I tried to install FastSD CPU in a new virtual environment and I found an issue that I'd like someone to confirm. As per the install.sh script, torch is first installed by itself before continuing to install the remaining requirements; however, here's a rundown of the installation commands:
uv pip install torch --index-url https://download.pytorch.org/whl/cpu <<< Installs torch 2.7.1+cpu
uv pip install -r "$BASEDIR/requirements.txt" <<< Uninstalls torch 2.7.1+cpu and installs torch 2.5.1
So the installation script ends up downloading torch twice. It seems the issue can be solved by upgrading typing-extensions to version 4.10.0 in requirements.txt and installing the CPU version of torchvision in the installation script:
uv pip install torch torchvision --index-url https://download.pytorch.org/whl/cpu
Can anyone confirm?.
I tried to install FastSD CPU in a new virtual environment and I found an issue that I'd like someone to confirm. As per the install.sh script, torch is first installed by itself before continuing to install the remaining requirements; however, here's a rundown of the installation commands:
uv pip install torch --index-url https://download.pytorch.org/whl/cpu<<< Installs torch 2.7.1+cpuuv pip install -r "$BASEDIR/requirements.txt"<<< Uninstalls torch 2.7.1+cpu and installs torch 2.5.1So the installation script ends up downloading torch twice. It seems the issue can be solved by upgrading typing-extensions to version 4.10.0 in requirements.txt and installing the CPU version of torchvision in the installation script:
uv pip install torch torchvision --index-url https://download.pytorch.org/whl/cpuCan anyone confirm?.