big-endian machine

[earonesty]

i want to be able to train a model on a linux box, but then ship to an ibm or sun box or any other for execution.

cannot load on a big-endian architecture when it was generated on a little-endian machine

Restoring existing nnet
Bad model version: -1834810029

but other than that i can get the program to work fine on ibm/sun

[akhti]

Well, seems that that you need some more portable format to do this kind of stuff, e.g. text. That's not so hard to force faster-rnnlm to use text format.

• replace FreadAllOrDie with istream<< in ReadHeader in nnet.cc:19
• replace fwrite with ostream>> in NNet::Save and NNet::SaveCompatible in nnet.cc:194
• update Dump and Load functions in util.h:37
• if you use maxent, you should also update MaxEnt::Dump and MaxEnt::Load in maxent.cc:43

Ok, maybe it's no absolutely trivial, but totally doable ;)