atari_wrappers.py: Input shape does not seem quite right

[pllee4]

Hi, isn't we need to stack first before invert? By inverting first, I believe the stacked axis is different than expected.

Reinforcement-Learning/Week3/atari_wrappers.py

Lines 160 to 169 in c57064f

	def make_env(env_name, fire=True):
	env = gym.make(env_name)
	env = MaxAndSkipEnv(env) ## Return only every `skip`-th frame
	if fire:
	env = FireResetEnv(env) ## Fire at the beginning
	env = WarpFrame(env) ## Reshape image
	env = ImageToPyTorch(env) ## Invert shape
	env = FrameStack(env, 4) ## Stack last 4 frames
	env = ScaledFloatFrame(env) ## Scale frames
	return env

The current code would produce input shape of
(1, 84, 336)

If we change to

    env = WarpFrame(env) ## Reshape image
    env = FrameStack(env, 4) ## Stack last 4 frames
    env = ImageToPyTorch(env) ## Invert shape

then we would be able to get shape of

(4, 84, 84)