Ziyu Wan

Sorry for the late reply. Could you please send me the image-mask pair so that I could check your case?

Great question. The anwser is yes, in my opinion. Actually, such image representation was very popular on the early-stage computers, which only owns limited bits to store the color information....

Yes. Currently the validation shares the same mask set with the training. You could also pass some extra parameters to change this behavior.

Sure. One gpu is enough for testing. The released code also supports for single-GPU training, yet will lead to a long optimization time.

This error is caused by your network issue. Please try to re-download the pre-trained models via the new added Baidu Driver.

Please try to re-download the pre-trained models via the new added Baidu Driver.

Could you please describe the errors you have met in more details?

Sorry for the late reply due to other ddl. Basically, yes. Feel free to raise any other questions further.

Hi, you could get the mask [here](https://nv-adlr.github.io/publication/partialconv-inpainting). You could also generate the mask on-the-fly by enabling `--random_stroke` option.

Hi, Are your V100 GPUs 32G?