Segmentation fault (core dumped) running lstmtraining

[John Lee Ward]

I am trying to follow along the tessearct 4 training tutorial on github. All is well until I try to run the lstmtraining.

I run this command on a tesseract 4.00 beta installation

lstmtraining --debug_interval 100   --traineddata ~/tesstutorial/engtrain/eng/eng.traineddata   --net_spec '[1,36,0,1 Ct3,3,16 Mp3,3 Lfys48 Lfx96 Lrx96 Lfx256 O1c111]'   --model_output ~/tesstutorial/engoutput/base --learning_rate 20e-4   --train_listfile ~/tesstutorial/engtrain/eng.training_files.txt   --eval_listfile ~/tesstutorial/engeval/eng.training_files.txt   --max_iterations 5000 &>~/tesstutorial/engoutput/basetrain.log

I immediately get this error: Segmentation fault (core dumped)

Looking at the command, the only thing that concerns me is the crytic --net_spec option. Thinking that maybe these arguments are causing a problem, I thought to see if my traineddata has layers of the same name. So I run this command:

combine_tessdata -d tessdata/eng.traineddata

Version string:4.0.0-beta.1

23:version:size=12, offset=192

And it does not show any of the layer information. So now I am more confused than ever. 

Tried restarting machine, same error. Nothing in basetrain.log. For what its worth, I am running Ubuntu 18.04 on a Dell xps. 

[John Lee Ward]

I noticed that the --traindeddata path was erroneous. I corrected that to point to a good eng.traineddata file but the error is the same. 

Segmentation Faul (core dumped) 

[John Lee Ward]

Problem resolved. Once I finally pointed to a valid eng.traineddata file, all was good.  

Sorry for the false alarm. Case closed.