Yes. As mentioned in the wiki regarding 4.0 training, it is very easy to overtrain if using large number of iterations for finetuning.
Please read the wiki page for more details. On 16-Oct-2017 11:17 PM, "Jerry Deng" <[email protected]> wrote: > Newbie here, any help is appreciated. I'm using some handwriting data to > Fine Tune train the english language model that I extracted from > eng.traineddata file. Prepared with box file and lstmf file. It works > with small test. When I actually run it on 1500+ lstmf file, it works fine > when I keep the max_iteration to be under 2200 or so. As soon as I went > over some threshold amount, the model suddenly became unusable and and spit > out only CAPITAL letters with some odd punctuations (and the error rate > shoot over 100%). One time it even failed due to a segmentation fault. > Does it sound like it's running out of memory or what are the possible > causes? > > -- > You received this message because you are subscribed to the Google Groups > "tesseract-ocr" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To post to this group, send email to [email protected]. > Visit this group at https://groups.google.com/group/tesseract-ocr. > To view this discussion on the web visit https://groups.google.com/d/ > msgid/tesseract-ocr/84abde41-209b-493b-8c07-d6d9ea9fb33a% > 40googlegroups.com > <https://groups.google.com/d/msgid/tesseract-ocr/84abde41-209b-493b-8c07-d6d9ea9fb33a%40googlegroups.com?utm_medium=email&utm_source=footer> > . > For more options, visit https://groups.google.com/d/optout. > -- You received this message because you are subscribed to the Google Groups "tesseract-ocr" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at https://groups.google.com/group/tesseract-ocr. To view this discussion on the web visit https://groups.google.com/d/msgid/tesseract-ocr/CAG2NduWDh6qEH64OA%2B-nc-b1WG2raDpo9G%2BcDmdxRFfDP5jB1Q%40mail.gmail.com. For more options, visit https://groups.google.com/d/optout.
