Ok this is confusing living heck out of me. I ran your code above and
iter$value() results in the dimensions I would expect. Hence, this appears
correct.
BUT, I tried again to simply run only my code below
```
train_iter <- mx.io.CSVIter(
data_csv = "A.csv", data_shape = c(3, 3, 1),
label_csv = "A.csv", label_shape = c(3, 3, 1),
batch_size = 1
)
train_iter$reset()
train_iter$iter.next()
train_iter$value()
```
and I still only get one value in the label. I am expecting $label to be
identical to $data
```
$label
[1] 1.373546
```
Now instead if I run:
```
train_iter <- mx.io.CSVIter(
data_csv = "A.csv", data_shape = c(3, 3, 1),
label_csv = "A.csv", label_shape = c(3, 1, 3),
batch_size = 1
)
train_iter$reset()
train_iter$iter.next()
train_iter$value()
```
I get
```
$`data`
, , 1, 1
[,1] [,2] [,3]
[1,] 1.1643714 2.689276 2.152658
[2,] 0.8696243 1.527834 1.233204
[3,] 2.1957829 2.738622 1.815747
$label
, , 1, 1
[,1]
[1,] 1.1643714
[2,] 0.8696243
[3,] 2.1957829
, , 2, 1
[,1]
[1,] 2.689276
[2,] 1.527834
[3,] 2.738622
, , 3, 1
[,1]
[1,] 2.152658
[2,] 1.233204
[3,] 1.815747
```
Is it possible that the label axis is rotated relative to the data axis? Are we
supposed to transpose the label axis until is matches the data axis?
[ Full content available at:
https://github.com/apache/incubator-mxnet/issues/12428 ]
This message was relayed via gitbox.apache.org for [email protected]
