cleanup

Author: stefanch

sgdml/cli.py

@@ -1181,6 +1181,7 @@ def reset(command=None, **kwargs):
             print(ui.info_str('[INFO]') + ' Benchmark cache is already empty.')
     else:
         print(' Cancelled.')
+    print('')
 
 
 def main():

sgdml/predict.py

@@ -211,7 +211,7 @@ def __init__(
                 The parameters `batch_size` and `num_workers` are only
                 relevant if this code runs on a CPU. Both can be set
                 automatically via the function
-                `set_opt_num_workers_and_batch_size_fast`. Enabling
+                `prepare_parallel`. Enabling
                 calculations via PyTorch is only recommended with GPU
                 support. CPU calcuations are faster with our NumPy
                 implementation.
@@ -354,7 +354,7 @@ def _set_num_workers(
         Note
         ----
                 This parameter can be optimally determined using
-                `set_opt_num_workers_and_batch_size_fast`.
+                `prepare_parallel`.
 
         Parameters
         ----------
@@ -414,7 +414,7 @@ def _set_batch_size(
         Note
         ----
                 This parameter can be optimally determined using
-                `set_opt_num_workers_and_batch_size_fast`.
+                `prepare_parallel`.
 
         Parameters
         ----------
@@ -441,6 +441,8 @@ def _set_bulk_mp(
 
     def set_opt_num_workers_and_batch_size_fast(self, n_bulk=1, n_reps=1): # deprecated
         """
+        Warning
+        -------
         Deprecated! Please use the function `prepare_parallel` in future projects.
         
         Parameters
@@ -497,7 +499,11 @@ def prepare_parallel(self, n_bulk=1, n_reps=1, return_is_from_cache=False):  # n
                 n_reps : int, optional
                         Number of repetitions (bigger value: more
                         accurate, but also slower).
-
+                return_is_from_cache : bool, optional
+                        If enabled, this function returns a second value
+                        indicating if the returned results were obtained
+                        from cache.
+    
         Returns
         -------
                 int
@@ -883,4 +889,3 @@ def predict(self, r):
         F = desc.r_to_d_desc_op(r, pdist, res[1:], self.ucell_size).reshape(1, -1)
         # F = res[1:].reshape(1,-1).dot(r_d_desc)
         return E, F
-        

sgdml/train.py

@@ -486,12 +486,6 @@ def train(  # noqa: C901
             j_end=M if use_cg else None,
         )
 
-        # if use_cg:
-        #    print(
-        #        ui.info_str('[INFO]')
-        #        + ' Nystroem preconditioner uses %s training points.' % M
-        #        )
-
         # test 2
 
         # use_ny = True
@@ -512,24 +506,13 @@ def train(  # noqa: C901
 
             lam = 1e-8
 
-            # print(M*R_d_desc.shape[2])
-
             # ny_idxs = np.random.choice(K.shape[0], M*R_d_desc.shape[2], replace=False)
-
             # K_mm = K[ny_idxs, :]
             # K_mm = K_mm[:, ny_idxs]
-
             # K_nm = K[:, ny_idxs]
 
-            # P_inv3 = -(1.0/(-lam)) * (np.eye(K.shape[0]) - K_nm.dot(np.linalg.solve((K_mm + K_nm.T.dot(K_nm)),K_nm.T)))
-
-            # P_inv3_new = (-1.0/lam)*np.eye(K.shape[0]) - (-1.0/lam)*K_nm.dot(np.linalg.solve(((-lam)*K_mm + K_nm.T.dot(K_nm)),K_nm.T))
-            # P_inv3 = P_inv3_new
-
             _lup = sp.linalg.lu_factor((-lam) * K_mm + K_nm.T.dot(K_nm))
             def mv(v):
-                # M = (-1.0/lam)*np.eye(K.shape[0]) - (-1.0/lam)*K_nm.dot(np.linalg.solve(((-lam)*K_mm + K_nm.T.dot(K_nm)),K_nm.T))
-                # M = (-1.0/lam)*np.eye(K.shape[0]) - (-1.0/lam)*K_nm.dot(np.linalg.solve(((-lam)*K_mm + K_nm.T.dot(K_nm)),K_nm.T.dot(v)))
                 P_v = -(-1.0 / lam) * (
                     K_nm.dot(sp.linalg.lu_solve(_lup, K_nm.T.dot(v))) - v
                 )
@@ -683,32 +666,10 @@ def callback(xk):
 
                 del K
 
-                # from scipy.sparse.linalg import cg
-
-                # alphas, status = cg(
-                #     -K_op,
-                #     y,
-                #     tol=1e-4,
-                #     maxiter=3 * n_atoms * n_train,
-                #     M=P_op,
-                #     callback=callback,
-                # )  # M=P_inv3
-                # alphas = -alphas
-
-
-                #import scipy.sparse.linalg as spla
-
-                #M2 = spla.spilu(-K_op)
-                #M = spla.LinearOperator(K_op.shape, M2.solve)
-
-
                 from scipy.sparse.linalg import cg
                 alphas, status = cg(-K_op, y, M=P_op, tol=1e-4, maxiter=3 * n_atoms * n_train, callback=callback)
                 alphas = -alphas
 
-                # sys.stdout.flush()
-                # print('\n')
-
         # test 2
 
         # alphas = K_mm_mn.dot(alphas)  # remove me later