lucidrains / lightweight-gan

Implementation of 'lightweight' GAN, proposed in ICLR 2021, in Pytorch. High resolution image generations that can be trained within a day or two
MIT License
1.62k stars 220 forks source link

showing results while training ? #119

Open galaelized opened 2 years ago

galaelized commented 2 years ago

how to show generator results after every epoch during training ?

this is my current configuration

 lightweight_gan \
  --data "/content/dataset/Dataset/" \
  --num-train-steps 100000 \
  --image-size 128 \
  --name GAN2DBlood5k \
  --batch-size 32 \
  --gradient-accumulate-every 5 \
  --disc-output-size 1 \
  --dual-contrast-loss \
  --attn-res-layers [] \
  --calculate_fid_every 1000\
  --greyscale \
  --amp

using --show-progress only works after training. Also it seems that there is no longer checkpoints per epoch

iScriptLex commented 2 years ago

--evaluate-every 1 - show generation every 1 epoch. --num-image-tiles 14 - make the set of 14x14 images when showing generation result. --save-every 500 - save checkpoint every 500 epochs.

galaelized commented 2 years ago

how to plot generator loss and discriminator loss?