Question on loss used in Vanilla REINFORCE implementation

I was looking at the definition of the loss for REINFORCE from Implementing New Algorithms (Basic) — rllab 0.0.1 documentation which had it defined as:
loss
I was wondering why in the example implementation (examples/reinforcement_learning/reinforce.py at 81f47e8ea49c74494d2aa8dc1c9c4ddc6c0eca73 · pytorch/examples · GitHub) there is no division by the number of steps (T)?