More tests

test.py

@@ -133,7 +133,7 @@ class Perlin_Noise():
 
 
 class Perlin_PCA_Noise():
-    def __init__(self, dim_a=2, kernel_func='SE_1.41_1', window=64, ssf=-1, f_sigma=1):
+    def __init__(self, dim_a=2, kernel_func='SE_1.41_1.0', window=128, ssf=-1, f_sigma=1):
         self.dim_a = dim_a
         self.kernel_func = kernel_func
         self.window = window
@@ -162,7 +162,7 @@ class Perlin_PCA_Noise():
 
 
 class PCA_Noise():
-    def __init__(self, dim_a=2, kernel_func='SE_1.41_1', window=64, ssf=-1, f_sigma=1):
+    def __init__(self, dim_a=2, kernel_func='SE_1.41_1.0', window=128, ssf=-1, f_sigma=1):
         self.dim_a = dim_a
         self.kernel_func = kernel_func
         self.window = window
@@ -187,6 +187,35 @@ class PCA_Noise():
         self.traj = [[0]*self.dim_a]
 
 
+class Human_PCA_Noise():
+    def __init__(self, dim_a=2, kernel_func='SE_1.414_1.0', window=128, ssf=-1, f_sigma=1):
+        self.dim_a = dim_a
+        self.kernel_func = kernel_func
+        self.window = window
+        self.ssf = ssf
+        self.f_sigma = f_sigma
+        self.index = 0
+        self.reset()
+
+    def __call__(self, obs, env):
+        if self.ssf != -1 and self.index % self.ssf == 0:
+            self.traj = [[0]*len(self.traj[0])]
+        traj = th.Tensor(self.traj).unsqueeze(0)
+        eps = human_input(obs, env)
+        epsilon = th.Tensor(eps).unsqueeze(0)
+        sample = self.dist.sample(traj, self.f_sigma, epsilon).squeeze(0)
+        self.traj.append(sample)
+        self.index += 1
+        return sample
+
+    def reset(self):
+        self.dist = pca.PCA_Distribution(
+            action_dim=self.dim_a, par_strength='CONT_DIAG', kernel_func=self.kernel_func, window=self.window)
+        self.dist.proba_distribution(th.Tensor([[0]*2]), th.Tensor([[1]*2]))
+
+        self.traj = [[0]*self.dim_a]
+
+
 class Colored_PCA_Noise():
     def __init__(self, beta=1, dim_a=2, samples=2**18, kernel_func='SE_1.41_1', window=64, ssf=-1):
         self.beta = beta
@@ -199,11 +228,11 @@ class Colored_PCA_Noise():
         self.reset()
 
     def __call__(self, obs, env):
-        epsilon = self.samples[:, self.index]
+        epsilon = th.Tensor(self.samples[:, self.index]).unsqueeze(0)
         if self.ssf != -1 and self.index % self.ssf == 0:
             self.traj = [[0]*len(self.traj[0])]
         traj = th.Tensor(self.traj).unsqueeze(0)
-        sample = self.dist.sample(traj).squeeze(0)
+        sample = self.dist.sample(traj, epsilon=epsilon).squeeze(0)
         self.traj.append(sample)
         self.index += 1
         return sample
@@ -223,7 +252,7 @@ def rand_seed():
 
 def choosePlayType():
     options = {'human': human_input, 'PCA': PCA_Noise(),
-               'REX': Colored_Noise(beta=0), 'PINK': Colored_Noise(beta=1), 'BROWN': Colored_Noise(beta=2), 'BETA.5': Colored_Noise(beta=.5), 'PINK_PCA': Colored_PCA_Noise(beta=1), 'Precise_PCA': PCA_Noise(f_sigma=0.33), 'Perlin': Perlin_Noise(scale=0.05, octaves=1), 'FastPerlin': Perlin_Noise(scale=0.2, octaves=1), 'SlowPerlin': Perlin_Noise(scale=0.0125, octaves=1), 'Perlin_3': Perlin_Noise(scale=0.05, octaves=3), 'Perlin_8': Perlin_Noise(scale=0.05, octaves=8), 'Perlin_PCA': Perlin_PCA_Noise()}
+               'REX': Colored_Noise(beta=0), 'PINK': Colored_Noise(beta=1), 'BROWN': Colored_Noise(beta=2), 'BETA.5': Colored_Noise(beta=.5), 'PINK_PCA': Colored_PCA_Noise(beta=1), 'Precise_PCA': PCA_Noise(f_sigma=0.33), 'Perlin': Perlin_Noise(scale=0.05, octaves=1), 'FastPerlin': Perlin_Noise(scale=0.2, octaves=1), 'SlowPerlin': Perlin_Noise(scale=0.0125, octaves=1), 'Perlin_3': Perlin_Noise(scale=0.05, octaves=3), 'Perlin_8': Perlin_Noise(scale=0.05, octaves=8), 'Perlin_PCA': Perlin_PCA_Noise(), 'Human_PCA': Human_PCA_Noise()}
     for i, name in enumerate(options):
         print('['+str(i)+'] '+name)
     while True: