nn_quantizer.py for CIFAR10 got error blob_dims == blob->shape() Cannot load blob from hdf5

I have download code from git clone https://github.com/ARM-software/ML-examples.git

I have executed ML-examples ->cmsisnn-cifar10 and successfully generate mean file mean.binaryproto,cifar10_test_lmdb and cifar10_train_lmdb.

through below command


While converting trained Caffe model into .pkl file, so that it can be converted to cmsis-nn source file, by below command.
python nn_quantizer.py --model models/cifar10_m7_train_test.prototxt --weights models/cifar10_m7_iter_300000.caffemodel.h5 --save models/cifar10_m7.pkl

got below error

**Check failed: blob_dims == blob->shape() Cannot load blob from hdf5;
shape mismatch. Source shape is 16 32 5 5 (12800) target shape is 32 32 5 5 (25600)
*** Check failure stack trace: *****

I don’t know why the above error occurs as I am using a trained Caffe model without any modification for reference I have attached prototxt file. I am new to Caffe, please help me.
cifar10_m7_train_test_prototxt.txt (2.75 KB)

Did you follow the instructions here:


Also, I’m porting TensorFlow again. We will have it done by the end of the week. This gets us off caffe finally.

I did not train the model, just try to convert it into .pkl file.

run below script without GPU, is it ok

python nn_quantizer.py