-
Notifications
You must be signed in to change notification settings - Fork 0
/
PolicyGradient.py
25 lines (21 loc) · 883 Bytes
/
PolicyGradient.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
import torch
import torch.nn as nn
from torch.distributions.categorical import Categorical
class PolicyGradient():
def __init__(self,input_size,output_size):
self.network=nn.Sequential(
nn.Linear(input_size,64),
nn.Tanh(),
nn.Linear(64,output_size),
nn.Identity())
def get_action(self,obs):
action=self.network(obs)
return Categorical(logits=action).sample().item()
def train_step(self,actions,advantages,observations,optimizer):
observations=torch.as_tensor(observations,dtype=torch.float32)
actions=torch.as_tensor(actions,dtype=torch.float32)
advantages=torch.as_tensor(advantages,dtype=torch.float32)
loss=-(self.network(observations).log_prob(actions)*advantages).mean()
loss.backward()
optimizer.step()
optimizer.zero_grad()