train_neuralxc_with_args.sh

#!/bin/bash

dataset_name=$1
all_dataset_names=( "Bibtex" "Delicious" "Mediamill" "Eurlex" )
val=0
for name in ${all_dataset_names[*]};
    do
        if [ "$name" == "$dataset_name" ] ; then
            val=1
        fi
    done

if [ "$val" == 0 ] ; then
    echo "Invalid dataset name specified"
    exit 1
fi

setup_path="./setups/$dataset_name"
data_root="./data/$dataset_name"
dataset_info="dataset_info.yml"
inp_enc_cfg="input_encoder_cfg.yml"
inp_dec_cfg="input_decoder_cfg.yml"
otp_enc_cfg="output_encoder_cfg.yml"
otp_dec_cfg="output_decoder_cfg.yml"
reg_cfg="regressor_cfg.yml"
opt_cfg="opt.yml"


# Please specify the options below
# device -> PyTorch device string
# epochs -> Number of epochs
device=
epochs=

# This are static, feel free to change them as required
batch_size=64
inp_ae_wgt=0.01
otp_ae_wgt=0.01
seed=1729
all_k=( 1 3 5 )
interval=20
init_scheme="xavier_uniform"

for k in ${all_k[*]};
    do
        python train_neuralxc.py --data_root $data_root --dataset_info "$setup_path/$dataset_info" \
                                 --input_encoder_cfg "$setup_path/$inp_enc_cfg" --input_decoder_cfg "$setup_path/$inp_dec_cfg" \
                                 --output_encoder_cfg "$setup_path/$otp_enc_cfg" --output_decoder_cfg "$setup_path/$otp_dec_cfg" \
                                 --regressor_cfg "$setup_path/$reg_cfg" --device $device --epochs $epochs --batch_size $batch_size \
                                 --input_ae_loss_weight $inp_ae_wgt --output_ae_loss_weight $otp_ae_wgt \
                                 --optimizer_cfg "$setup_path/$opt_cfg" --seed $seed --plot --k $k --interval $interval \
                                 --init_scheme $init_scheme
    done