DiffBIR

DiffBIR copied to clipboard

这是内存不足了吗？(RuntimeError: Unable to find a valid cuDNN algorithm to run convolution)

5

我按照步骤安装，然后运行。在转换图片时出现如下错误： RuntimeError: Unable to find a valid cuDNN algorithm to run convolution CUDA 安装没有问题啊，这是不是表示GPU内存不够了？

ruanqizhen

Failure during training

3

I started training on A100 GPU with about 2000 training images. It completed about 900 Epochs, then the process ended abruptly without any errors. I can see several checkpoint `step`...

kirrukirru

Multi GPU Support?

2

Loving the results, but maxing out GPU 24GB VRAM. Can this be run with mutli GPUs, where it will continue on the second GPU so it doesn't run out of...

AIisCool

About the Training Time

1

Dear author,thanks for your excellent work. I have a question how many days did it take to train stage2 on 8 A100s?

Chantec

How to train DiffBIR for face retouching?

5

Hi, thanks for sharing your awesome work. I want to train DiffBIR on 1024 x1024 [ffhqr dataset](https://github.com/skylab-tech/ffhqr-dataset). I want to use [ffhq dataset](https://github.com/NVlabs/ffhq-dataset/tree/master) as input (1024x024). I have modified [codeformer](https://github.com/XPixelGroup/DiffBIR/blob/main/dataset/codeformer.py)...

luser350

optimizer selection

6

hi, thanks for this excellent work. I have noticed that in the code of train-stage1.py line 106, the optimizer is AdamW opt = torch.optim.AdamW( swinir.parameters(), lr=cfg.train.learning_rate, weight_decay=0 ) but in...

xuanxu92

有关restoration module的问题

1

很棒的工作！ 我有一个问题想请教一下。文章里提到：“During inference, we replace our trained restoration mod- ule with off-the-shelf task-specific restoration models: BSR- Net [73]2 for BSR, SwinIR [29]3 used in DifFace [67] for BFR, and SCUNet-PSNR...

JZPeterPan

测试了4090，250px的放大4倍也会爆显存，有没有办法设置显存使用大小？

thom802311

有关测试需要的显存问题

1

您好，首先很感谢能推出v2版本，能够便于运行和理解。目前我是用24G显存的3090运行代码。请问为什么在V1版本上能进行4倍超分的图像，在V2版本中只能进行2倍超分。想知道什么原因导致的？

AIchemist369

使用tiled对大图像执行SR时失败了

1

非常感谢您的作品。 在使用参数 `--tiled --tile_size 512 --tile_stride 256`，并在执行前使用了 ``` export PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True export PYTORCH_CUDA_ALLOC_CONF=max_split_size_mb:64 ``` 对1024px\*1024px的图像进行SR（V2）时，在24G显存和48G显存的机器上均可以成功执行。 但当对1536px\*1536px的图像进行SR（V2）时，在24G显存和48G显存的机器上均无法执行。 错误信息如下所示，该错误信息是在48G显存的机器上出现的，它尝试申请81GB内存，注定是失败了。 在24G显存的机器上执行时，它尝试申请9GB内存，但仍然失败了。 对同一张图像处理时，申请的内存大小相差甚远，这点感觉也很奇怪。 通过代码了解到仅在Stage2时用到了tiled选项，而在Stage1时是对整个图像进行处理。 请问是否有办法解决呢 十分感谢，期待您的回复！ ``` Traceback (most recent call last): File "/home/yangfan/work/DiffBIR/inference.py", line...

YangFan1997