@IvanKobzarev do we have a libtorch for vulkan backend?
Also, I tried running .ptl file generated with backend=‘vulkan’ on a Android phone with c++ interface. I see that the tensor.device().is_vulkan() returns True (tensors are getting copied to vulkan) but when I run the module.forward() it throws an error Throws Could not run ‘prepacked::conv2d_clamp_prepack’ with arguments from the ‘vulkan’ backend. Can you help me here?