You still need to run Facial Destylization, since we need to prepare paired training data to train DualStyleGAN (i.e., need to obtain the intrinsic and extrinsic style codes).

Then, in Progressive Fine-Tuning, you can skip step 1 & 2 by loading my pretrained model, and only run step 3.

from dualstylegan.

Thanks for answering it!
So does it mean I can skip step2 in facial destylization but need to do step 3? Or should I do both?

from dualstylegan.

You can try with or without step 2.
I don't know which one is better.

from dualstylegan.

Thanks!
I tried to train step2 in Facial Destylization, but faced an error.

I prepared 300 images with alignment and super resolution script, 1024x1024, and saved them in /data/mystyle/images/train
and then run this command
python ./model/stylegan/prepare_data.py --out ./data/mystyle/lmdb/ --n_worker 4 --size 1024 ./data/mystyle/images/
It created two files in lmdb folder which are data.mdb and lock.mdb.

and run this command,
python -m torch.distributed.launch --nproc_per_node=1 --master_port=8765 finetune_stylegan.py --iter 600 --batch 4 --ckpt ./checkpoint/stylegan2-ffhq-config-f.pt --style mystyle --augment ./data/mystyle/lmdb/
but faced this error. Could you point out what I am wrong here?

I am using RTX A5000 GPU 24G VRAM

load model: ./checkpoint/stylegan2-ffhq-config-f.pt
0%| | 0/600 [00:00<?, ?it/s]/home/donghyun/Desktop/training/DualStyleGAN/model/stylegan/op/conv2d_gradfix.py:88: UserWarning: conv2d_gradfix not supported on PyTorch 1.12.1+cu113. Falling back to torch.nn.functional.conv2d().
warnings.warn(
0%| | 0/600 [00:01<?, ?it/s]
Traceback (most recent call last):
File "finetune_stylegan.py", line 391, in
train(args, loader, generator, discriminator, g_optim, d_optim, g_ema, device)
File "finetune_stylegan.py", line 159, in train
r1_loss = d_r1_loss(real_pred, real_img)
File "/home/donghyun/Desktop/training/DualStyleGAN/util.py", line 71, in d_r1_loss
grad_real, = autograd.grad(
File "/home/donghyun/anaconda3/envs/dualstylegan_env/lib/python3.8/site-packages/torch/autograd/init.py", line 276, in grad
return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
File "/home/donghyun/anaconda3/envs/dualstylegan_env/lib/python3.8/site-packages/torch/autograd/function.py", line 253, in apply
return user_fn(self, *args)
File "/home/donghyun/Desktop/training/DualStyleGAN/model/stylegan/non_leaking.py", line 352, in backward
grad_input, grad_grid = GridSampleBackward.apply(grad_output, input, grid)
File "/home/donghyun/Desktop/training/DualStyleGAN/model/stylegan/non_leaking.py", line 361, in forward
grad_input, grad_grid = op(grad_output, input, grid, 0, 0, False)
TypeError: 'tuple' object is not callable
ERROR:torch.distributed.elastic.multiprocessing.api:failed (exitcode: 1) local_rank: 0 (pid: 17115) of binary: /home/donghyun/anaconda3/envs/dualstylegan_env/bin/python
Traceback (most recent call last):
File "/home/donghyun/anaconda3/envs/dualstylegan_env/lib/python3.8/runpy.py", line 194, in _run_module_as_main
return _run_code(code, main_globals, None,
File "/home/donghyun/anaconda3/envs/dualstylegan_env/lib/python3.8/runpy.py", line 87, in _run_code
exec(code, run_globals)
File "/home/donghyun/anaconda3/envs/dualstylegan_env/lib/python3.8/site-packages/torch/distributed/launch.py", line 193, in
main()
File "/home/donghyun/anaconda3/envs/dualstylegan_env/lib/python3.8/site-packages/torch/distributed/launch.py", line 189, in main
launch(args)
File "/home/donghyun/anaconda3/envs/dualstylegan_env/lib/python3.8/site-packages/torch/distributed/launch.py", line 174, in launch
run(args)
File "/home/donghyun/anaconda3/envs/dualstylegan_env/lib/python3.8/site-packages/torch/distributed/run.py", line 752, in run
elastic_launch(
File "/home/donghyun/anaconda3/envs/dualstylegan_env/lib/python3.8/site-packages/torch/distributed/launcher/api.py", line 131, in call
return launch_agent(self._config, self._entrypoint, list(args))
File "/home/donghyun/anaconda3/envs/dualstylegan_env/lib/python3.8/site-packages/torch/distributed/launcher/api.py", line 245, in launch_agent
raise ChildFailedError(
torch.distributed.elastic.multiprocessing.errors.ChildFailedError:

finetune_stylegan.py FAILED

Failures:
<NO_OTHER_FAILURES>

Root Cause (first observed failure):
[0]:
time : 2022-11-11_01:36:27
host : kr-03
rank : 0 (local_rank: 0)

from dualstylegan.

https://github.com/williamyang1991/DualStyleGAN/tree/main/model/stylegan/op_cpu
you can change to CPU compatible version to avoid using conv2d_gradfix.py

from dualstylegan.

conv2d_gradfix.py is from stylegan2-pytorch, or you can check similar issues in
https://github.com/rosinality/stylegan2-pytorch/issues
to see if there are any solution

from dualstylegan.

I see. I thought this is because of the multi GPU setting.

I wonder about your training process when creating different models such as cartoon or Pixar.
Did you train all the models from Facial Destylization to optimization? or skip step 2 in Facial Destylization or any other steps?

from dualstylegan.

all process
only skip Progressive Fine-Tuning step 1 & 2 by loading the pretrained model

from dualstylegan.