thanks will keep trying.
G

On Tuesday, December 5, 2023 at 10:41:56 PM UTC+2 Keith Smith wrote:

> From one novice to another ...
>
> 1. Yes, that is my understanding of how to run further iterations.
>
> 2. Yes, EPOCHS says to iterate that many times over your set of tests.  I 
> think I have heard the recommended number of EPOCHS in general is 2, though 
> I don't know how much science is behind that.  I think 100 EPOCHs is too 
> many and will over fit.
>
> On Tue, Dec 5, 2023 at 3:47 AM Ghinwa Choueiter <gchou...@gmail.com> 
> wrote:
>
>> Hi there,
>>
>> I trained a model as follows
>>
>> export TESSDATA_DIR=/<path>/tessdata_best/
>> export LANGDATA_DIR=/<path>/tesstrain/data
>>
>> nohup make LANG_TYPE=RTL \
>>       MODEL_NAME=ara_plus \
>>       PSM=13 \
>>       START_MODEL=ara \
>>       TESSDATA=$TESSDATA_DIR \
>>       LANGDATA_DIR=$LANGDATA_DIR \
>>       EPOCHS=100 \
>>       RATIO_TRAIN=0.90 \
>>       DEBUG_INTERVAL=-1 training >> data/ara_plus.log &
>>
>> 1. once I have the initial model, how would I run further iterations on 
>> the same data. Should I copy ara_plus.traineddata to  $TESSDATA_DIR and 
>> specify START_MODEL=ara_plus? Or is there another way.
>>
>> 2. When I specify EPOCHS > 0 then I see that the Makefile sets the 
>> iterations to - EPOCHS. What is that actually doing? Will it actually 
>> iterate = EPOCHS * data points. I see we are using SGD so LSTM training is 
>> running each data point separately. 
>>
>> thank you.
>> G
>>
>> -- 
>> 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 tesseract-oc...@googlegroups.com.
>> To view this discussion on the web visit 
>> https://groups.google.com/d/msgid/tesseract-ocr/b6e1ab30-75f8-4a93-a80d-a95cb72e5b22n%40googlegroups.com
>>  
>> <https://groups.google.com/d/msgid/tesseract-ocr/b6e1ab30-75f8-4a93-a80d-a95cb72e5b22n%40googlegroups.com?utm_medium=email&utm_source=footer>
>> .
>>
>

-- 
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 tesseract-ocr+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/tesseract-ocr/d6308a05-74d4-4135-926f-a26dfb8a2454n%40googlegroups.com.

Reply via email to