Merge pull request #8 from dfki-ric/refactor/cleanup_cython

Clean up cython module

benchmarks/benchmark_cartesian_dmp.py

@@ -0,0 +1,26 @@
+from functools import partial
+import numpy as np
+from movement_primitives.dmp import CartesianDMP
+import timeit
+
+
+start_y = np.array([0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0])
+goal_y = np.array([1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0])
+dt = 0.01
+int_dt = 0.001
+
+dmp = CartesianDMP(execution_time=1.0, dt=dt, n_weights_per_dim=6, int_dt=int_dt)
+dmp.configure(start_y=start_y, goal_y=goal_y)
+dmp.set_weights(1000 * np.random.randn(*dmp.get_weights().shape))
+
+times = timeit.repeat(partial(dmp.open_loop, quaternion_step_function="cython"), repeat=10, number=1)
+print("RK4 + Cython")
+print("Mean: %.5f; Std. dev.: %.5f" % (np.mean(times), np.std(times)))
+
+times = timeit.repeat(partial(dmp.open_loop, quaternion_step_function="python"), repeat=10, number=1)
+print("RK4 + Python")
+print("Mean: %.5f; Std. dev.: %.5f" % (np.mean(times), np.std(times)))
+
+times = timeit.repeat(partial(dmp.open_loop, step_function="euler", quaternion_step_function="python"), repeat=10, number=1)
+print("Euler + Python")
+print("Mean: %.5f; Std. dev.: %.5f" % (np.mean(times), np.std(times)))

benchmarks/benchmark_dmp.py

@@ -12,7 +12,20 @@
 
 dmp = DMP(n_dims=n_dims, execution_time=1.0, dt=dt, n_weights_per_dim=6, int_dt=int_dt)
 dmp.configure(start_y=start_y, goal_y=goal_y)
-dmp.forcing_term.weights = 1000 * np.random.randn(*dmp.forcing_term.weights.shape)
+dmp.set_weights(1000 * np.random.randn(*dmp.get_weights().shape))
 
 times = timeit.repeat(partial(dmp.open_loop, step_function="rk4"), repeat=10, number=1)
+print("RK4")
+print("Mean: %.5f; Std. dev.: %.5f" % (np.mean(times), np.std(times)))
+
+times = timeit.repeat(partial(dmp.open_loop, step_function="euler"), repeat=10, number=1)
+print("Euler (Python)")
+print("Mean: %.5f; Std. dev.: %.5f" % (np.mean(times), np.std(times)))
+
+times = timeit.repeat(partial(dmp.open_loop, step_function="euler-cython"), repeat=10, number=1)
+print("Euler (Cython)")
+print("Mean: %.5f; Std. dev.: %.5f" % (np.mean(times), np.std(times)))
+
+times = timeit.repeat(partial(dmp.open_loop, step_function="rk4-cython"), repeat=10, number=1)
+print("RK4 (Cython)")
 print("Mean: %.5f; Std. dev.: %.5f" % (np.mean(times), np.std(times)))

benchmarks/benchmark_dmp_phase.py

@@ -0,0 +1,16 @@
+from functools import partial
+import numpy as np
+from movement_primitives.dmp._canonical_system import canonical_system_alpha
+from movement_primitives.dmp._canonical_system import phase as phase_python
+from movement_primitives.dmp_fast import phase as phase_cython
+import timeit
+
+
+goal_t = 1.0
+start_t = 0.0
+int_dt = 0.001
+alpha = canonical_system_alpha(0.01, goal_t, start_t, int_dt)
+times = timeit.repeat(partial(phase_python, 0.5, alpha, goal_t, start_t, int_dt), repeat=1000, number=1000)
+print("Mean: %.5f; Std. dev.: %.5f" % (np.mean(times), np.std(times)))
+times = timeit.repeat(partial(phase_cython, 0.5, alpha, goal_t, start_t, int_dt), repeat=1000, number=1000)
+print("Mean: %.5f; Std. dev.: %.5f" % (np.mean(times), np.std(times)))

benchmarks/benchmark_dual_dmp.py

@@ -1,18 +1,22 @@
+from functools import partial
 import numpy as np
 from movement_primitives.dmp import DualCartesianDMP
 import timeit
 
 
 start_y = np.array([0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0])
 goal_y = np.array([1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0])
-dt = 0.001
-int_dt = 0.0001
+dt = 0.01
+int_dt = 0.001
 
 dmp = DualCartesianDMP(execution_time=1.0, dt=dt, n_weights_per_dim=6, int_dt=int_dt)
 dmp.configure(start_y=start_y, goal_y=goal_y)
-dmp.forcing_term.weights = 1000 * np.random.randn(*dmp.forcing_term.weights.shape)
+dmp.set_weights(1000 * np.random.randn(*dmp.get_weights().shape))
 
-times = timeit.repeat(dmp.open_loop, repeat=10, number=1)
+times = timeit.repeat(partial(dmp.open_loop, step_function="cython"), repeat=10, number=1)
+print("Cython")
+print("Mean: %.5f; Std. dev.: %.5f" % (np.mean(times), np.std(times)))
+
+times = timeit.repeat(partial(dmp.open_loop, step_function="python"), repeat=10, number=1)
+print("Python")
 print("Mean: %.5f; Std. dev.: %.5f" % (np.mean(times), np.std(times)))
-# Pure python
-# Mean: 0.58188; Std. dev.: 0.00225

movement_primitives/dmp/_canonical_system.py

@@ -69,7 +69,3 @@ def phase(t, alpha, goal_t, start_t, int_dt=0.001, eps=1e-10):
     execution_time = goal_t - start_t
     b = max(1.0 - alpha * int_dt / execution_time, eps)
     return b ** ((t - start_t) / int_dt)
-
-
-# uncomment to overwrite with Cython implementation
-#from ..dmp_fast import phase