Hongbo Zhu

` s_test` is used to calculate the inverse HVP estimation by Stochastic Second-Order Algorithm. `grad_z` can be used to calculate any test point $\nabla_{\theta}L(z_{\text{test}}, \hat{\theta})$ or any training data point...

python train_ggcnn.py --description training_on_jacquard --network ggcnn --batches-per-epoch [value] --dataset jacquard --dataset-path ./Jacquard --vis Just set the [value] as you want

> Can this network handle 1280*720 pictures An easy way is to crop your images and make length = width, such as from 1280* 720 to 720* 720, then resize...

Here is the 'lastframe.zip' given by the other. https://drive.google.com/file/d/1WFiwRi_sMA_McZnkbEjh8Rnl-Im7_9Mk/view It maybe helpful for someone in the same situation as me.