修改了算法代码,并建立了一个简单的训练脚本.修改bert处理二维输入,移除PPO的permute参数

Algorithm/policy.py

@@ -1,7 +1,10 @@
 import torch
 import numpy as np
 from torch import nn
-from .utils import build_mlp, reparameterize, evaluate_lop_pi
+try:
+    from .utils import build_mlp, reparameterize, evaluate_lop_pi
+except ImportError:
+    from utils import build_mlp, reparameterize, evaluate_lop_pi
 
 class StateIndependentPolicy(nn.Module):