About training options of fashion dataset

[Jarvisss]

There is an example training command in README, but it's for market dataset, do you use the same parameters for the fashion dataset ?
I use the same parameters as market for fashion to train for 60k iterations, but the results are not so good as in paper.
Could you please share the training options of fashion to obtain the result in your paper, that would be really helpful for me, thx!

[Jarvisss]

here is my training command on fashion
CUDA_VISIBLE_DEVICES=2 python train.py --output_dir output/fashion/model_2 --checkpoints_dir output/fashion/model_2 --warp_skip mask --dataset fashion --nb_inputs 2 --l1_penalty_weight 0.001 --nn_loss_area_size 5 --batch_size 1 --content_loss_layer block1_conv2 --number_of_epochs 60 --dmax 6 --kernel_size_last=1 --fusion_type att_dec_rec --return_att 1 --gan_penalty_weight 0.1

I modify the --nn_loss_area_size to be 5 as Deformable GANs suggest, but I'm not sure if the --dmax and --kernel_size should be modified for fashion dataset, could you plz help me?
@Stephlat