I don’t know what the actual use case is, but take a look at the RGB color space and how a color is defined in it (i.e. via 3 uint8 values). Then create a tensor with the desired color by assigning the R, G, and B value to it.
If you want to use the numpy array as a tensor, you could use tensor = torch.from_numpy(image). Alternatively, if you want to create the same “red tensor” this should work: