add repr for genome and gene;

add ipynb test for testing whether add node or add conn will not change the output for the network.

tensorneat/algorithm/neat/gene/base.py

@@ -40,3 +40,6 @@ class BaseGene(StatefulBaseClass):
     @property
     def length(self):
         return len(self.fixed_attrs) + len(self.custom_attrs)
+
+    def repr(self, state, gene, precision=2):
+        raise NotImplementedError

tensorneat/algorithm/neat/gene/conn/base.py

@@ -22,3 +22,12 @@ class BaseConnGene(BaseGene):
             jax.vmap(self.forward, in_axes=(None, None, 0))(state, attrs, batch_inputs),
             attrs,
         )
+
+    def repr(self, state, conn, precision=2, idx_width=3, func_width=8):
+        in_idx, out_idx = conn[:2]
+        in_idx = int(in_idx)
+        out_idx = int(out_idx)
+
+        return "{}(in: {:<{idx_width}}, out: {:<{idx_width}})".format(
+            self.__class__.__name__, in_idx, out_idx, idx_width=idx_width
+        )

tensorneat/algorithm/neat/gene/conn/default.py

@@ -60,3 +60,19 @@ class DefaultConnGene(BaseConnGene):
     def forward(self, state, attrs, inputs):
         weight = attrs[0]
         return inputs * weight
+
+    def repr(self, state, conn, precision=2, idx_width=3, func_width=8):
+        in_idx, out_idx, weight = conn
+
+        in_idx = int(in_idx)
+        out_idx = int(out_idx)
+        weight = round(float(weight), precision)
+
+        return "{}(in: {:<{idx_width}}, out: {:<{idx_width}}, weight: {:<{float_width}})".format(
+            self.__class__.__name__,
+            in_idx,
+            out_idx,
+            weight,
+            idx_width=idx_width,
+            float_width=precision + 3,
+        )

tensorneat/algorithm/neat/gene/node/base.py

@@ -39,3 +39,11 @@ class BaseNodeGene(BaseGene):
             ),
             attrs,
         )
+
+    def repr(self, state, node, precision=2, idx_width=3, func_width=8):
+        idx = node[0]
+
+        idx = int(idx)
+        return "{}(idx={:<{idx_width}})".format(
+            self.__class__.__name__, idx, idx_width=idx_width
+        )

tensorneat/algorithm/neat/gene/node/default.py

@@ -122,3 +122,28 @@ class DefaultNodeGene(BaseNodeGene):
         )
 
         return z
+
+    def repr(self, state, node, precision=2, idx_width=3, func_width=8):
+        idx, bias, res, agg, act = node
+
+        idx = int(idx)
+        bias = round(float(bias), precision)
+        res = round(float(res), precision)
+        agg = int(agg)
+        act = int(act)
+
+        if act == -1:
+            act_func = Act.identity
+        else:
+            act_func = self.activation_options[act]
+        return "{}(idx={:<{idx_width}}, bias={:<{float_width}}, response={:<{float_width}}, aggregation={:<{func_width}}, activation={:<{func_width}})".format(
+            self.__class__.__name__,
+            idx,
+            bias,
+            res,
+            self.aggregation_options[agg].__name__,
+            act_func.__name__,
+            idx_width=idx_width,
+            float_width=precision + 3,
+            func_width=func_width
+        )

tensorneat/algorithm/neat/gene/node/default_without_response.py

@@ -98,3 +98,26 @@ class NodeGeneWithoutResponse(BaseNodeGene):
         )
 
         return z
+
+    def repr(self, state, node, precision=2, idx_width=3, func_width=8):
+        idx, bias, agg, act = node
+
+        idx = int(idx)
+        bias = round(float(bias), precision)
+        agg = int(agg)
+        act = int(act)
+
+        if act == -1:
+            act_func = Act.identity
+        else:
+            act_func = self.activation_options[act]
+        return "{}(idx={:<{idx_width}}, bias={:<{float_width}}, aggregation={:<{func_width}}, activation={:<{func_width}})".format(
+            self.__class__.__name__,
+            idx,
+            bias,
+            self.aggregation_options[agg].__name__,
+            act_func.__name__,
+            idx_width=idx_width,
+            float_width=precision + 3,
+            func_width=func_width,
+        )

tensorneat/algorithm/neat/gene/node/kan_node.py

@@ -25,3 +25,8 @@ class KANNode(BaseNodeGene):
 
     def forward(self, state, attrs, inputs, is_output_node=False):
         return Agg.sum(inputs)
+
+    def repr(self, state, node, precision=2):
+        idx = node[0]
+        idx = int(idx)
+        return "{}(idx: {})".format(self.__class__.__name__, idx)