DocsNode Tutorial

Node Editor Tutorial

Build a simple neural network using the visual Node Editor. No coding required!

Goal

Create a 2-layer Dense network for MNIST classification:

Input (784) → Dense (128, ReLU) → Dense (10, Softmax) → Output

Step 1: Open the Node Editor

Go to View → Node Editor (or press Ctrl+1)

Action	Control
Pan	Middle-mouse drag or Space+drag
Zoom	Mouse wheel
Select	Left-click

Step 2: Add an Input Node

Right-click on the canvas
Navigate to Data → DatasetInput
Configure in Properties: Input Shape: 784, Batch Size: 32

+------------------+
|  DatasetInput    |
+------------------+
| Dataset: [None]  |
|   Batch: 32      |
+------------------+
|        [output]→ |
+------------------+

Step 3-5: Add Dense Layers

Right-click → Layers → Dense
Configure first Dense: Units: 128, Activation: ReLU
Connect DatasetInput output to Dense input (drag between pins)
Add second Dense: Units: 10, Activation: Softmax
Connect the layers together

Step 6: Add Output Node

Right-click → Output → ModelOutput
Connect the second Dense to the Output

+-------------+    +-------------+    +-------------+    +-------------+
| DatasetInput|-→--| Dense(128)  |-→--| Dense(10)   |-→--| ModelOutput |
+-------------+    | ReLU        |    | Softmax     |    +-------------+
                   +-------------+    +-------------+

Step 7: Validate the Graph

Go to Nodes → Validate Graph (or Ctrl+Shift+V)

A green checkmark appears if valid. Common errors:

"Missing input connection" → Ensure all inputs are connected
"Shape mismatch" → Check layer dimensions

Step 8: Generate Code

PyTorch Output

import torch.nn as nn

class GeneratedModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.dense_1 = nn.Linear(784, 128)
        self.relu_1 = nn.ReLU()
        self.dense_2 = nn.Linear(128, 10)
        self.softmax = nn.Softmax(dim=-1)

    def forward(self, x):
        x = self.dense_1(x)
        x = self.relu_1(x)
        x = self.dense_2(x)
        x = self.softmax(x)
        return x

PyCyxWiz Output

import pycyxwiz as cx

dense_1 = cx.Dense(784, 128)
relu_1 = cx.ReLU()
dense_2 = cx.Dense(128, 10)
softmax = cx.Softmax()

def forward(x):
    x = dense_1.forward(x)
    x = relu_1.forward(x)
    x = dense_2.forward(x)
    x = softmax.forward(x)
    return x

Common Node Types

Category	Nodes
Data	DatasetInput, DataLoader
Layers	Dense, Conv2D, LSTM, Flatten
Activations	ReLU, Sigmoid, Tanh, Softmax, GELU
Normalization	BatchNorm, LayerNorm
Output	ModelOutput, LossOutput

What You Learned

Adding nodes from the context menu
Connecting nodes with links
Configuring node properties
Validating the graph
Generating Python code