Nanjun He

Results 13 issues of Nanjun He

error: cannot call constructor ‘Regiongrowing::Regiongrowing’ directly [-fpermissive]

Does your code support Multiple GPU training?

Hi It seems that the function `get_1x_lr_params` can only Iterate `nn.Conv2d` in the top node of the backbone. such as the `conv1` , while fails to Iterate 'nn.Conv2d' in subnode,...

![image](https://user-images.githubusercontent.com/27271981/112472422-128a2e80-8da8-11eb-9611-b2f20ba7a52b.png)

The model and loaded state dict do not match exactly size mismatch for rpn_head.rpn_cls.weight: copying a param with shape torch.Size([3, 256, 1, 1]) from checkpoint, the shape in current model...

Hi, I am try to ReIP. the training code. I am wondering how long does it take for training stage one. Best

![image](https://user-images.githubusercontent.com/27271981/147735300-7572cd4b-3622-45f2-a61a-82713ddd66a8.png)

- This is the reconstruction result of 100 Epoch of VQVAE. ![recons_VQVAE_Epoch_99](https://user-images.githubusercontent.com/27271981/204431756-8dd9346a-5ed3-4560-951c-558f1f31e455.png)