rename noise policies properly

kengz · Apr 19, 2017 · be6c2a9 · be6c2a9
1 parent c650b04
commit be6c2a9
Show file tree

Hide file tree

Showing 3 changed files with 8 additions and 33 deletions.
diff --git a/rl/policy/noise.py b/rl/policy/noise.py
@@ -60,7 +60,7 @@ def update(self, sys_vars):
             self.n_step = sys_vars['epi']
 
 
-class AnnealedGaussian(LinearNoisePolicy):
+class AnnealedGaussianPolicy(LinearNoisePolicy):
 
     '''
     Base class of random noise policy for DDPG
@@ -71,7 +71,7 @@ class AnnealedGaussian(LinearNoisePolicy):
     def __init__(self, env_spec, exploration_anneal_episodes,
                  mu, sigma, sigma_min,
                  **kwargs):  # absorb generic param without breaking
-        super(AnnealedGaussian, self).__init__(
+        super(AnnealedGaussianPolicy, self).__init__(
             env_spec, exploration_anneal_episodes)
         self.size = env_spec['action_dim']
         self.mu = mu
@@ -92,12 +92,12 @@ def current_sigma(self):
         return sigma
 
 
-class GaussianWhiteNoise(AnnealedGaussian):
+class GaussianWhiteNoisePolicy(AnnealedGaussianPolicy):
 
     def __init__(self, env_spec, exploration_anneal_episodes=20,
                  mu=0., sigma=.3, sigma_min=None,
                  **kwargs):  # absorb generic param without breaking
-        super(GaussianWhiteNoise, self).__init__(
+        super(GaussianWhiteNoisePolicy, self).__init__(
             env_spec, exploration_anneal_episodes,
             mu, sigma, sigma_min)
 
@@ -106,7 +106,7 @@ def sample(self):
         return sample
 
 
-class OUNoise(AnnealedGaussian):
+class OUNoisePolicy(AnnealedGaussianPolicy):
 
     '''
     Based on
@@ -116,7 +116,7 @@ class OUNoise(AnnealedGaussian):
     def __init__(self, env_spec, exploration_anneal_episodes=20,
                  theta=.15, mu=0., sigma=.3, dt=1e-2, x0=None, sigma_min=None,
                  **kwargs):  # absorb generic param without breaking
-        super(OUNoise, self).__init__(
+        super(OUNoisePolicy, self).__init__(
             env_spec, exploration_anneal_episodes,
             mu, sigma, sigma_min,
             **kwargs)

diff --git a/rl/spec/component_locks.json b/rl/spec/component_locks.json
@@ -20,10 +20,10 @@
       "DDPG"
     ],
     "Policy": [
-      "GaussianWhiteNoise",
+      "GaussianWhiteNoisePolicy",
       "LinearNoisePolicy",
       "NoNoisePolicy",
-      "OUNoise"
+      "OUNoisePolicy"
     ]
   },
   "actor_critic": {

diff --git a/rl/spec/dev_experiment_specs.json b/rl/spec/dev_experiment_specs.json
@@ -120,30 +120,5 @@
       "gamma": [0.97, 0.99],
       "lr": [0.01, 0.1]
     }
-  },
-  "dev": {
-    "problem": "CartPole-v0",
-    "Agent": "DDPG",
-    "HyperOptimizer": "GridSearch",
-    "Memory": "LinearMemoryWithForgetting",
-    "Optimizer": "AdamOptimizer",
-    "Policy": "OUNoise",
-    "PreProcessor": "NoPreProcessor",
-    "param": {
-      "lr": 0.01,
-      "decay": 0.0,
-      "gamma": 0.99,
-      "hidden_layers": [64],
-      "hidden_layers_activation": "sigmoid",
-      "output_layer_activation": "linear",
-      "exploration_anneal_episodes": 10,
-      "auto_architecture": false,
-      "num_hidden_layers": 3,
-      "first_hidden_layer_size": 512
-    },
-    "param_range": {
-      "gamma": [0.97, 0.99],
-      "lr": [0.01, 0.1]
-    }
   }
 }