Couldn't check accuracy of quantized tensorflow model

[katherineyun]

Hi, I am currently working with a custom tensorflow model. So far, the quantization was successful, but I would like to know the accuracy of quantization results.

The following command is for testing quantized caffe network:
decent_q test -model quantize_results/quantize_train_test.prototxt -weights quantize_results/quantize_train_test.caffemodel -test_iter 100

But I have encountered this error (running decent_q test in tf container):

decent_q: error: argument command: invalid choice: 'test' (choose from 'quantize', 'inspect', 'dump')

Also, what command I should use to test the accuracy of a quantized tensorflow network?

Any suggestion or help is appreciated!

[Thesane]

you can use the validate script
https://github.com/Xilinx/ml-suite/tree/master/examples/tensorflow

python run.py --validate --model /opt/models/tensorflow/inception_v1_baseline.pb --output_dir work --input_nodes data --output_nodes loss3_loss3 --c_input_nodes data --c_output_nodes loss3_loss3 --input_shapes ?,224,224,3`

[katherineyun]

you can use the validate script
https://github.com/Xilinx/ml-suite/tree/master/examples/tensorflow

python run.py --validate --model /opt/models/tensorflow/inception_v1_baseline.pb --output_dir work --input_nodes data --output_nodes loss3_loss3 --c_input_nodes data --c_output_nodes loss3_loss3 --input_shapes ?,224,224,3`

Hi Ahmed, thank you for your answer! Do you know if there is a way to check the quantization accuracy without deploying the network onto FPGA?