Internal type error when using trained model

Support
1

Hi,

I’ve tried getting started and small experiments using opennmt with success.

But, on my real problem, I’ve some issues. For information, my data set has 5.000.000 lines which produce a 160mb model to be trained.

After 7 days training, when I try to use the trained model I get this error with or without GPU:

root@9f0faa133529:~/torch/opennmt# th translate.lua -model ../data/os-model-14043_epoch13_15.07_10478.t7  -src data/src-test.txt  -output pred.txt -gpuid 1
[07/23/17 14:04:15 INFO] Using GPU(s): 1	
[07/23/17 14:04:15 INFO] Loading '../data/os-model-14043_epoch13_15.07_10478.t7'...	
[07/23/17 14:04:17 INFO] Model seq2seq trained on bitext	
/root/torch/install/bin/luajit: /root/torch/install/share/lua/5.1/nn/MapTable.lua:15: bad argument #1 to 'unpack' (table expected, got boolean)
stack traceback:
	[C]: in function 'unpack'
	/root/torch/install/share/lua/5.1/nn/MapTable.lua:15: in function '_extend'
	/root/torch/install/share/lua/5.1/nn/MapTable.lua:36: in function 'updateOutput'
	./onmt/modules/Network.lua:11: in function 'forward'
	./onmt/translate/Translator.lua:271: in function 'translateBatch'
	./onmt/translate/Translator.lua:396: in function 'translate'
	translate.lua:110: in function 'main'
	translate.lua:196: in main chunk
	[C]: in function 'dofile'
	/root/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:145: in main chunk
	[C]: at 0x00406670

I dont know how to fix that…

Please, can you help me?

Thanks!

2

Hi,

You should update your Torch installation.

3

@guillaumekln thank you for the answer!

But, after the update, I got this output:

root@7afe740a3e28:~/torch/opennmt# th translate.lua -model ../data/os-model-29221_epoch13_15.07_release.t7 -src data/src-test.txt -output pred.txt
[07/26/17 16:18:46 INFO] Loading '../data/os-model-29221_epoch13_15.07_release.t7'...	
/root/torch/install/bin/luajit: /root/torch/install/share/lua/5.1/torch/File.lua:375: unknown object
stack traceback:
	[C]: in function 'error'
	/root/torch/install/share/lua/5.1/torch/File.lua:375: in function 'readObject'
	/root/torch/install/share/lua/5.1/torch/File.lua:368: in function 'readObject'
	/root/torch/install/share/lua/5.1/nn/Module.lua:192: in function 'read'
	/root/torch/install/share/lua/5.1/torch/File.lua:351: in function 'readObject'
	/root/torch/install/share/lua/5.1/torch/File.lua:369: in function 'readObject'
	/root/torch/install/share/lua/5.1/torch/File.lua:369: in function 'readObject'
	/root/torch/install/share/lua/5.1/torch/File.lua:353: in function 'readObject'
	/root/torch/install/share/lua/5.1/torch/File.lua:368: in function 'readObject'
	/root/torch/install/share/lua/5.1/torch/File.lua:369: in function 'readObject'
	/root/torch/install/share/lua/5.1/torch/File.lua:369: in function 'readObject'
	...
	/root/torch/install/share/lua/5.1/torch/File.lua:369: in function 'readObject'
	/root/torch/install/share/lua/5.1/torch/File.lua:369: in function 'readObject'
	/root/torch/install/share/lua/5.1/torch/File.lua:409: in function 'load'
	./onmt/translate/Translator.lua:110: in function '__init'
	/root/torch/install/share/lua/5.1/torch/init.lua:91: in function 'new'
	translate.lua:53: in function 'main'
	translate.lua:196: in main chunk
	[C]: in function 'dofile'
	/root/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:150: in main chunk
	[C]: at 0x00406670

I’m using an x86_64 linux on my training machine and production machine (RedHat Linux 6.4 and Ubuntu distros). After reading previous topics I tried to run a test on my training machine, but I have the same issue.

I have no idea… can you help me to solve that?

Thanks

4

Hi,

I apologize for the insistence… but can you help me to fix that?

Thanks for support,

5

Hi @welingtonveiga,

It’s hard to debug actually as I never encountered this issue and the error is not very explicit. Did you try to train on the Ubuntu machine instead?

6

Hi @guillaumekln!
When I tried it on my Ubuntu machine everything is OK.
Now, I’m running a simple example on my main training server to give you more information about the problem.
Thanks,