#Quantization Analysis Tool

The quant_analysis.py provided in the algorithm package is an analysis tool for model deployment scenarios. It is used to verify the quantization sensitivity before deploying the model from floating-point (FP32) to quantization (such as INT8). The execution tool supports single-machine multi-card and multi-machine distributed operation.

#Usage

Run the tools/quant_analysis.py script to load the specified configuration file, perform model quantization analysis, and output the analysis results to the designated directory.

#Parameters Introduction

#Usgae Example

python3 tools/quant_analysis.py \
        -c ./configs/bev/bevformer_tiny_resnet50_detection_nuscenes.py \
        -ids 0,1

For a detailed introduction to the configuration file, please refer to the [Configuration] section [config configuration] (../config/keywords).