feat: add metrics.svg

extensions/grapycal_builtin/grapycal_builtin/interaction/imagePasteNode.py

@@ -60,11 +60,10 @@ def render_from_fig(
     return buf
 
 @contextmanager
-def open_fig() -> Generator[Tuple[Figure, Axes], None, None]:
+def open_fig(equal=False) -> Generator[Tuple[Figure, Axes], None, None]:
     fig = plt.figure()
     ax = fig.gca()
     ax.set_facecolor("black")
-    ax.set_aspect("equal")
     try:
         yield fig, ax
     finally:
@@ -153,7 +152,7 @@ def build_node(self):
         self.cmap = self.add_attribute(
             "cmap",
             StringTopic,
-            "gray",
+            "viridis",
             editor_type="options",
             options=[
                 "gray",
@@ -412,7 +411,7 @@ def preprocess_data(self, data):
         return data
 
     def update_image(self, data):
-        with open_fig() as (fig, ax):
+        with open_fig(equal=True) as (fig, ax):
             for d in data:
                 if len(d.shape) == 3:
                     for slice in d:

extensions/grapycal_builtin/grapycal_builtin/interaction/printNode.py

@@ -1,9 +1,18 @@
+import io
+import pprint
+
 from grapycal.sobjects.controls import TextControl
 from grapycal.sobjects.edge import Edge
-from grapycal.sobjects.node import RESTORE_FROM, Node
+from grapycal.sobjects.node import Node
 from grapycal.sobjects.port import InputPort
 
 
+def get_pprint_str(data):
+    output = io.StringIO(newline="")
+    pprint.pprint(data, stream=output)
+    return output.getvalue()
+
+
 class PrintNode(Node):
     '''
     Display the data received from the input edge.
@@ -24,13 +33,13 @@ def build_node(self):
     def edge_activated(self, edge, port):
         self.flash_running_indicator()
         data = edge.get()
-        self.text_control.text.set(str(data))
+        self.text_control.text.set(get_pprint_str(data))
 
     def input_edge_added(self, edge: Edge, port: InputPort):
         if edge.is_data_ready():
             self.flash_running_indicator()
             data = edge.get()
-            self.text_control.text.set(str(data))
+            self.text_control.text.set(get_pprint_str(data))
 
     def input_edge_removed(self, edge: Edge, port: InputPort):
         self.text_control.text.set('')

extensions/grapycal_torch/dataset.py

@@ -33,9 +33,10 @@ def task(self):
 
         ds = []
         for i in range(size):
-            ds.append(raw_ds[i])
+            pair = raw_ds[i]
+            ds.append({'image': pair[0], 'label': pair[1]})
 
         if self.include_labels.get() == 'False':
-            ds = [x[0] for x in ds]
+            ds = [x['image'] for x in ds]
 
         self.out.push(ds)

extensions/grapycal_torch/loss.py

@@ -11,6 +11,7 @@ class BCEWithLogitsLossNode(SimpleModuleNode):
     def build_node(self):
         super().build_node()
         self.label.set('BCEWithLogitsLoss')
+        self.icon_path.set('metrics')
 
     def create_module(self) -> nn.Module:
         return nn.BCEWithLogitsLoss()
@@ -29,6 +30,7 @@ class CrossEntropyLossNode(SimpleModuleNode):
     def build_node(self):
         super().build_node()
         self.label.set('CrossEntropyLoss')
+        self.icon_path.set('metrics')
 
     def create_module(self) -> nn.Module:
         return nn.CrossEntropyLoss()
@@ -47,6 +49,7 @@ class MSELossNode(SimpleModuleNode):
     def build_node(self):
         super().build_node()
         self.label.set('MSELoss')
+        self.icon_path.set('metrics')
 
     def create_module(self) -> nn.Module:
         return nn.MSELoss()

extensions/grapycal_torch/metrics.py

@@ -12,6 +12,7 @@ def build_node(self):
         super().build_node()
         self.label.set('Accuracy')
         self.shape.set('simple')
+        self.icon_path.set('metrics')
 
     def calculate(self, prediction, target) -> Any:
         #TODO: if target is one-hot encoded, convert it to class labels

extensions/grapycal_torch/networkDef.py

@@ -23,6 +23,7 @@ class NetworkCallNode(Node):
     def build_node(self,name:str="Network"):
         self.label.set('')
         self.shape.set('normal')
+        self.icon_path.set('nn')
         self.network_name = self.add_attribute('network name',StringTopic,editor_type='text',init_value=name)
         self.network_name.add_validator(lambda x,_: x != '') # empty name may confuse users
         self.mode_control = self.add_option_control(name='mode',options=['train','eval'],value='train',label='Mode')
@@ -142,6 +143,7 @@ def build_node(self,name:str="Network",inputs:List[str]|None=None):
             inputs = ['x']
 
         self.shape.set('normal')
+        self.icon_path.set('nn')
 
         # setup attributes
         # The self.outs attribute is actually "inputs" of the network, but it was mistakenly named "outs" and I didn't want to change it to avoid breaking backwards compatibility
@@ -229,6 +231,7 @@ def build_node(self,name:str="Network",outputs:List[str]|None=None):
         if outputs is None:
             outputs = ['y']
         self.shape.set('normal')
+        self.icon_path.set('nn')
 
         # setup attributes
         self.ins = self.add_attribute('ins',ListTopic,editor_type='list',init_value=outputs,display_name='outputs')

extensions/grapycal_torch/tensor_operations.py

@@ -1,11 +1,12 @@
-from ast import pattern
 from typing import Any
-from grapycal import FunctionNode, IntTopic, StringTopic
-from grapycal.extension.utils import NodeInfo
-from grapycal.sobjects.controls.textControl import TextControl
+
+import einops
 import torch
 import torch.nn.functional as F
-import einops
+from grapycal import FunctionNode, IntTopic
+from grapycal.extension.utils import NodeInfo
+from grapycal.sobjects.controls.textControl import TextControl
+
 
 class CatNode(FunctionNode):
     category = 'torch/operations'