If I have this tensor ks=[1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0. 8.0] and a function myfunc(k:torch.tensor, s:float)->torch.tensor: # do something return torch.tensor([2.3, 3.4, 5.1, 2.2]) Return value of myfunc is not a constant of the above numbers. Its just to illustrate the tensor size. At the moment, I am looping as follows. ks_indices = torch.tensor(list(range(4)) for i in range(5): k_set = torch.index_select(ks, 0, ks_indices.view(-1)) result_tensor = myfunc(k_set, ks[0]) ks = torch.roll(ks, -1) Is it possible to avoid the loop and make a single tensor call to get all five result_tensor's as a 5x4 tensor in one go? Continue reading...
