Training huge hdf5 file in batches trigger DeferredInitializationError #19320
-
Hi All, I am a newbie to mxnet, however, I did follow couple of tutorials online and thought of training a model on a huge hdf5 file ndarrays not images. But, I got hanged up on "DeferredInitilizatioError". However, from the tutorials and the google searches I looked at tells me this representation is correct... Is there something I am missing in the code? Thank you in advance...
|
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 1 reply
-
If you don't specify |
Beta Was this translation helpful? Give feedback.
If you don't specify
in_units=
ingluon.nn.Dense(1)
, the number ofin_units
will be inferred from the first forward pass. If you don't want to infer them, just specify thein_units=
argument when adding the Dense layer.