Dequantizing the Vitis ai quantized model to evaluate its performance on the test set in Pytorch.
- Requires Vitis AI, I have used vitis-ai-pytorch docker container
- Adjust the directory structure accordingly or change the dataset file path and model file path in the code.
example: evaluate model on subset of 1000 images from val
python eval_quant.py --subset_len 1000