update results of training.

kgcnn/models/casting.py

@@ -163,7 +163,7 @@ def template_cast_list_input(model_inputs,
 
     if input_tensor_type in ["padded", "masked"]:
         if mask_assignment is None or not isinstance(mask_assignment, (list, tuple)):
-            raise ValueError()
+            raise ValueError("Mask assignment information is required or invalid.")
 
         reduced_mask = [x for x in mask_assignment if x is not None]
         if len(reduced_mask) == 0:

kgcnn/ops/scatter.py

@@ -13,6 +13,16 @@ def compute_output_spec(self, indices, values, shape):
 
 
 def scatter_reduce_max(indices, values, shape):
+    r"""Scatter values at indices into new tensor of shape.
+
+    Args:
+        indices (Tensor): 1D Indices of shape `(M, )` .
+        values (Tensor): Vales of shape `(M, ...)` .
+        shape (tuple): Target shape.
+
+    Returns:
+        Tensor: Scattered values of `shape` .
+    """
     if any_symbolic_tensors((indices, values, shape)):
         return _ScatterMax().symbolic_call(indices, values, shape)
     return kgcnn_backend.scatter_reduce_max(indices, values, shape)
@@ -27,6 +37,16 @@ def compute_output_spec(self, indices, values, shape):
 
 
 def scatter_reduce_min(indices, values, shape):
+    r"""Scatter values at indices into new tensor of shape.
+
+    Args:
+        indices (Tensor): 1D Indices of shape `(M, )` .
+        values (Tensor): Vales of shape `(M, ...)` .
+        shape (tuple): Target shape.
+
+    Returns:
+        Tensor: Scattered values of `shape` .
+    """
     if any_symbolic_tensors((indices, values, shape)):
         return _ScatterMin().symbolic_call(indices, values, shape)
     return kgcnn_backend.scatter_reduce_min(indices, values, shape)
@@ -41,6 +61,16 @@ def compute_output_spec(self, indices, values, shape):
 
 
 def scatter_reduce_mean(indices, values, shape):
+    r"""Scatter values at indices into new tensor of shape.
+
+    Args:
+        indices (Tensor): 1D Indices of shape `(M, )` .
+        values (Tensor): Vales of shape `(M, ...)` .
+        shape (tuple): Target shape.
+
+    Returns:
+        Tensor: Scattered values of `shape` .
+    """
     if any_symbolic_tensors((indices, values, shape)):
         return _ScatterMean().symbolic_call(indices, values, shape)
     return kgcnn_backend.scatter_reduce_mean(indices, values, shape)
@@ -55,6 +85,16 @@ def compute_output_spec(self, indices, values, shape):
 
 
 def scatter_reduce_sum(indices, values, shape):
+    r"""Scatter values at indices into new tensor of shape.
+
+    Args:
+        indices (Tensor): 1D Indices of shape `(M, )` .
+        values (Tensor): Vales of shape `(M, ...)` .
+        shape (tuple): Target shape.
+
+    Returns:
+        Tensor: Scattered values of `shape` .
+    """
     if any_symbolic_tensors((indices, values, shape)):
         return _ScatterSum().symbolic_call(indices, values, shape)
     return kgcnn_backend.scatter_reduce_sum(indices, values, shape)
@@ -74,6 +114,16 @@ def compute_output_spec(self, indices, values, shape):
 
 
 def scatter_reduce_softmax(indices, values, shape, normalize: bool = False):
+    r"""Scatter values at indices to normalize values via softmax.
+
+    Args:
+        indices (Tensor): 1D Indices of shape `(M, )` .
+        values (Tensor): Vales of shape `(M, ...)` .
+        shape (tuple): Target shape of scattered tensor.
+
+    Returns:
+        Tensor: Values with softmax computed by grouping at indices.
+    """
     if any_symbolic_tensors((indices, values, shape)):
         return _ScatterSoftmax(normalize=normalize).symbolic_call(indices, values, shape)
     return kgcnn_backend.scatter_reduce_softmax(indices, values, shape, normalize=normalize)

