为CIFAR10运行nn_quantizer.py时出现错误blob_dims == blob-> shape()无法从hdf5加载blob

时间:2019-10-14 06:38:58

标签: machine-learning model deep-learning caffe cmsis

我正在执行ML-examples ->cmsisnn-cifar10。成功生成均值文件mean.binaryproto,ancifar10_test_lmdb and cifar10_train_lmdb。在将经过训练的Caffe模型转换为.pkl文件时,可以通过以下命令将其转换为cmsis-nn源文件。

python nn_quantizer.py --model models/cifar10_m7_train_test.prototxt --weights models/cifar10_m7_iter_300000.caffemodel.h5 --save models/cifar10_m7.pkl

遇到错误

    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)

帮帮我,我是Caffe的新手。

0 个答案:

没有答案