"""
Multi-Agent Deep Deterministic Policy Gradient
Paper link:
https://proceedings.neurips.cc/paper/2017/file/68a9750337a418a86fe06c1991a1d64c-Paper.pdf
Implementation: MindSpore
Trick: Parameter sharing for all agents, with agents' one-hot IDs as actor-critic's inputs.
"""
from mindspore.nn import MSELoss
from xuance.mindspore import ms, Module, Tensor, optim, ops
from xuance.mindspore.learners import LearnerMAS
from xuance.mindspore.utils import clip_grads
from xuance.common import List
from argparse import Namespace
from operator import itemgetter
[docs]
class MADDPG_Learner(LearnerMAS):
def __init__(self,
config: Namespace,
model_keys: List[str],
agent_keys: List[str],
policy: Module,
callback):
super(MADDPG_Learner, self).__init__(config, model_keys, agent_keys, policy, callback)
self.optimizer = {
key: {
'actor': optim.Adam(params=self.policy.parameters_actor[key], lr=self.config.learning_rate_actor,
eps=1e-5),
'critic': optim.Adam(params=self.policy.parameters_critic[key], lr=self.config.learning_rate_critic,
eps=1e-5)}
for key in self.model_keys}
self.scheduler = {
key: {'actor': optim.lr_scheduler.LinearLR(self.optimizer[key]['actor'], start_factor=1.0,
end_factor=self.end_factor_lr_decay, total_iters=self.config.running_steps),
'critic': optim.lr_scheduler.LinearLR(self.optimizer[key]['critic'], start_factor=1.0,
end_factor=self.end_factor_lr_decay, total_iters=self.config.running_steps)}
for key in self.model_keys}
self.gamma = config.gamma
self.tau = config.tau
self.mse_loss = MSELoss()
# Get gradient function
self.grad_fn_actor = {key: ms.value_and_grad(self.forward_fn_actor, None,
self.optimizer[key]['actor'].parameters, has_aux=True)
for key in self.model_keys}
self.grad_fn_critic = {key: ms.value_and_grad(self.forward_fn_critic, None,
self.optimizer[key]['critic'].parameters, has_aux=True)
for key in self.model_keys}
self.policy.set_train()
[docs]
def forward_fn_actor(self, batch_size, obs, obs_joint, actions, ids, mask_values, agent_key):
_, actions_eval = self.policy(observation=obs, agent_ids=ids)
if self.use_parameter_sharing:
act_eval = actions_eval[agent_key].reshape(batch_size, self.n_agents, -1).reshape(batch_size, -1)
else:
a_joint = {k: actions_eval[k] if k == agent_key else actions[k] for k in self.agent_keys}
act_eval = ops.cat(itemgetter(*self.agent_keys)(a_joint), axis=-1).reshape(batch_size, -1)
_, q_policy = self.policy.Qpolicy(joint_observation=obs_joint, joint_actions=act_eval,
agent_ids=ids, agent_key=agent_key)
q_policy_i = q_policy[agent_key].reshape(-1)
loss_a = -(q_policy_i * mask_values).sum() / mask_values.sum()
return loss_a, q_policy_i, act_eval
[docs]
def forward_fn_critic(self, obs_joint, actions_joint, ids, mask_values, q_target, agent_key):
_, q_eval = self.policy.Qpolicy(joint_observation=obs_joint, joint_actions=actions_joint, agent_ids=ids,
agent_key=agent_key)
q_eval_a = q_eval[agent_key].reshape(-1)
td_error = (q_eval_a - ops.stop_gradient(q_target)) * mask_values
loss_c = (td_error ** 2).sum() / mask_values.sum()
return loss_c, q_eval_a, td_error
[docs]
def update(self, sample):
self.iterations += 1
# prepare training data
sample_Tensor = self.build_training_data(sample,
use_parameter_sharing=self.use_parameter_sharing,
use_actions_mask=False)
batch_size = sample_Tensor['batch_size']