training/hyper/hyper_qm9_orbitals.py

@@ -85,7 +85,9 @@
                     {"shape": [None, 3], "name": "node_coordinates", "dtype": "float32", "ragged": True},
                     {"shape": [None, 2], "name": "range_indices", "dtype": "int64", "ragged": True}
                 ],
-                "input_embedding": {"node": {"input_dim": 95, "output_dim": 128}},
+                "input_tensor_type": "ragged",
+                "input_embedding": None,
+                "input_node_embedding": {"input_dim": 95, "output_dim": 128},
                 "bessel_basis": {"num_radial": 20, "cutoff": 5.0, "envelope_exponent": 5},
                 "pooling_args": {"pooling_method": "sum"}, "conv_args": {"units": 128, "cutoff": None},
                 "update_args": {"units": 128}, "depth": 3, "verbose": 10,

training/results/ESOLDataset/CMPNN/CMPNN_ESOLDataset_score.yaml

@@ -0,0 +1,137 @@
+OS: posix_linux
+backend: tensorflow
+cuda_available: 'True'
+data_unit: mol/L
+date_time: '2023-12-20 23:54:16'
+device_id: '[LogicalDevice(name=''/device:CPU:0'', device_type=''CPU''), LogicalDevice(name=''/device:GPU:0'',
+  device_type=''GPU'')]'
+device_memory: '[]'
+device_name: '[{}, {''compute_capability'': (7, 0), ''device_name'': ''Tesla V100-SXM2-32GB''}]'
+epochs:
+- 600
+- 600
+- 600
+- 600
+- 600
+execute_folds: null
+kgcnn_version: 4.0.0
+loss:
+- 0.00023183324083220214
+- 0.00017752200074028224
+- 0.00018914483371190727
+- 0.0001874043809948489
+- 1.900034476420842e-05
+max_loss:
+- 1.1973497867584229
+- 1.0974987745285034
+- 1.4608932733535767
+- 1.2774747610092163
+- 1.125367522239685
+max_scaled_mean_absolute_error:
+- 1.715080976486206
+- 1.685046911239624
+- 1.8969279527664185
+- 1.7868362665176392
+- 1.7247272729873657
+max_scaled_root_mean_squared_error:
+- 2.3039710521698
+- 2.257887125015259
+- 2.608121633529663
+- 2.406620979309082
+- 2.245151996612549
+max_val_loss:
+- 0.7626262307167053
+- 1.6030727624893188
+- 0.7904647588729858
+- 0.939492404460907
+- 0.7619210481643677
+max_val_scaled_mean_absolute_error:
+- 1.3947796821594238
+- 2.2608180046081543
+- 1.5816482305526733
+- 1.5697236061096191
+- 1.4567921161651611
+max_val_scaled_root_mean_squared_error:
+- 1.8257416486740112
+- 2.6560630798339844
+- 1.8949168920516968
+- 2.023801803588867
+- 1.8428680896759033
+min_loss:
+- 0.00023183324083220214
+- 0.00017752200074028224
+- 0.00018914483371190727
+- 0.00018653765437193215
+- 1.8690105207497254e-05
+min_scaled_mean_absolute_error:
+- 0.005625245161354542
+- 0.005176743026822805
+- 0.005119773093611002
+- 0.0039019521791487932
+- 0.0043538580648601055
+min_scaled_root_mean_squared_error:
+- 0.032370418310165405
+- 0.028836989775300026
+- 0.029924051836133003
+- 0.029095962643623352
+- 0.009313425980508327
+min_val_loss:
+- 0.09925275295972824
+- 0.0914703980088234
+- 0.09436161816120148
+- 0.0922454372048378
+- 0.08730437606573105
+min_val_scaled_mean_absolute_error:
+- 0.4868481457233429
+- 0.46619078516960144
+- 0.4638049304485321
+- 0.4697567820549011
+- 0.4539531469345093
+min_val_scaled_root_mean_squared_error:
+- 0.6407178640365601
+- 0.6302186250686646
+- 0.6600759625434875
+- 0.6400931477546692
+- 0.6043533086776733
+model_class: make_model
+model_name: CMPNN
+model_version: '2023-10-30'
+multi_target_indices: null
+number_histories: 5
+scaled_mean_absolute_error:
+- 0.005625245161354542
+- 0.005188747774809599
+- 0.005194343626499176
+- 0.004212634172290564
+- 0.004786244127899408
+scaled_root_mean_squared_error:
+- 0.032370418310165405
+- 0.028836989775300026
+- 0.029924051836133003
+- 0.029157962650060654
+- 0.009391490370035172
+seed: 42
+time_list:
+- '0:12:35.362480'
+- '0:12:35.969232'
+- '0:12:42.460222'
+- '0:12:35.954069'
+- '0:12:26.994579'
+val_loss:
+- 0.13186116516590118
+- 0.1403244137763977
+- 0.11817561089992523
+- 0.12358944118022919
+- 0.1467534303665161
+val_scaled_mean_absolute_error:
+- 0.5213338732719421
+- 0.5559774041175842
+- 0.5131708979606628
+- 0.5310410857200623
+- 0.5166580080986023
+val_scaled_root_mean_squared_error:
+- 0.7463422417640686
+- 0.7871690988540649
+- 0.7463598251342773
+- 0.7365252375602722
+- 0.7358969449996948

