Open goldwater668 opened 1 year ago
Yes, you need to have a W&B account and log in to your account using wand login
before running the program
@jiyegege why do you do this?
@seawater668 If you don't need W&B, you can also change the code to cancel this way, W&B is just to facilitate me to record training loss。you also can use wandb offline
disable W&B sync
@jiyegege AttributeError: 'VGGCaffePreTrained' object has no attribute 'features'
@seawater668 what’s your Python version,my version is 3.8.16
@jiyegege Python 3.8.16
@seawater668 I have run the code on both my macos and liunx machines, and there is no error in your feedback. But according to your feedback, the code error may appear in the backtone.py
file, have you changed this file
@jiyegege This file has not been changed, does vgg19.npy need to be downloaded in advance
@jiyegege File "train.py", line 123, in main model.to_onnx('animeGan.onnx', input_sample=torch.randn(1, 3, 256, 256))
RuntimeError: ONNX export failed: Couldn't export operator aten::upsample_bilinear2d
After training for 40 epochs, the test result does not change the real face. What I am training is the cartoonization of the face.
@seawater668 If you're just stylizing faces, you'll probably need to make your own data sets and extract face parts for stylizing. There's animeGan's pytorch open source project on github for faces
The face part has been extracted
AttributeError: 'VGGCaffePreTrained' object has no attribute 'features' wandb: Waiting for W&B process to finish... (failed 1). wandb: You can sync this run to the cloud by running: wandb: wandb sync ./wandb/offline-run-20230318_194510-3tktpfql wandb: Find logs at: ./wandb/offline-run-20230318_194510-3tktpfql/logs