In the ocrd-train Makefile, here is the code for finetuning
ifdef START_MODEL $(LAST_CHECKPOINT): unicharset lists $(PROTO_MODEL) mkdir -p data/checkpoints lstmtraining --traineddata $(PROTO_MODEL) --old_traineddata $(TESSDATA)/$(START_MODEL).traineddata --continue_from data/$(START_MODEL)/$(START_MODEL).lstm --net_spec "[1,36,0,1 Ct3,3,16 Mp3,3 Lfys48 Lfx96 Lrx96 Lfx256 O1chead -n1 data/unicharset]" --model_output data/checkpoints/$(MODEL_NAME) --learning_rate 20e-4 --train_listfile data/list.train --eval_listfile data/list.eval --max_iterations 10000 Why do we need the following line? I thought it was only used in training from scratch. --net_spec "[1,36,0,1 Ct3,3,16 Mp3,3 Lfys48 Lfx96 Lrx96 Lfx256 O1chead -n1 data/unicharset]" \ Should the learning rate be set lower for fine-tuning? The learning rate for training from scratch is 20e-4, so it would seem that the learning rate for fine-tuning should be significantly lower? --learning_rate 20e-4 \ -- 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/d1c9058e-3885-4ae7-9602-cf9033ccd87f%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