training/results/ESOLDataset/CMPNN/CMPNN_hyper.json

@@ -0,0 +1 @@
+{"model": {"class_name": "make_model", "module_name": "kgcnn.literature.CMPNN", "config": {"name": "CMPNN", "inputs": [{"shape": [null, 41], "name": "node_attributes", "dtype": "float32"}, {"shape": [null, 11], "name": "edge_attributes", "dtype": "float32"}, {"shape": [null, 2], "name": "edge_indices", "dtype": "int64"}, {"shape": [null, 1], "name": "edge_indices_reverse", "dtype": "int64"}, {"shape": [], "name": "total_nodes", "dtype": "int64"}, {"shape": [], "name": "total_edges", "dtype": "int64"}, {"shape": [], "name": "total_reverse", "dtype": "int64"}], "input_tensor_type": "padded", "input_node_embedding": {"input_dim": 95, "output_dim": 64}, "input_edge_embedding": {"input_dim": 5, "output_dim": 64}, "node_initialize": {"units": 300, "activation": "relu"}, "edge_initialize": {"units": 300, "activation": "relu"}, "edge_dense": {"units": 300, "activation": "linear"}, "node_dense": {"units": 300, "activation": "linear"}, "edge_activation": {"activation": "relu"}, "verbose": 10, "depth": 5, "dropout": null, "use_final_gru": true, "pooling_gru": {"units": 300}, "pooling_kwargs": {"pooling_method": "sum"}, "output_embedding": "graph", "output_mlp": {"use_bias": [true, false], "units": [300, 1], "activation": ["relu", "linear"]}}}, "training": {"fit": {"batch_size": 50, "epochs": 600, "validation_freq": 1, "verbose": 2, "callbacks": []}, "compile": {"optimizer": {"class_name": "Adam", "config": {"learning_rate": {"class_name": "ExponentialDecay", "module": "keras.optimizers.schedules", "config": {"initial_learning_rate": 0.001, "decay_steps": 1600, "decay_rate": 0.5, "staircase": false}}}}, "loss": "mean_squared_error"}, "cross_validation": {"class_name": "KFold", "config": {"n_splits": 5, "random_state": 42, "shuffle": true}}, "scaler": {"class_name": "StandardLabelScaler", "config": {"with_std": true, "with_mean": true, "copy": true}}}, "data": {"data_unit": "mol/L"}, "info": {"postfix": "", "postfix_file": "", "kgcnn_version": "4.0.0"}, "dataset": {"class_name": "ESOLDataset", "module_name": "kgcnn.data.datasets.ESOLDataset", "config": {}, "methods": [{"set_attributes": {}}, {"map_list": {"method": "set_edge_indices_reverse"}}, {"map_list": {"method": "count_nodes_and_edges"}}, {"map_list": {"method": "count_nodes_and_edges", "total_edges": "total_reverse"}}]}}

