def _normalize_data(self, obs, ret):
obs_normalized = normalize(obs, self.normalization['obs'][0],
self.normalization['obs'][1])
ret_normalized = normalize(ret, self.normalization['ret'][0],
self.normalization['ret'][1])
return obs_normalized, ret_normalized
def _normalize_data(self, obs, act, obs_next=None):
obs_normalized = normalize(obs, self.normalization['obs'][0],
self.normalization['obs'][1])
actions_normalized = normalize(act, self.normalization['act'][0],
self.normalization['act'][1])
if obs_next is not None:
delta = obs_next - obs
deltas_normalized = normalize(delta,
self.normalization['delta'][0],
self.normalization['delta'][1])
return obs_normalized, actions_normalized, deltas_normalized
else:
return obs_normalized, actions_normalized
def _normalize_data(self, obs, act, delta=None):
assert len(obs) == len(act) == self.num_models
assert self.normalization is not None
norm_obses = []
norm_acts = []
norm_deltas = []
for i in range(self.num_models):
norm_obs = normalize(obs[i], self.normalization[i]['obs'][0], self.normalization[i]['obs'][1])
norm_act = normalize(act[i], self.normalization[i]['act'][0], self.normalization[i]['act'][1])
norm_obses.append(norm_obs)
norm_acts.append(norm_act)
if delta is not None:
assert len(delta) == self.num_models
norm_delta = normalize(delta[i], self.normalization[i]['delta'][0], self.normalization[i]['delta'][1])
norm_deltas.append(norm_delta)
if delta is not None:
return norm_obses, norm_acts, norm_deltas
return norm_obses, norm_acts