Learnables Compression
Compression type, specified as "none"
or
"bfloat16"
Since R2023b
Model Configuration Pane: Code Generation / Interface
Description
Perform learnables compression and generate C/C++ code using Brain Floating Point format
(bfloat16
).
Dependencies
To enable this parameter, set the Target Library
parameter to none
.
Settings
"none"
(default) | "bfloat16"
To enable learnables compression, use "bfloat16"
.
Recommended Settings
Application | Setting |
---|---|
Debugging | |
Traceability | |
Efficiency | |
Safety precaution |
Programmatic Use
Parameter:
DLLearnablesCompression
|
Type: character vector |
Values:
"none" | "bfloat16"
|
Default:
"none"
|
Version History
Introduced in R2023b