training/results/QM7Dataset/DimeNetPP/DimeNetPP_QM7Dataset_score.yaml

@@ -0,0 +1,137 @@
+OS: posix_linux
+backend: tensorflow
+cuda_available: 'True'
+data_unit: kcal/mol
+date_time: '2023-12-21 04:45:11'
+device_id: '[LogicalDevice(name=''/device:CPU:0'', device_type=''CPU''), LogicalDevice(name=''/device:GPU:0'',
+  device_type=''GPU'')]'
+device_memory: '[]'
+device_name: '[{}, {''compute_capability'': (7, 0), ''device_name'': ''Tesla V100-SXM2-32GB''}]'
+epochs:
+- 872
+- 872
+- 872
+- 872
+- 872
+execute_folds: null
+kgcnn_version: 4.0.0
+loss:
+- 0.0005192547105252743
+- 0.0005859421798959374
+- 0.0005881146644242108
+- 0.00038828517426736653
+- 0.00042653322452679276
+max_loss:
+- 0.39550185203552246
+- 0.3835352957248688
+- 0.39535853266716003
+- 0.40081343054771423
+- 0.3717119097709656
+max_scaled_mean_absolute_error:
+- 88.4721450805664
+- 85.88572692871094
+- 88.60206604003906
+- 89.8046646118164
+- 83.19475555419922
+max_scaled_root_mean_squared_error:
+- 131.0218505859375
+- 130.70944213867188
+- 129.65948486328125
+- 136.13157653808594
+- 121.79318237304688
+max_val_loss:
+- 0.07793247699737549
+- 0.05247681587934494
+- 0.06055280938744545
+- 0.08880073577165604
+- 0.0839301124215126
+max_val_scaled_mean_absolute_error:
+- 17.380496978759766
+- 11.777695655822754
+- 13.62238883972168
+- 19.90665054321289
+- 18.728012084960938
+max_val_scaled_root_mean_squared_error:
+- 23.846885681152344
+- 15.797113418579102
+- 18.345048904418945
+- 26.381744384765625
+- 23.828720092773438
+min_loss:
+- 0.0005192547105252743
+- 0.0005859421798959374
+- 0.0005805905675515532
+- 0.00038581088301725686
+- 0.0004265273455530405
+min_scaled_mean_absolute_error:
+- 0.11637404561042786
+- 0.13132883608341217
+- 0.1302138864994049
+- 0.08653981983661652
+- 0.09565576165914536
+min_scaled_root_mean_squared_error:
+- 0.20463432371616364
+- 0.20020626485347748
+- 0.203572615981102
+- 0.13961230218410492
+- 0.16126485168933868
+min_val_loss:
+- 0.017145873978734016
+- 0.015065383166074753
+- 0.01469708327203989
+- 0.015139556489884853
+- 0.015097062103450298
+min_val_scaled_mean_absolute_error:
+- 3.8343989849090576
+- 3.376465082168579
+- 3.295900583267212
+- 3.3965675830841064
+- 3.384349822998047
+min_val_scaled_root_mean_squared_error:
+- 10.640360832214355
+- 5.686461925506592
+- 7.498864650726318
+- 7.203327178955078
+- 6.20181131362915
+model_class: make_model
+model_name: DimeNetPP
+model_version: '2023-12-04'
+multi_target_indices: null
+number_histories: 5
+scaled_mean_absolute_error:
+- 0.11637404561042786
+- 0.13132883608341217
+- 0.13185767829418182
+- 0.08708549290895462
+- 0.09565749019384384
+scaled_root_mean_squared_error:
+- 0.20463432371616364
+- 0.20020626485347748
+- 0.203572615981102
+- 0.13961230218410492
+- 0.16126485168933868
+seed: 42
+time_list:
+- '6:59:19.071354'
+- '7:00:28.823901'
+- '6:50:30.302822'
+- '6:46:21.878251'
+- '6:55:38.207034'
+val_loss:
+- 0.01725739613175392
+- 0.015085577964782715
+- 0.014698205515742302
+- 0.015139556489884853
+- 0.015109173022210598
+val_scaled_mean_absolute_error:
+- 3.8581318855285645
+- 3.381181478500366
+- 3.2967283725738525
+- 3.3965675830841064
+- 3.3870768547058105
+val_scaled_root_mean_squared_error:
+- 10.91750431060791
+- 5.7236785888671875
+- 7.511663436889648
+- 7.305181503295898
+- 6.205348014831543

