HongChuang/ultralytics-ascend
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 3

ROS quickstart Docs page cleanup (#13835)

Browse source 
Co-authored-by: UltralyticsAssistant <web@ultralytics.com>
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>
This commit is contained in:
Francesco Mattioli 2024-06-20 18:41:19 +02:00 • committed by GitHub
parent 0ca3cd3fbe
commit 31de5d044f
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
7 changed files with 282 additions and 161 deletions
Download patch file Download diff file Expand all files Collapse all files

4
docs/en/integrations/tensorrt.md
View file

@ -137,9 +137,9 @@ The arguments provided when using [export](../modes/export.md) for an Ultralytic
- `batch` : The maximum batch-size that will be used for inference. During inference smaller batches can be used, but inference will not accept batches any larger than what is specified.
!!! note
!!! note
During calibration, twice the `batch` size provided will be used. Using small batches can lead to inaccurate scaling during calibration. This is because the process adjusts based on the data it sees. Small batches might not capture the full range of values, leading to issues with the final calibration, so the `batch` size is doubled automatically. If no batch size is specified `batch=1`, calibration will be run at `batch=1 * 2` to reduce calibration scaling errors.
During calibration, twice the `batch` size provided will be used. Using small batches can lead to inaccurate scaling during calibration. This is because the process adjusts based on the data it sees. Small batches might not capture the full range of values, leading to issues with the final calibration, so the `batch` size is doubled automatically. If no batch size is specified `batch=1`, calibration will be run at `batch=1 * 2` to reduce calibration scaling errors.
Experimentation by NVIDIA led them to recommend using at least 500 calibration images that are representative of the data for your model, with INT8 quantization calibration. This is a guideline and not a _hard_ requirement, and <u>**you will need to experiment with what is required to perform well for your dataset**.</u> Since the calibration data is required for INT8 calibration with TensorRT, make certain to use the `data` argument when `int8=True` for TensorRT and use `data="my_dataset.yaml"`, which will use the images from [validation](../modes/val.md) to calibrate with. When no value is passed for `data` with export to TensorRT with INT8 quantization, the default will be to use one of the ["small" example datasets based on the model task](../datasets/index.md) instead of throwing an error.