setup Anaconda, Spyder, did first couple tutorials, pytorch working i…

…n environment

gentle_intro_to_autograd.py

@@ -0,0 +1,22 @@
+import torch, torchvision
+
+# Load a pretrained resnet18 model
+model = torchvision.models.resnet18(pretrained=True)
+
+# Random tensor data to represent an image with three channels and height/width of 64
+data = torch.rand(1,3,64,64)
+# Image's label init'd to some random values
+labels = torch.rand(1,1000)
+
+# Run the input data through the model to make a prediction
+prediction = model(data) # forward pass
+
+# Calculate the error and backpropagate through the network
+loss = (prediction - labels).sum()
+loss.backward() # backward pass
+
+# Load an optimizer with a learning rate of 0.01 and momentum of 0.9
+optim = torch.optim.SGD(model.parameters(), lr=1e-2, momentum=0.9)
+
+# Initiate gradient descent
+optim.step() # gradient descent

tensors.py

@@ -0,0 +1,49 @@
+import torch
+import numpy as np
+
+# Examples of Tensors being initialized
+data = [[1,2],[3,4]]
+
+# Directly from data, data type is inferred
+xData = torch.tensor(data)
+
+# From a NumPy array
+npArray = np.array(data)
+xNP = torch.from_numpy(npArray)
+
+# From another tensor, retaining properties of the original
+xOnes = torch.ones_like(xData)
+print(f"Ones Tensor: \n {xOnes} \n")
+
+# From another tensor, overriding the datatype
+xRand = torch.rand_like(xData, dtype=torch.float)
+print(f"Random Tensor: \n {xRand} \n")
+
+# shape defines the tensor dimensions
+shape = (2,3,)
+randomTensor = torch.rand(shape)
+onesTensor = torch.ones(shape)
+zerosTensor = torch.zeros(shape)
+
+print(f"Random Tensor: \n {randomTensor} \n")
+print(f"Ones Tensor: \n {onesTensor} \n")
+print(f"Zeros Tensor: \n {zerosTensor} \n")
+
+# Tensors have attributes for their shape, datatype, and device
+attrDemoTensor = torch.rand(3,4)
+print(f"Shape: {attrDemoTensor.shape}")
+print(f"Datatype: {attrDemoTensor.dtype}")
+print(f"Device: {attrDemoTensor.device} \n")
+
+# Torch has a ton of tensor operations buil-in, apparently similar to NumPy
+tensor = torch.ones(4,4)
+print(f"{tensor} \n")
+tensor[:,1] = 0
+print(f"{tensor} \n")
+
+t1 = torch.cat([tensor, tensor, tensor], dim=1)
+print(f"{t1} \n")
+
+# Operations with a _ suffix are performed in-place
+tensor.add_(5)
+print(tensor)

test.py

@@ -0,0 +1,6 @@
+import torch
+
+print("Hello World!")
+
+x = torch.rand(5,3)
+print(x)