RuizhouLiu

Issue about Multi-GPU training

2

Hi there, I was stuck with an issue that training mask dino model on multiple gpus is too slow. I train the mask dino on my own dataset, but the...

Multi-process for dataloader when set num_workers larger than 0 will raise an error by RGCN model

2

Hi, when I set num_workers > 0, an error are raised. I think it one reason raising the error is the DGL, but i am not sure. Here is the...

An issue raised when debugging.

1

There is an issue `error running command svifpd.watch-track-enable: e.settracked is not a function. this is likely caused by the extension that contributes svifpd.watch-track-enable.` raised when I want to check what...

How to run?

1

Awesome work! But I cannot find guidance for training and testing, could you provide more details to run your code? Thanks!

Hi, I try to re-construct your code with DDP for faster multi-GPU training. However, I encountered some issue. ``` File "/data2/liurzh/Decent_parallel/train.py", line 243, in main() File "/data2/liurzh/Decent_parallel/train.py", line 109, in...

Are there some tips for adding custom models and dataset in the `joliGEN` lib?

1

Thanks for your works! Are there some tips for adding custom models and dataset in the `joliGEN` lib?

In `CUTModel`, the implementations of `compute_G_loss_cut` function may not consist with original implementations.

1

Hi, there! When I run `cut` model, I found that the way of computing `self.loss_G_NCE` and `self.loss_G_NCE_Y` is same. I run config file `example_gan_glasses2noglasses.json`. In `self.loss_G_NCE_Y`: ``` self.loss_G_NCE_Y = self.calculate_NCE_loss(...