# MIT License
# Copyright (c) 2021 MARL @ SJTU
# Author: Ming Zhou
# Permission is hereby granted, free of charge, to any person obtaining a copy
# of this software and associated documentation files (the "Software"), to deal
# in the Software without restriction, including without limitation the rights
# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
# copies of the Software, and to permit persons to whom the Software is
# furnished to do so, subject to the following conditions:
# The above copyright notice and this permission notice shall be included in all
# copies or substantial portions of the Software.
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
# SOFTWARE.
from argparse import Namespace
from typing import Type, Dict, Any, Sequence
import torch
import numpy as np
from torch import optim
from malib.rl.common.trainer import Trainer
from malib.utils.data import Postprocessor
from malib.utils.typing import AgentID
from malib.utils.tianshou_batch import Batch
[docs]class PGTrainer(Trainer):
[docs] def setup(self):
self.optimizer: Type[optim.Optimizer] = getattr(
optim, self.training_config["optimizer"]
)(self.policy.parameters()["actor"], lr=self.training_config["lr"])
self.lr_scheduler: torch.optim.lr_scheduler.LambdaLR = None
self.ret_rms = None
[docs] def post_process(self, batch: Batch, agent_filter: Sequence[AgentID]) -> Batch:
# v_s_ = np.full(indices.shape, self.ret_rms.mean)
unnormalized_returns, _ = Postprocessor.compute_episodic_return(
batch, gamma=self.training_config["gamma"], gae_lambda=1.0
)
if self.training_config["reward_norm"] and self.ret_rms is not None:
batch["returns"] = (unnormalized_returns - self.ret_rms.mean) / np.sqrt(
self.ret_rms.var + self._eps
)
self.ret_rms.update(unnormalized_returns)
else:
batch["returns"] = unnormalized_returns
batch["logits"], _ = self.policy.actor(
batch.obs, state=batch.get("state", None)
)
return batch
[docs] def train(self, batch: Batch) -> Dict[str, Any]:
self.optimizer.zero_grad()
logits = batch.logits
dist = self.policy.dist_fn.proba_distribution(logits)
act = batch.act
ret = batch.returns
log_prob = dist.log_prob(act).reshape(len(ret), -1).transpose(0, 1)
loss = -(log_prob * ret).mean()
loss.backward()
self.optimizer.step()
if self.lr_scheduler is not None:
self.lr_scheduler.step()
return {"avg_loss": loss.item()}