Release Notes
Notes Before Use
This version is applicable to J6E/M/P/H/B, it is the official release of the algorithm toolchain, the version properties based on different chip platforms are as follows:
-
For J6E/M Linux platform: Commercial-ready version.
-
For J6E/M QNX platform: Restricted commercial version.
-
For J6P/H Linux platform: Commercial-ready version.
-
For J6B QNX platform: Commercial-ready version.
-
For S6P Linux platform: Integration version.
Please make sure to read the “J6 Algorithm Toolchain License Agreement” carefully before using the toolchain. If you choose to continue using the toolchain, it means that you have fully understood, comprehended and accepted the “J6 Algorithm Toolchain License Agreement”.
Release information
Horizon OpenExplorer 3.8.0(release_date:202602)
Main Functional Updates
-
Add support for the Starry Platform (S6P). The target platform can be specified by setting
march = nash-starry-p. -
Optimize the model export process in Horizon_plugin_pytorch, reducing export time and improving export efficiency.
-
Add the
quantization.hbdk4.checkAPI to Horizon_plugin_pytorch for early validation of potential issues during model export. -
Add the
load_calib_state_dictAPI to Horizon_plugin_pytorch. This API allows dynamically adjusting the quantization dtype of specified nodes after Histogram Calibration without recalibration. It automatically recalculates the corresponding scales based on the saved calibration statistics. For details, see Reusing Calibration Results when Adjusting Quantization dtype section. -
Add Android support to UCP DNN.
-
Add the
extract_functionAPI to HBDK, enabling extraction of specified subgraphs from a model.
Compatibility Statement
-
J6B DSP hardware does not support 101 symmetric border padding. It has been unified to zero padding. The affected operators are:
-
hbVPPyrUp.
-
hbVPPyrDown with HB_VP_INTER_GAUSSIAN interpolation mode.
-
hbVPHBDemosaicing.
-
-
To support the new calibration workflow and avoid FakeQuant misalignment after introducing int16, the default behavior of QconfigSetter has been changed: add + ReLU fusion is now enabled by default. For existing models, please use the load functionality for compatibility.
Use Restrictions
The inference preemption function for models containing the TopK operator depends on the BSP version, which needs to meet: J6EM-Linux >= v3.1.0, J6PH-Linux-Matrix >= v3.0.0, J6B-QNX >= v2.0.0.
Deprecated Functions
None.
Fix Issues
-
Fix the issue where hb_compile reports incorrect device information in certain scenarios.
-
Fix the issue where hb_model_info displays incorrect values for some parameters in certain scenarios.
-
Fix the issue where hb_analyzer reports inaccurate CPU statistics in HBM scenarios.
Known Issues
None.
