Use a CPU to run tflite model, this is not recommended and in most cases OpenVINO can be used in CPU mode with better results. 
+- [CPU Detector (not recommended for actual use](#cpu-detector-not-recommended): Use a CPU to run tflite model, this is not recommended and in most cases OpenVINO can be used in CPU mode with better results.
 
 :::
 

docs/docs/configuration/semantic_search.md

@@ -43,9 +43,32 @@ The text model is used to embed tracked object descriptions and perform searches
 
 Differently weighted CLIP models are available and can be selected by setting the `model_size` config option:
 
-:::tip
+```yaml
+semantic_search:
+  enabled: True
+  model_size: small
+```
+
+- Configuring the `large` model employs the full Jina model and will automatically run on the GPU if applicable.
+- Configuring the `small` model employs a quantized version of the model that uses much less RAM and runs faster on CPU with a very negligible difference in embedding quality.
 
-The CLIP models are downloaded in ONNX format, which means they will be accelerated using GPU hardware when available. This depends on the Docker build that is used. See [the object detector docs](../configuration/object_detectors.md) for more information.
+### GPU Acceleration
+
+The CLIP models are downloaded in ONNX format, and the `large` model can be accelerated using GPU hardware, when available. This depends on the Docker build that is used.
+
+:::info
+
+If the correct build is used for your GPU and the `large` model is configured, then the GPU will be detected and used automatically.
+
+**AMD**
+- ROCm will automatically be detected and used for semantic search in the `-rocm` Frigate image.
+
+**Intel**
+- OpenVINO will automatically be detected and used as a detector in the default Frigate image.
+
+**Nvidia**
+- Nvidia GPUs will automatically be detected and used as a detector in the `-tensorrt` Frigate image.
+- Jetson devices will automatically be detected and used as a detector in the `-tensorrt-jp(4/5)` Frigate image.
 
 :::
 
@@ -55,9 +78,6 @@ semantic_search:
   model_size: small
 ```
 
-- Configuring the `large` model employs the full Jina model and will automatically run on the GPU if applicable.
-- Configuring the `small` model employs a quantized version of the model that uses much less RAM and runs faster on CPU with a very negligible difference in embedding quality.
-
 ## Usage and Best Practices
 
 1. Semantic search is used in conjunction with the other filters available on the Search page. Use a combination of traditional filtering and semantic search for the best results.

frigate/config/config.py

@@ -667,7 +667,7 @@ def load(cls, **kwargs):
             migrate_frigate_config(config_path)
 
         # Finally, load the resulting configuration file.
-        with open(config_path, "a+") as f:
+        with open(config_path, "a+" if new_config else "r") as f:
             # Only write the default config if the opened file is non-empty. This can happen as
             # a race condition. It's extremely unlikely, but eh. Might as well check it.
             if new_config and f.tell() == 0: