continue keras core integration

aimat-lab · Sep 12, 2023 · 50e856a · 50e856a
1 parent a6fee65
commit 50e856a
Show file tree

Hide file tree

Showing 3 changed files with 8 additions and 10 deletions.
diff --git a/kgcnn/layers_core/casting.py b/kgcnn/layers_core/casting.py
@@ -34,12 +34,12 @@ def __init__(self, reverse_indices: bool = True, dtype_batch: str = "int64", dty
     def build(self, input_shape):
         self.built = True
 
-    def compute_output_shape(self, input_shape):
-        out_shape = [tuple([None] + list(input_shape[0][2:])), tuple(list(reversed(input_shape[1][2:])) + [None]),
-            (None, ), (None, ), (None, ), (None, )]
-        if len(input_shape) == 5:
-            out_shape = out_shape + [tuple([None] + list(input_shape[4][2:]))]
-        return out_shape
+    # def compute_output_shape(self, input_shape):
+    #     out_shape = [tuple([None] + list(input_shape[0][2:])), tuple(list(reversed(input_shape[1][2:])) + [None]),
+    #         (None, ), (None, ), (None, ), (None, )]
+    #     if len(input_shape) == 5:
+    #         out_shape = out_shape + [tuple([None] + list(input_shape[4][2:]))]
+    #     return out_shape
 
     def call(self, inputs: list, **kwargs):
         """Changes node and edge indices into a Pytorch Geometric (PyG) compatible tensor format.
@@ -217,10 +217,9 @@ def call(self, inputs: list, **kwargs):
             node_mask_flatten = ops.reshape(node_mask, [-1])
             nodes_flatten = pad_left(nodes_flatten)
             node_len = cat_one(node_len)
+            node_mask_flatten = pad_left(node_mask_flatten)
             nodes_id = repeat_static_length(ops.arange(ops.shape(node_len)[0], dtype=self.dtype_batch), node_len,
                                             total_repeat_length=ops.shape(nodes_flatten)[0])
-
-        if self.padded_disjoint:
             nodes_id = ops.where(node_mask_flatten, nodes_id, ops.convert_to_tensor(0, dtype=self.dtype_batch))
 
         return [nodes_flatten, nodes_id, node_len]

diff --git a/kgcnn/literature_core/Schnet/_model.py b/kgcnn/literature_core/Schnet/_model.py
@@ -138,7 +138,6 @@ def make_model(inputs: list = None,
     if expand_distance:
         ed = GaussBasisLayer(**gauss_args)(ed)
 
-    print(n, ed, disjoint_indices)
     # Model
     n = Dense(interaction_args["units"], activation='linear')(n)
     for i in range(0, depth):

diff --git a/training_core/hyper/hyper_esol.py b/training_core/hyper/hyper_esol.py
@@ -16,7 +16,7 @@
                 "input_node_embedding": {"input_dim": 95, "output_dim": 64},
                 "input_edge_embedding": {"input_dim": 25, "output_dim": 1},
                 "gcn_args": {"units": 140, "use_bias": True, "activation": "relu"},
-                "depth": 0, "verbose": 10,
+                "depth": 5, "verbose": 10,
                 "output_embedding": "graph",
                 "output_mlp": {"use_bias": [True, True, False], "units": [140, 70, 1],
                                "activation": ["relu", "relu", "linear"]},