train.sh

output="runs"
device="cuda0"

## Originally the batch size was 10 and epoch size was 200 for hs
## Originally the batch size was 10, epoch size was 50, valid_per_batch and save_per_batch were 4000 each for django
if [ "$2" == "gColab" ]; then
	echo "Google Colab running"
  	batch_size=10
  	max_epoch_size=10
else
  	batch_size=10
  	max_epoch_size=200
fi


if [ "$1" == "hs" ]; then
	# hs dataset
	echo "training hs dataset"
	echo "batch size : ${batch_size}"
	echo "max epoch size : ${max_epoch_size}"
	dataset="hs.freq3.pre_suf.unary_closure.bin"
	commandline="-batch_size ${batch_size} -max_epoch ${max_epoch_size} -valid_per_batch 280 -save_per_batch 280 -decode_max_time_step 350 -optimizer adadelta -rule_embed_dim 128 -node_embed_dim 64 -valid_metric bleu"
	datatype="hs"
else
	# django dataset
	batch_size=10
	max_epoch_size=2
	valid_per_batch_val=500
	save_per_batch_val=500
	echo "training django dataset"
	echo "batch size : ${batch_size}"
	echo "max epoch size : ${max_epoch_size}"
	echo "valid_per_batch_val : ${valid_per_batch_val}"
	echo "save_per_batch_val : ${save_per_batch_val}"
	dataset="django.cleaned.dataset.freq5.par_info.refact.space_only.bin"
	commandline="-batch_size ${batch_size} -max_epoch ${max_epoch_size} -valid_per_batch ${valid_per_batch_val} -save_per_batch ${save_per_batch_val} -decode_max_time_step 100 -optimizer adam -rule_embed_dim 128 -node_embed_dim 64 -valid_metric bleu"
	datatype="django"
fi

# train the model
THEANO_FLAGS="mode=FAST_RUN,device=${device},floatX=float32" python -u code_gen.py \
	-data_type ${datatype} \
	-data data/${dataset} \
	-output_dir ${output} \
	${commandline} \
	train

# decode testing set, and evaluate the model which achieves the best bleu and accuracy, resp.
for model in "model.best_bleu.npz" "model.best_acc.npz"; do
	THEANO_FLAGS="mode=FAST_RUN,device=${device},floatX=float32" python code_gen.py \
	-data_type ${datatype} \
	-data data/${dataset} \
	-output_dir ${output} \
	-model ${output}/${model} \
	${commandline} \
	decode \
	-saveto ${output}/${model}.decode_results.test.bin

	python code_gen.py \
		-data_type ${datatype} \
		-data data/${dataset} \
		-output_dir ${output} \
		evaluate \
		-input ${output}/${model}.decode_results.test.bin
done