libsvm iter bug

[eric-haibin-lin]

Memory corruption when reading the first 2k lines of kdda.t

[eric-haibin-lin]

minimum reproduce-able input:

1 1:1 2:1 3:1 4:1 5:1 6:1 7:1 8:0.301 9:0.602 10:1 11:1 12:1 13:1 14:1 15:1 16:1 17:1 18:1 19:1 20:1 21:1 22:1
1 2:1 3:1 4:1 5:1 8:0.301 10:1 11:1 12:1 14:1 15:1 17:1 19:1 20:1 23:1 24:1 25:1 26:1 27:1 28:0.602 29:1 30:1 31:1 32:1 33:1

    kdda = os.path.join(os.getcwd(), 'kdda.500')
    data_train = mx.io.LibSVMIter(data_libsvm=kdda,
                                  #label_libsvm=kdda,
                                  data_shape=(20216830, ),
                                  #label_shape=f[3],
                                  batch_size=1)
    it = iter(data_train)
    for batch in it:
        print(batch.data[0].asnumpy())
        #break

1e804c1