BasicSR

BasicSR copied to clipboard

用自己的数据集训练EDVR

6

您好，非常感谢您作出这个实用的SR工具，现在我想用自己制作的视频数据集去训练EDVR，但是我的预设有些不太清楚的地方

Secant1998

Hello Xintao, I am so glad to use your EDVR model and I want to implement it in C++. I use the Torch.jit.trace and save it as "edvr.pt", but an...

ACALJJ32

NameError: name 'deform_conv_ext' is not defined.

2

Hello Xintao, I have done this: BASICSR_EXT=True python setup.py develop before I started to train. And also confirmed the .so file exists in basicsr/ops. The error is still there. Thanks...

RuijiaoSun

kyz20

RuntimeError:mat1 and mat2 shapes cannot be multiplied(8*2048 and 8192*512)

1

When i train the StyleGAN2 i meet the follow error, i only only changed the yml file. and my own data size is 128*128, how to solve the problem?any advice...

sanersbug

wandb only system information is displayed, without training process and model parameters

5

when i use wandb，there is only system information is displayed, without training process and model parameters  what i can do？

qwerdf4

ZW-PRO

Something wrong with the output image. The system could't load image and shows "CRC" error

2

Dear Xinntao, During training or testing step, I have come across some problems with the output image, on Ubuntu system, some of the images cannot be opened, and it shows...

futureprecd

关于ESRGAN的训练问题

2

涛哥，我在训练ESRGAN的时候，先预训练生成器，然后加载生成器的预训练模型一起训练ESRGAN，训练ESRGAN时我发现出来两个模型，一个是生成器的模型，一个是判别器的模型，在这我有两个问题想你请教： 1.训练ESRGAN和单独训练生成器出来的两个关于生成器的模型有什么不同吗？ 2.测试ESRGAN的时候是加载共同训练ESRGAN得到的生成器的模型？ 如果是这样的话 判别器的模型意义是什么呢？ 感谢涛哥百忙之中能帮忙解答一下..

zkh123456789

How to reduce the training time of BasicSR and accelerate the training process

1

When I use 'torch.backends.cudnn.benchmark=True' to accelerate the training process,the results showed that the training time was not shortened. And can I use fp16 to accelerate the training process?

xuboming8