dvschultz / ai Goto Github PK

Very convenient repo. Thanks for your effort.

Was trying to figure out how to modify the augmentation policy but its really not clear where and how you do that. At present, its set up to default to random augments. The training command line has a boolean for augmentation, so we can turn it on or off there but looking in the RFAugment.py I see a gazillion different augmentation policies but no indication what flag to pass or where to select Policy=""? Any suggestions?

Hi there,

first of all, thanks for the great work!

I have an issue when running the StyleGAN2_Colab_Train.ipynb notebook on Colab. After downloading the Stylegan2 repo and converting the images to TF records, I run the training script (run_training.py), but the script exits while compling the TensorFlow plugin "upfirdn_2d.cu" (see image).

Any idea of why this happens be and how to fix this?

Hi, I've tried to use a Pix2PixHD trained model (480 entry images and 480 canny-edged, 200 epochs, fine results) with a movie. I've extracted all frames from video, and the applied canny edges to get my inputs. Once through my model, I've found out that the results from consecutive frames, even being almost the same, are quite different . This leads to flickering video once I create movie from synthetised frames...is this an error or misuse?

Thanks in advance

Working through the paperspace version of StyleGan2 training video - all works well until I go to train, then I get this error:

NotImplementedError: Cannot convert a symbolic Tensor (Inputs/minibatch_gpu_in:0) to a numpy array.

All images are 1024x1024, and TF records look the same as in the tutorial. Not sure how to proceed...

Can't upload a .pkl file from my machine to the collab. Fails everytime.

UPDATE: Waited an hour. Working now.

Hello! I'm trying to follow this and I'm running into an issue, I think it's maybe to do with pip? I get as far as running the bash install-stylegan2.sh command before things go wrong. When I run that command it installs loads of things and then starts giving me warnings like this:
INFO: pip is looking at multiple versions of six to determine which version is compatible with other requirements. This could take a while. INFO: pip is looking at multiple versions of requests to determine which version is compatible with other requireme nts. This could take a while. INFO: pip is looking at multiple versions of protobuf to determine which version is compatible with other requireme nts. This could take a while.
This is repeated for a number of other modules. Eventually it says:

ERROR: Cannot install -r requirements.txt (line 28) and Keras-Applications==1.0.7 because these package versions have conflicting dependencies.
The conflict is caused by:
    The user requested Keras-Applications==1.0.7
    tensorflow-gpu 1.15.0 depends on keras-applications>=1.0.8
To fix this you could try to:
1. loosen the range of package versions you've specified
2. remove package versions to allow pip attempt to solve the dependency conflict
ERROR: ResolutionImpossible: for help visit https://pip.pypa.io/en/latest/user_guide/#fixing-conflicting-dependencies

It then completes the installation and it all seems ok, but when I try to run the run_generator.py command it fails immediately, saying Traceback (most recent call last): File "run_generator.py", line 9, in <module> import PIL.Image ModuleNotFoundError: No module named 'PIL'

I think I've worked out that this means that pip has not managed to correctly install the various modules it needs to run this command...so I tried to install them myself. I started with keras-applications v1.0.8 and Pillow. That changed the error message to missing tensorflow, so I installed tensorflow 1.15. Then the error message said it was missing opensimplex, so I
installed that.

Once I had done all that I again ran the run_generator.py command, which generated a lot more activity but it eventually failed with the last line saying RuntimeError: No GPU devices found

I'm pretty clueless as to what has gone wrong. Is there some set of dependencies breaking somewhere? Or am I just doing something wildly wrong? It seems weird to me that all of those errors would then result in a "missing GPU device", so is that error completely unrelated to the previous errors?

Got this error after the command:
!python run_generator.py generate-images --network=/content/drive/My\ Drive/stylegan2-colab-test/stylegan2/results/00002-stylegan2-birdaus-1gpu-config-f/submit_config.pkl --seeds=3875451-3876000 --truncation-psi=0.7

Everything prior were done as in the tutorial but with my own dataset. But then I got:

Local submit - run_dir: results/00006-generate-images
dnnlib: Running run_generator.generate_images() on localhost...
Loading networks from "/content/drive/My Drive/stylegan2-colab-test/stylegan2/results/00002-stylegan2-birdaus-1gpu-config-f/submit_config.pkl"...
Traceback (most recent call last):
  File "run_generator.py", line 490, in <module>
    main()
  File "run_generator.py", line 485, in main
    dnnlib.submit_run(sc, func_name_map[subcmd], **kwargs)
  File "/content/drive/My Drive/stylegan2-colab-test/stylegan2/dnnlib/submission/submit.py", line 343, in submit_run
    return farm.submit(submit_config, host_run_dir)
  File "/content/drive/My Drive/stylegan2-colab-test/stylegan2/dnnlib/submission/internal/local.py", line 22, in submit
    return run_wrapper(submit_config)
  File "/content/drive/My Drive/stylegan2-colab-test/stylegan2/dnnlib/submission/submit.py", line 280, in run_wrapper
    run_func_obj(**submit_config.run_func_kwargs)
  File "/content/drive/My Drive/stylegan2-colab-test/stylegan2/run_generator.py", line 120, in generate_images
    _G, _D, Gs = pretrained_networks.load_networks(network_pkl)
  File "/content/drive/My Drive/stylegan2-colab-test/stylegan2/pretrained_networks.py", line 76, in load_networks
    G, D, Gs = pickle.load(stream, encoding='latin1')
ValueError: too many values to unpack (expected 3)

Hey I am not sure why, but you are trying to install dataset-tools a second time without --user and it is pointing to a wrong path:

Thanks for sharing Convert_pkl_to_pt.ipynb. I do not have a problem with the model provided when run in google colab, but encounter an issue when using my own custom model. My custom model is trained with 512x512 images with stylegan2-ada-pytorch with all default settings.

Traceback (most recent call last):
File "/content/stylegan2-pytorch/convert_weight.py", line 236, in
generator, discriminator, g_ema = pickle.load(f)
ModuleNotFoundError: No module named 'torch_utils'

After I pip install torch_utils, I got the following message:
Traceback (most recent call last):
File "/content/stylegan2-pytorch/closed_form_factorization.py", line 18, in
ckpt = torch.load(args.ckpt)
File "/usr/local/lib/python3.7/dist-packages/torch/serialization.py", line 608, in load
return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
File "/usr/local/lib/python3.7/dist-packages/torch/serialization.py", line 777, in _legacy_load
magic_number = pickle_module.load(f, **pickle_load_args)
ModuleNotFoundError: No module named 'torch_utils.persistence'

What should I do?