python animation_demo.py --config config/end2end.yaml --checkpoint ./ckpt/final_3DV.tar --source_image_pth ./assets/EM.jpeg --driving_video_pth ./assets/02.mp4 --relative --adapt_scale --find_best_frame --cpu gives me:

/Users/user/miniconda3/envs/safa3/lib/python3.7/site-packages/skimage/transform/_warps.py:105: UserWarning: The default mode, 'constant', will be changed to 'reflect' in skimage 0.15.
  warn("The default mode, 'constant', will be changed to 'reflect' in "
/Users/user/miniconda3/envs/safa3/lib/python3.7/site-packages/skimage/transform/_warps.py:110: UserWarning: Anti-aliasing will be enabled by default in skimage 0.15 to avoid aliasing artifacts when down-sampling images.
  warn("Anti-aliasing will be enabled by default in skimage 0.15 to "
animation_demo.py:32: YAMLLoadWarning: calling yaml.load() without Loader=... is deprecated, as the default Loader is unsafe. Please read https://msg.pyyaml.org/load for full details.
  config = yaml.load(f)
blend_scale:  1
/Users/user/miniconda3/envs/safa3/lib/python3.7/site-packages/torch/functional.py:478: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at  /Users/runner/work/pytorch/pytorch/pytorch/aten/src/ATen/native/TensorShape.cpp:2895.)
  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]
/Users/user/miniconda3/envs/safa3/lib/python3.7/site-packages/torchvision/models/_utils.py:209: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and will be removed in 0.15, please use 'weights' instead.
  f"The parameter '{pretrained_param}' is deprecated since 0.13 and will be removed in 0.15, "
/Users/user/miniconda3/envs/safa3/lib/python3.7/site-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and will be removed in 0.15. The current behavior is equivalent to passing `weights=MobileNet_V2_Weights.IMAGENET1K_V1`. You can also use `weights=MobileNet_V2_Weights.DEFAULT` to get the most up-to-date weights.
  warnings.warn(msg)
creating the FLAME Decoder
/Users/user/miniconda3/envs/safa3/lib/python3.7/site-packages/pytorch3d/io/obj_io.py:533: UserWarning: Mtl file does not exist: ./modules/data/template.mtl
  warnings.warn(f"Mtl file does not exist: {f}")
[W NNPACK.cpp:51] Could not initialize NNPACK! Reason: Unsupported hardware.
128it [03:03,  1.48s/it]
Best frame: 120
/Users/user/miniconda3/envs/safa3/lib/python3.7/site-packages/torch/nn/functional.py:4216: UserWarning: Default grid_sample and affine_grid behavior has changed to align_corners=False since 1.3.0. Please specify align_corners=True if the old behavior is desired. See the documentation of grid_sample for details.
  "Default grid_sample and affine_grid behavior has changed "
Traceback (most recent call last):
  File "animation_demo.py", line 216, in <module>
    relative=opt.relative, adapt_movement_scale=opt.adapt_scale, cpu=opt.cpu)
  File "animation_demo.py", line 83, in make_animation
    driving_initial = driving[:, :, 0].cuda()
  File "/Users/user/miniconda3/envs/safa3/lib/python3.7/site-packages/torch/cuda/__init__.py", line 211, in _lazy_init
    raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enabled

it seems to come from the animation_demo.py file with the absence of the if not cpu condition in line 83.

How can I modify the file to solve this properly?

First, I want to congratulate you on an ingenious paper. The idea of using FLAME with face alignment and masking is simply brilliant. I am getting very good results with your current code. However, I would like to ask if you could provide some brief advice on how to raise the resolution (e.g. from 256x256 to 512x512). I attempted to adjust settings in the code, but there were persistent tensor mismatch problems. I presume the dataset must be re-trained at higher resolution? Please advise, if you will have time.

I download SAFA model checkpoint from Google Drive,and use data.pkl to the checkpoint.But there are some errors. File "/home/mpl/anaconda3/envs/py37/lib/python3.7/site-packages/torch/serialization.py", line 764, in _legacy_load magic_number = pickle_module.load(f, **pickle_load_args) _pickle.UnpicklingError: A load persistent id instruction was encountered, but no persistent_load function was specified.

I have downloaded face-parsing.PyTorch, however it attempted to import Resnet18, which is not possible in the publicly available PyTorch versions. The proper way to import is:

from resnet_pytorch import ResNet model = ResNet.from_pretrained('resnet18', num_classes=10)

The following error is thrown if attempting to use pytorch-resnet:

File "/SAFA/face_parsing/model.py", line 112, in forward feat8, feat16, feat32 = self.resnet(x) ValueError: not enough values to unpack (expected 3, got 1)

The face-parsing project does not indicate how Resnet has been implemented. How have you installed it on your system?

According to the documentation (pytorch version 1.11.0): "torch.meshgrid(*tensors) currently has the same behavior as calling numpy.meshgrid(*arrays, indexing=’ij’). In the future torch.meshgrid will transition to indexing=’xy’ as the default."

Without this commit, the following warning is issued:

"UserWarning: Default grid_sample and affine_grid behavior has changed to align_corners=False since 1.3.0. Please specify align_corners=True if the old behavior is desired. See the documentation of grid_sample for details."

Since "align_corners=False" (the default value) is specified twice in the code, I presume that the places where the argument align_corners are not used are meant to have "align_corners=True". I may be wrong, though.

As of imageIO 2.16.0 (Feb22) there are now a v2 and v3 namespaces in addition to the top-level namespace. In order to avoid a deprecation warning, the call to imageio.imread has been replaced by imageio.v2.imread.

The version of imageio in requirements.txt has also been changed to the most recent version of the module (2.22.2).

Prior to generating the output video, I would like to log the normal maps for each frame, if possible. I see that normal maps are being generated in the tdmm_estimator prior to generation and stored in render_ops, but what I need is a 2nd pass normal map for the final, deformed source face in the generator.

In class OcclusionAwareGenerator, could you suggest where is the final data array from which the normal map could be derived, or should it be re-calculated in some way? I haven't had much time to dig into the code, and am hoping that you will immediately know what I am talking about and provide some quick pointers.