Can't update a tensor by assignment

I am trying to add two tensors and save the result into a third tensor that I update every loop. Adding tensors seems hit or miss, as the velocity calculation works no problem and updates but for some reason my position does not change at all and refuses to update. For context, this snippet is a part of a function inside a class, but the variables y_0 and y_1. dy_0 and dy_1 are local variables defined at the top of this function. The output of self.eval_func.evaluate(*self.eval_func.tRP_tensor)).data does give me the proper tensor so this is not the issue.

Here is the relevant snippet:

for i, itt in enumerate(range(self.sim_time)):
    # if t is needed, it is calculated here
    cur_real_time = self.start_val + itt * self.dt

    # Calclate velocity from force
    temp1 = torch.mul(self.dt, self.eval_func.evaluate(*self.eval_func.tRP_tensor)).data
    # print(, = +  # noqa

    # calculate new position from velocity
    temp2 = torch.mul(self.dt, = +

    # update variables this iterations initial and final values
    # for the next iteration = =
    self.update_args(cur_real_time, y_1, dy_1)

    if cur_real_time - last_output_time >= self.output_step:
        # this block is option status output with ETA if check_val
        # is passed as True to the functuion
        if check_val:
            ratio = i / self.sim_time
            cur_time = - start
            ETA = cur_time.seconds / ratio
            print('\rCurrently {:.2f}%. Running for {} seconds : E.T.A {:.2f} Seconds. T - done = {:.2f}'.format(100 * ratio, cur_time, ETA, ETA - cur_time.seconds), end='')  # noqa

        # put the final result for this iteration in the output
        last_output_time = cur_real_time

print('\rCompleted in {}!'.format(
return np.array(out_list)

In this code, y_1 (and y_0) refuses to update with the value of y_1 = y_0 + dt * dy_1. I have this issue sometimes with dy_0 in another function and believe it to the be same issue in a Runge Kutta 4 function but have no idea why it is happening.

The temp1 and temp2 variables exist for debugging purposes so I could see what was actually being assigned when I preformed operations. It has the same behavior if I substitute the definition of temp1 and temp2 into the code and remove the variables.

Thank you very much.

Solved. The error for reference was that I didn’t make each tensor of the same dtype and instead left y_1 without explicitly being declared as dtype=float.