Add some dummy gradient examples with pytorch

grad.py

+import torch
+
+def print_gradient_formula(operation, grad_formula):
+    print(f"Gradient formula for {operation}: {grad_formula}\n")
+
+def compute_gradients():
+
+
+    # Example 1: Gradient of Element-wise Multiplication
+    print("Example 1: Element-wise Multiplication")
+    beta = torch.tensor([3.0, 4.0, 5.0], requires_grad=False)
+    X1 = torch.tensor([2.0, 3.0, 4.0], requires_grad=True)
+    print(f"beta  : {beta}")
+    print(f"X1    : {X1}")
+    Y1 = beta * X1
+    Y1.sum().backward()
+    print(f"Gradient of Y1 with respect to X1: {X1.grad}")
+    print_gradient_formula("β * X", "∂Y/∂X = β")
+
+    # Reset gradients for the next operation
+    X1.grad.zero_()
+
+    # Example 2: Gradient of Element-wise Division
+    print("Example 2: Element-wise Division")
+    X2 = torch.tensor([2.0, 4.0, 8.0], requires_grad=True)
+    Y2 = torch.tensor([1.0, 2.0, 4.0], requires_grad=False)
+    print(f"X2 : {X2}")
+    print(f"Y2 : {Y2}")
+    Z2 = X2 / Y2
+    Z2.sum().backward()
+    print(f"Gradient of Z2 with respect to X2: {X2.grad}")
+    print_gradient_formula("X / Y", "∂Z/∂X = 1 / Y")
+
+    # Reset gradients for the next operation
+    X2.grad.zero_()
+
+    # Example 3: Gradient of Element-wise Power
+    print("Example 3: Element-wise Power")
+    X3 = torch.tensor([1.0, 2.0, 3.0], requires_grad=True)
+    exponent = torch.tensor([2.0, 3.0, 4.0], requires_grad=False)
+    print(f"X3  : {X3}")
+    print(f"pow : {exponent}")
+    Z3 = X3.pow(exponent)
+    Z3.sum().backward()
+    print(f"Gradient of Z3 with respect to X3: {X3.grad}")
+    print_gradient_formula("X^exponent", "∂Z/∂X = exponent * X^(exponent - 1)")
+
+    # Reset gradients for the next operation
+    X3.grad.zero_()
+
+    # Example 4: Gradient of Element-wise Square Root
+    print("Example 4: Element-wise Square Root")
+    X4 = torch.tensor([4.0, 9.0, 16.0], requires_grad=True)
+    print(f"X4 : {X4}")
+    Z4 = X4.sqrt()
+    Z4.sum().backward()
+    print(f"Gradient of Z4 with respect to X4: {X4.grad}")
+    print_gradient_formula("sqrt(X)", "∂Z/∂X = 1 / (2 * sqrt(X))")
+
+    # Reset gradients for the next operation
+    X4.grad.zero_()
+
+    # Example 5: Combining Multiple Operations
+    print("Example 5: Combining Multiple Operations")
+    X5 = torch.tensor([4.0, 9.0, 16.0], requires_grad=True)
+    Y5 = torch.tensor([2.0, 3.0, 4.0], requires_grad=False)
+    print(f"X5 : {X5}")
+    print(f"Y5 : {Y5}")
+    Z5 = (X5.pow(2) / Y5).sqrt()
+    Z5.sum().backward()
+    print(f"Gradient of Z5 with respect to X5: {X5.grad}")
+    print_gradient_formula("sqrt((X^2) / Y)", "∂Z/∂X = (1 / (2 * sqrt((X^2) / Y))) * (2 * X / Y)")
+
+if __name__ == "__main__":
+    compute_gradients()
+
+

tensor.py

+# Testing Tensor type in aidge
+
+import numpy as np
+import aidge_core
+
+array = np.random.rand(16,28*28)
+tensor = aidge_core.Tensor(array)
+
+
+capacity = tensor.capacity()
+print(f"Capacity   {capacity}")
+print(f"Dims       {tensor.dims()}")
+print(f"dtype      {tensor.dtype()}")
+print(f"Coords     {tensor.get_coord(0)}")
+print(f"")