/VAE-pytorch

computer vision project VAE with pytorch

Primary LanguagePython

VAE pytorch

computer vision project with pytorch (VAE)

python main.py \
--mode train \
--lambda_reg 1 \
--train_img_dir data/상의 \
--img_size 256 --latent_dim 64 \
--max_channel 256 --decoder_channel_in 256 \
--print_every 1 --save_every 10 --total_iters 200 \
--batch_size 64 --lr 1e-4 \
--checkpoint_dir expr/denormalization
/home/ubuntu/anaconda3/envs/pytorch_p36/bin/python main.py \
--mode train \
--lambda_reg 1 \
--train_img_dir data/상의 \
--img_size 256 --latent_dim 64 \
--max_channel 128 --decoder_channel_in 128 \
--print_every 1 --save_every 10 --total_iters 200 \
--batch_size 128 --lr 1e-4 \
--checkpoint_dir expr/channel_128

recon_weight 1000

/home/ubuntu/anaconda3/envs/pytorch_p36/bin/python main.py \
--mode train \
--train_img_dir data/상의 \
--img_size 256 --latent_dim 64 \
--max_channel 256 --decoder_channel_in 128 \
--print_every 1 --save_every 10 --total_iters 200 \
--batch_size 64 --lr 1e-4 \
--checkpoint_dir expr/recon_1000_channel_256 \
--recon_weight 1000 --resume_iter 90

recon_weight 10000

/home/ubuntu/anaconda3/envs/pytorch_p36/bin/python main.py \
--mode train \
--train_img_dir data/상의 \
--img_size 256 --latent_dim 64 \
--max_channel 256 --decoder_channel_in 128 \
--print_every 1 --save_every 10 --total_iters 200 \
--batch_size 64 --lr 1e-4 \
--checkpoint_dir expr/recon_10000_channel_256 \
--recon_weight 10000

test

/home/ubuntu/anaconda3/envs/pytorch_p36/bin/python main.py \
--mode train \
--train_img_dir test_data/ \
--img_size 256 --latent_dim 64 \
--max_channel 256 --decoder_channel_in 128 \
--print_every 1 --save_every 100 --total_iters 500 \
--batch_size 2 --lr 1e-4 \
--checkpoint_dir expr/test \
--recon_weight 10000

recon_weight 1000, kl_weight 10

/home/ubuntu/anaconda3/envs/pytorch_p36/bin/python main.py \
--mode train \
--train_img_dir /home/ubuntu/data/상의 \
--img_size 256 --latent_dim 64 \
--max_channel 256 --decoder_channel_in 128 \
--print_every 1 --save_every 10 --total_iters 200 \
--batch_size 64 --lr 1e-4 \
--checkpoint_dir expr/recon_1000_kl_10_channel_256 \
--recon_weight 1000 --kl_weight 10