The problem can be solved by changing e1.forward() to e1.forward(is_train=True) The difference here is that if you don't specify is_train=True, the executor does not expect to have a backward after and thus won't initialize e1.outputs which leads to seg fault in future calculations.
[ Full content available at: https://github.com/apache/incubator-mxnet/issues/12269 ] This message was relayed via gitbox.apache.org for [email protected]
