DWSF
DWSF copied to clipboard
bytedance
code repository for Practical Deep Dispersed Watermarking with Synchronization
张量尺寸错误 RuntimeError: output with shape [1, 128, 128] doesn't match the broadcast shape [3, 128, 128]
完整的报错信息 Train Epoch: 0 [0/394] D_loss: 1.4287535548210144 = DisF:0.9066857099533081 + DisT:0.5220678448677063 En_De_loss: 0.6271777153015137 = Gen:0.5223720073699951 + En:1.126967191696167=MSE-0.49768513441085815+MSSIM-125.85639953613281, De:0.4012618660926819 Traceback (most recent call last): File "train_ed.py", line 90, in for batch_idx,...
In the generate_random_coor function, there is splitsizeH = max(splitSize * 1.3, min(height // 4, splitSize * 4)) splitsizeW = max(splitSize * 1.3, min(width // 4, splitSize * 4)) Why is...
在某些较低版本的 torch/python 版本中，可能会触发找不到 AdamW 优化器，改为正确大小写拼写可解决 ``` # cherrling @ viper in ~/code/DWSF on git:main x [8:51:24] $ ./train_ed.sh Traceback (most recent call last): File "train_ed.py", line 64, in optimizer =...
What if the watermark capacity is increased during model training? For example, with 60 bit watermark information, can the model still achieve this effect? I am currently training a 60...