training/results/QM7Dataset/DimeNetPP/DimeNetPP_hyper.json

@@ -0,0 +1 @@
+{"model": {"class_name": "make_model", "module_name": "kgcnn.literature.DimeNetPP", "config": {"name": "DimeNetPP", "inputs": [{"shape": [null], "name": "node_number", "dtype": "int64", "ragged": true}, {"shape": [null, 3], "name": "node_coordinates", "dtype": "float32", "ragged": true}, {"shape": [null, 2], "name": "range_indices", "dtype": "int64", "ragged": true}, {"shape": [null, 2], "name": "angle_indices", "dtype": "int64", "ragged": true}], "input_tensor_type": "ragged", "input_embedding": null, "input_node_embedding": {"input_dim": 95, "output_dim": 128, "embeddings_initializer": {"class_name": "RandomUniform", "config": {"minval": -1.7320508075688772, "maxval": 1.7320508075688772}}}, "emb_size": 128, "out_emb_size": 256, "int_emb_size": 64, "basis_emb_size": 8, "num_blocks": 4, "num_spherical": 7, "num_radial": 6, "cutoff": 5.0, "envelope_exponent": 5, "num_before_skip": 1, "num_after_skip": 2, "num_dense_output": 3, "num_targets": 1, "extensive": false, "output_init": "zeros", "activation": "swish", "verbose": 10, "output_embedding": "graph", "use_output_mlp": false, "output_mlp": {}}}, "training": {"fit": {"batch_size": 10, "epochs": 872, "validation_freq": 10, "verbose": 2, "callbacks": []}, "compile": {"optimizer": {"class_name": "Adam", "config": {"learning_rate": {"class_name": "kgcnn>LinearWarmupExponentialDecay", "config": {"learning_rate": 0.001, "warmup_steps": 150.0, "decay_steps": 200000.0, "decay_rate": 0.01}}, "use_ema": true, "amsgrad": true}}, "loss": "mean_absolute_error"}, "scaler": {"class_name": "QMGraphLabelScaler", "config": {"atomic_number": "node_number", "scaler": [{"class_name": "StandardLabelScaler", "config": {"with_std": true, "with_mean": true, "copy": true}}]}}, "multi_target_indices": null}, "data": {"data_unit": "kcal/mol"}, "info": {"postfix": "", "postfix_file": "", "kgcnn_version": "4.0.0"}, "dataset": {"class_name": "QM7Dataset", "module_name": "kgcnn.data.datasets.QM7Dataset", "config": {}, "methods": [{"map_list": {"method": "set_range", "max_distance": 5, "max_neighbours": 1000}}, {"map_list": {"method": "set_angle"}}]}}