bfloat16

关于GPU的使用率上不去的问题 About the problem of GPU usage not increasing

1

运行命令:python inference_realesrgan_video.py -i inputs/01.mp4 -n realesr-animevideov3 -s 2 --suffix outx2 command:python inference_realesrgan_video.py -i inputs/01.mp4 -n realesr-animevideov3 -s 2 --suffix outx2 即使加上--num_process_per_gpu参数后,运行速度依然非常缓慢 Even after adding the --num_process_per_gpu parameter, still very slow...

### What feature would you like to see?  As shown in the picture above (this picture is from PS), add multiple other encoding columns of equal width to facilitate...

Optimised graphics card recognition and half-precision recognition

2

The torch itself recognises obsolete graphics drivers, so there is no need to restrict older cards. Requiring graphics memory greater than or equal to 5.5G (that is, 6G or more,...

Note: if you do not provide all of the following information I will directly ignore and close this issue - Il2CppDumper version 6.7.40 - Target Unity version (optional) 2019.4.31f1 -...

Garbro has a bit of a problem determining whether a file is encrypted or not (CatSystem2 .int)

1

When unpacking "Grisaia no Meikyuu -LE LABYRINTHE DE LA GRISAIA-" and "Grisaia no Rakuen -LE EDEN DE LA GRISAIA-", Garbro didn't pop up a dialogue box asking for a key,...

Issues Related to TensorRT Accelerated Inference

1

After separating STFT and ISTFT from the `BSRoformer` class, I was able to successfully export the model to ONNX, and `trtexec` could convert the ONNX model to a TensorRT engine....

Any plans to support exporting onnx and TensorRT engine?

7

for bs roformer model

强制切片推理会导致模型误判景深

1

感谢作者开源的HAT，在动漫数据集上面进行训练取得了非常好的效果。 目前遇到的问题是：显存不足需要强制切片推理，并且图片前后景有虚化，人物应该是清晰的场景很容易误判景深（如下图绿色框所示）：  是否有可能能把全局的信息同步给每个tile，以此来纠正模型超分每个tile的行为？（尝试过”滑窗“等切片策略，但依然不能根除bad case）