From 8f83e080c32ee4b99d3db60ad2e3e3408387c43a Mon Sep 17 00:00:00 2001 From: "Jim Wang @ Intel" Date: Tue, 16 Jan 2024 18:51:22 -0700 Subject: [PATCH] docs: fix the formatting json issue of troubleshooting page (#463) Signed-off-by: Jim Wang --- docs_src/troubleshooting.md | 19 +++++++++---------- 1 file changed, 9 insertions(+), 10 deletions(-) diff --git a/docs_src/troubleshooting.md b/docs_src/troubleshooting.md index ac3660bb..18de9263 100644 --- a/docs_src/troubleshooting.md +++ b/docs_src/troubleshooting.md @@ -4,13 +4,12 @@ Q: Why is the performance sometimes on CPU better than on GPU, when running pipe A: The performance of pipeline benchmarking strongly depends on the models. Specifically for `yolov5s` object detection, it is recommended to use the model precision FP32 when it is running on device `GPU`. If supported, then you can change the model precision by going to the folder `configs/opencv-ovms/models/2022` from the root of project folder and editing the `base_path` for that particular model in the `config_template.json` file. For example, you can change the the base_path of `FP16` to `FP32` assuming the precision `FP32` of the model yolov5s is available: - ```json - ... - "config": { - "name": "yolov5s", - "base_path": "/models/yolov5s/FP32-INT8", - "layout": "NHWC:NCHW", - ... - } - - ``` +```json + ... + "config": { + "name": "yolov5s", + "base_path": "/models/yolov5s/FP32-INT8", + "layout": "NHWC:NCHW", + ... + } +```