Why the padding is different for state, action, reward?

[CeyaoZhang]

decision-transformer/gym/experiment.py

Lines 147 to 154 in c9e6ac0

	s[-1] = np.concatenate([np.zeros((1, max_len - tlen, state_dim)), s[-1]], axis=1)
	s[-1] = (s[-1] - state_mean) / state_std
	a[-1] = np.concatenate([np.ones((1, max_len - tlen, act_dim)) * -10., a[-1]], axis=1)
	r[-1] = np.concatenate([np.zeros((1, max_len - tlen, 1)), r[-1]], axis=1)
	d[-1] = np.concatenate([np.ones((1, max_len - tlen)) * 2, d[-1]], axis=1)
	rtg[-1] = np.concatenate([np.zeros((1, max_len - tlen, 1)), rtg[-1]], axis=1) / scale
	timesteps[-1] = np.concatenate([np.zeros((1, max_len - tlen)), timesteps[-1]], axis=1)
	mask.append(np.concatenate([np.zeros((1, max_len - tlen)), np.ones((1, tlen))], axis=1))

It's easy to understand padding the state with np.zero(,), but why use np.ones(,)* -10 to pad the action and np.ones(,) * 2 to pad the done flag?