pic2text

BUPT-NNDL-COURSE-DESIGN

linux-instruction

using

conda env create -f environment.yaml
conda activate pic2text 

finish conda env create

using

python set_vocab.py

set vocabulary

using

python main.py fit --config /your/configs/config.yaml

train the model

autosave

autoresume

using

tensorboard --logdir=your/root/path

to see training log

windows-instruction

using

conda env create -f environment.yaml
conda activate pic2text 

finish conda env create

reinstall conda from pytorch offical website, dpending on your CUDA version

you may need to install or update some extra package if you see error

change your file path in config! they are different from linux

using

set PYTHONUTF8=1
python main-forwindows.py fit --config /your/configs/config.yaml

else same to Linux