RunsenFeng

你好，原代码中默认是使用cuda设备运行模型，所以当检测不到cuda设备时，即纯cpu运行时，会报错。现已更新相应纯cpu运行的功能，你可以再pull下来运行试试。 另外，如果你的机器本身有nvidia GPU，但是 torch.cuda.is_available() 的输出为 False，可能nvidia driver的版本比较老，导致cuda toolkit 没有正确安装。

Besides the above answers, we recommend to install minicoda for virtual environment managements on windows. You can refer to the answers of issue [#17](https://github.com/geekyutao/Inpaint-Anything/issues/17).

Yes, the project will autodetect and use the cuda devices, if your torch is correctly compiled with cuda. You can check that using `torch.cuda.is_available()` in python scripts.

Hi, I have successfully run remove_anything.py on win11. Here is my environment setup: 1. Download and install the [Miniconda3 Windows 64-bit](https://docs.conda.io/en/latest/miniconda.html). 2. Run `Anaconda Powershell Prompt (miniconda3)` as administrator. 3....

> Hi, I have successfully run remove_anything.py on win11. Here is my environment setup: > > 1. Download and install the [Miniconda3 Windows 64-bit](https://docs.conda.io/en/latest/miniconda.html). > 2. Run `Anaconda Powershell Prompt...

The main reason is that: the sam feature of a newly uploaded image is pre-computed before you click "Predict Masks Using SAM", which requires several seconds to finish; and if...