Search code examples
deep-learningjupyter-notebookmxnet

jupyter notebook kernel dies when use mxnet ImageRecordIter with cifar10_val.rec

The cell is like below, trying to use cifar10_val.rec.The file is there.

def get_data_from_cifar(): 
    train = mx.io.ImageRecordIter() 
    val = mx.io.ImageRecordIter() 
    return train, val 

train, val = get_data_from_cifar()

Everytime I run the notebook it dies,telling

The kernel appears to have died. It will restart automatically.

There is a post about jupyter notebook kernel dies when using pandas,it's about memory,however it has to_sparse() to deal with that.

The memory in my pc is not huge also.However,while running the cell,the taks manager doesn't show the memory like to be exhausted.

Could this be some other problem?

UPDATE:

run scripts in python interpreter shell it tells as @leezu said

304: [16:36:58] src/io/image_aug_default.cc:282: Check failed: static_cast(res.rows) >= param_.data_shape1 && static_cast(res.cols) >= param_.data_shape[2] input image size smaller than input shape

guess I used wrong data_shape(3,128,128).Would update after change them acording to cifar10 instance.

Solution

  • Find out what cifar10 have is 60000 32*32 images.After change data_shape to (3,32,32) it doesn't die anymore.

    Actually have to set the label_width to 1 also. However the error can be track from the terminal.