def load_state_dict(self, ckpt):
self._actor.load_state_dict(ckpt['actor_state_dict'])
self._critic.load_state_dict(ckpt['critic_state_dict'])
self._ob_norm.load_state_dict(ckpt['ob_norm_state_dict'])
self._network_cuda(self._config.device)
self._actor_optim.load_state_dict(ckpt['actor_optim_state_dict'])
self._critic_optim.load_state_dict(ckpt['critic_optim_state_dict'])
optimizer_cuda(self._actor_optim, self._config.device)
optimizer_cuda(self._critic_optim, self._config.device)
def load_state_dict(self, ckpt):
self._log_alpha.data = torch.tensor(ckpt['log_alpha'],
requires_grad=True,
device=self._config.device)
self._actor.load_state_dict(ckpt['actor_state_dict'])
self._critic1.load_state_dict(ckpt['critic1_state_dict'])
self._critic2.load_state_dict(ckpt['critic2_state_dict'])
self._critic1_target.load_state_dict(self._critic1.state_dict())
self._critic2_target.load_state_dict(self._critic2.state_dict())
self._ob_norm.load_state_dict(ckpt['ob_norm_state_dict'])
self._network_cuda(self._config.device)
self._alpha_optim.load_state_dict(ckpt['alpha_optim_state_dict'])
self._actor_optim.load_state_dict(ckpt['actor_optim_state_dict'])
self._critic1_optim.load_state_dict(ckpt['critic1_optim_state_dict'])
self._critic2_optim.load_state_dict(ckpt['critic2_optim_state_dict'])
optimizer_cuda(self._alpha_optim, self._config.device)
optimizer_cuda(self._actor_optim, self._config.device)
optimizer_cuda(self._critic1_optim, self._config.device)
optimizer_cuda(self._critic2_optim, self._config.device)
def load_state_dict(self, ckpt):
self._actor.load_state_dict(ckpt["actor_state_dict"])
self._actor_target.load_state_dict(self._actor.state_dict())
self._critic1.load_state_dict(ckpt["critic1_state_dict"])
self._critic2.load_state_dict(ckpt["critic2_state_dict"])
self._critic1_target.load_state_dict(self._critic1.state_dict())
self._critic2_target.load_state_dict(self._critic2.state_dict())
self._network_cuda(self._config.device)
self._actor_optim.load_state_dict(ckpt["actor_optim_state_dict"])
self._critic1_optim.load_state_dict(ckpt["critic1_optim_state_dict"])
self._critic2_optim.load_state_dict(ckpt["critic2_optim_state_dict"])
optimizer_cuda(self._actor_optim, self._config.device)
optimizer_cuda(self._critic1_optim, self._config.device)
optimizer_cuda(self._critic2_optim, self._config.device)
def load_state_dict(self, ckpt):
self._log_alpha.data = torch.tensor(ckpt['log_alpha'],
requires_grad=True,
device=self._config.device)
self._actor.load_state_dict(ckpt['actor_state_dict'])
self._critic1.load_state_dict(ckpt['critic1_state_dict'])
self._critic2.load_state_dict(ckpt['critic2_state_dict'])
self._critic1_target.load_state_dict(self._critic1.state_dict())
self._critic2_target.load_state_dict(self._critic2.state_dict())
self._alpha_optim.load_state_dict(ckpt['alpha_optim_state_dict'])
self._actor_optim.load_state_dict(ckpt['actor_optim_state_dict'])
self._critic1_optim.load_state_dict(ckpt['critic1_optim_state_dict'])
self._critic2_optim.load_state_dict(ckpt['critic2_optim_state_dict'])
optimizer_cuda(self._alpha_optim, self._config.device)
optimizer_cuda(self._actor_optim, self._config.device)
optimizer_cuda(self._critic1_optim, self._config.device)
optimizer_cuda(self._critic2_optim, self._config.device)
if self._config.policy == 'cnn' and self._config.unsup_algo == 'curl':
self._curl.load_state_dict(ckpt['curl_state_dict'])
self._encoder_optim.load_state_dict(
ckpt['encoder_optim_state_dict'])
self._cpc_optim.load_state_dict(ckpt['cpc_optim_state_dict'])
optimizer_cuda(self._encoder_optim, self._config.device)
optimizer_cuda(self._cpc_optim, self._config.device)
self._network_cuda(self._config.device)
def load_state_dict(self, ckpt):
self._dqn.load_state_dict(ckpt['dqn_state_dict'])
self._network_cuda(self._config.device)
self._dqn_optim.load_state_dict(ckpt['dqn_optim_state_dict'])
optimizer_cuda(self._dqn_optim, self._config.device)