What means about output encoding?
some quentions about quantization
Posted: Tue, 2021-11-09 23:17
Forums - some quentions about quantization
Opinions expressed in the content posted here are the personal opinions of the original authors, and do not necessarily reflect those of Qualcomm Incorporated or its subsidiaries (“Qualcomm”). The content is provided for informational purposes only and is not meant to be an endorsement or representation by Qualcomm or any other party. This site may also provide links or references to non-Qualcomm sites and resources. Qualcomm makes no representations, warranties, or other commitments whatsoever about any non-Qualcomm sites or third-party resources that may be referenced, accessible from, or linked to this site.
Hello,
Quantisation is one of the way to reduce the AI computation demands and increase power efficiency. It is an umbrella term that covers a lot of different techniques to convert input values from a large set to output values in a smaller set. Coming to SNPE, non-quantized DLC files use 32 bit floating point representations of network parameters whereas Quantized DLC files use fixed point representations of network parameters, generally 8 bit weights and 8 or 32bit biases.
By default snpe-xxx-to-dlc tool convert original model into non quantised model. To quantize the model to 8 bit fixed point, use snpe-xxx-quantize (xxx is caffe, caffe2, onnx, tf).
The Quantisation algorithm takes the set of floating point values as input and outputs corresponding set of 8-bit fixed point values and encoding parameters (encoding-min and encoding-max). These output encoding parameters define the range and floating point values that will be representable by the fixed point format.
encoding-min: specifies the smallest floating point value that will be represented by the fixed point value of 0
encoding-max: specifies the largest floating point value that will be represented by the fixed point value of 255
For more information on Quantisation refer the following links:
https://www.qualcomm.com/news/onq/2019/03/12/heres-why-quantization-matt...
https://developer.qualcomm.com/sites/default/files/docs/snpe/quantized_m...