I was doing the PyTorch Deep Learning course from FreeCodeCamp and the doubt is:

weight = 0.7
bias = 0.3
start = 0
end = 1
step = 0.02

X = torch.arange(start, end, step).unsqueeze(dim=1)
y=weight*X + bias
X[:10], y[:10]
train_split=int(0.8*len(X))
X_train, y_train = X[:train_split], y[:train_split]
X_test, y_test=X[train_split:], y[train_split:]

Why the unsqueeze function is used to make the tensor of size [50, 1] and not [50]? The mentor was telling that it will cause error but I don't know why the error is happening?

Can you answer this question using maths and as well as basic fundamentals without maths as well ?

After trying to train the model I am getting this error:

class LinearRegressionModelv2(nn.Module):
  def __init__(self):
    super().__init__()
    self.linear_layer = nn.Linear(in_features=1, out_features=1)

  def forward(self, x: torch.Tensor) -> torch.Tensor:
    return self.linear_layer(x)

torch.manual_seed(42)
model_v2 = LinearRegressionModelv2()

y_prediction = model_v2(X_train)

IndexError: Dimension out of range (expected to be in range of [-1, 0], but got -2)

I was doing the PyTorch Deep Learning course from FreeCodeCamp and the doubt is:

weight = 0.7
bias = 0.3
start = 0
end = 1
step = 0.02

X = torch.arange(start, end, step).unsqueeze(dim=1)
y=weight*X + bias
X[:10], y[:10]
train_split=int(0.8*len(X))
X_train, y_train = X[:train_split], y[:train_split]
X_test, y_test=X[train_split:], y[train_split:]

Why the unsqueeze function is used to make the tensor of size [50, 1] and not [50]? The mentor was telling that it will cause error but I don't know why the error is happening?

Can you answer this question using maths and as well as basic fundamentals without maths as well ?

After trying to train the model I am getting this error:

class LinearRegressionModelv2(nn.Module):
  def __init__(self):
    super().__init__()
    self.linear_layer = nn.Linear(in_features=1, out_features=1)

  def forward(self, x: torch.Tensor) -> torch.Tensor:
    return self.linear_layer(x)

torch.manual_seed(42)
model_v2 = LinearRegressionModelv2()

y_prediction = model_v2(X_train)

IndexError: Dimension out of range (expected to be in range of [-1, 0], but got -2)

• python
• machine-learning
• deep-learning
• pytorch

• 3 You need to include more information for it to be helpful since this doesn't show what your X_train is or what your model_v2 was actually trained on. Generally, though, unsqueeze would be to get dimensions to match. In this case it's probably for making the first dimension an index into the dataset, where each datapoint is a single value – Chrispresso Commented Feb 3 at 15:35
• Added information about X_train and model_v2 – SouraOP Commented Feb 4 at 9:34
• Which error are you referring to? At which location does it happen, can you plase post a meaningful traceback. Your code runs fine for me, and if I remove the unsqueeze I get a different error than the one you have posted. – Daraan Commented Feb 4 at 9:40

1 Answer 1

Mathematically, the formula for Linear regression is given by

(y = X.W^T + b )

here,

• Input tensor, X, has shape (batch_size, in_features)
• Weight matrix, W has shape (out_features, in_features)
• Bias, B is a vector of shape (out_features).

When you make X in 1D (e.g., [50] in your case), matrix multiplication is undefined because the dimensions don't align. Therefore, to ensure correct computation input data (X) should be 2 dimensional, where:

• each row represents a sample
• and each column represents a feature.