ISSUE RESOLVED: Video:" PyTorch Workflow" - video #49 (Making Predictions)

[sankar82]

At 4:56, when you run y_preds after setting right the indentation error of forward function, y_preds should return 10 predicted values. For some weird reason, I am getting a single value as output (in step 6 below).

I have tried to rerun the whole thing multiple times right from the beginning. I also checked X_test and y_test, they both contain 10 values. Please see the entire code below.

# 1. Creating dataset
Create known parameters
weight = 0.7
bias = 0.3

Create
start = 0
end = 1
step = 0.02
X = torch.arange(start,end,step).unsqueeze(dim=1)
y = weight * X + bias
X[:10],y[:10]

# 2. Create train/test split
train_split = int(0.8 * len(X))
X_train,y_train = X[:train_split],y[:train_split] # indexing
X_test,y_test = X[train_split:],y[train_split:]

len(X_train), len(y_train),len(X_test),len(y_test)

# 3. Defining the model
from torch import nn

class LinearRegressionModel(nn.Module):
  def __init__(self):
    super().__init__() 
    self.weights = nn.Parameter(torch.randn(1, 
                                            requires_grad=True,
                                            dtype=torch.float))
    
    self.bias = nn.Parameter(torch.randn(1,
                                         requires_grad=True,
                                         dtype=torch.float))
    
  Forward method to define the computation in the model
  def forward(self, x: torch.Tensor) -> torch.Tensor:
    return self.weights * self.bias 

# 4. Instantiating the model
Create a random seed
torch.manual_seed(42)

Create an instance of the model (this is a subclass)
model_0 = LinearRegressionModel()

# 5. Predicting the y-value
with torch.inference_mode():
  y_preds = model_0(X_test)

y_preds

# 6. Wrong output
The output I get is tensor([0.0434]). 

Thanks a lot and appreciate any help on this. I am sure it is something silly, but for the love of PyTorch I am not able to figure it out.

[sankar82]

I am not able to delete the question. But I found out the mistake. I had written the formula wrong within forward(). I fixed it now. Thanks.

[mrdbourke]

Glad to hear you got it fixed! I get caught writing the forward() method wrong sometimes too.

Thank you for updating your answer

[sankar82]

Thanks Daniel. Just wanted to say I am loving this course and perhaps the best course I have taken. I was always a superficial learned, never went into depths. But this time I am taking time to learn and literally 'burning' the fundamentals into my brain. I love your curiosity and method of teaching as well as learning.

In fact, had you not shown how to calculate y_pred without inference_mode(), I would have never figured out the problem. Yours showed AddBackward while mine showed MulBackward. I realized my mistake instantly. Thank you for this wonderful course.