obs = sample_Tensor['obs']
actions = sample_Tensor['actions']
obs_next = sample_Tensor['obs_next']
rewards = sample_Tensor['rewards']
terminals = sample_Tensor['terminals']
agent_mask = sample_Tensor['agent_mask']
IDs = sample_Tensor['agent_ids']
if self.use_parameter_sharing:
key = self.model_keys[0]
bs = batch_size * self.n_agents
obs_joint = obs[key].reshape(batch_size, -1)
next_obs_joint = obs_next[key].reshape(batch_size, -1)
actions_joint = actions[key].reshape(batch_size, -1)
rewards[key] = rewards[key].reshape(batch_size * self.n_agents)
terminals[key] = terminals[key].reshape(batch_size * self.n_agents)
else:
bs = batch_size
obs_joint = ops.cat(itemgetter(*self.agent_keys)(obs), axis=-1).reshape(batch_size, -1)
next_obs_joint = ops.cat(itemgetter(*self.agent_keys)(obs_next), axis=-1).reshape(batch_size, -1)
actions_joint = ops.cat(itemgetter(*self.agent_keys)(actions), axis=-1).reshape(batch_size, -1)
info = self.callback.on_update_start(self.iterations, method="update",
policy=self.policy, sample_Tensor=sample_Tensor, bs=bs,
obs_joint=obs_joint, next_obs_joint=next_obs_joint,
actions_joint=actions_joint)
# get actions
_, actions_next = self.policy.Atarget(next_observation=obs_next, agent_ids=IDs)
# get values
if self.use_parameter_sharing:
key = self.model_keys[0]
actions_next_joint = actions_next[key].reshape(batch_size, self.n_agents, -1).reshape(batch_size, -1)
else:
actions_next_joint = ops.cat(itemgetter(*self.model_keys)(actions_next), -1).reshape(batch_size, -1)
_, q_next = self.policy.Qtarget(joint_observation=next_obs_joint, joint_actions=actions_next_joint,
agent_ids=IDs)
for key in self.model_keys:
mask_values = agent_mask[key]
# update critic
q_next_i = q_next[key].reshape(bs)
q_target = rewards[key] + (1 - terminals[key]) * self.gamma * q_next_i
(loss_c, q_eval_a, td_error), grads_critic = self.grad_fn_critic[key](obs_joint, actions_joint, IDs,
mask_values, q_target, key)
if self.use_grad_clip:
grads_critic = clip_grads(grads_critic, Tensor(-self.grad_clip_norm), Tensor(self.grad_clip_norm))
self.optimizer[key]['critic'](grads_critic)
# update actor
(loss_a, q_policy_i, act_eval), grads_actor = self.grad_fn_actor[key](
batch_size, obs, obs_joint, actions, IDs, mask_values, key)
if self.use_grad_clip:
grads_actor = clip_grads(grads_actor, Tensor(-self.grad_clip_norm), Tensor(self.grad_clip_norm))
self.optimizer[key]['actor'](grads_actor)
self.scheduler[key]['actor'].step()
self.scheduler[key]['critic'].step()
learning_rate_actor = self.scheduler[key]['actor'].get_last_lr()[0]
learning_rate_critic = self.scheduler[key]['critic'].get_last_lr()[0]
info.update({
f"{key}/learning_rate_actor": learning_rate_actor.asnumpy(),
f"{key}/learning_rate_critic": learning_rate_critic.asnumpy(),
f"{key}/loss_actor": loss_a.asnumpy(),
f"{key}/loss_critic": loss_c.asnumpy(),
f"{key}/predictQ": q_eval_a.mean().asnumpy()
})
info.update(self.callback.on_update_agent_wise(self.iterations, key, info=info, method="update",
mask_values=mask_values, q_policy_i=q_policy_i,
act_eval=act_eval, q_eval_a=q_eval_a, q_next_i=q_next_i,
q_target=q_target, td_error=td_error))
self.policy.soft_update(self.tau)
info.update(self.callback.on_update_end(self.iterations, method="update", policy=self.policy, info=info))
return info