Multi-class classification

[moranaudbxhb]

video lesson 87 .
topic : Creating a training and testing loop for a multi-class PyTorch model
problem : I am getting results that are so weird . I need to understand why

Epoch: 0 | Loss: 1.04324, Acc: 65.50% | Test Loss: 0.57861, Test Acc: 95.50%
Epoch: 10 | Loss: 0.05260, Acc: 98.75% | Test Loss: 0.00351, Test Acc: 100.00%
Epoch: 20 | Loss: 0.15444, Acc: 99.00% | Test Loss: 0.00025, Test Acc: 100.00%
Epoch: 30 | Loss: 0.16679, Acc: 99.25% | Test Loss: 0.05451, Test Acc: 99.50%
Epoch: 40 | Loss: 24442644.00000, Acc: 26.12% | Test Loss: 4618876616704.00000, Test Acc: 31.50%
Epoch: 50 | Loss: nan, Acc: 25.12% | Test Loss: nan, Test Acc: 24.50%
Epoch: 60 | Loss: nan, Acc: 25.12% | Test Loss: nan, Test Acc: 24.50%
Epoch: 70 | Loss: nan, Acc: 25.12% | Test Loss: nan, Test Acc: 24.50%
Epoch: 80 | Loss: nan, Acc: 25.12% | Test Loss: nan, Test Acc: 24.50%
Epoch: 90 | Loss: nan, Acc: 25.12% | Test Loss: nan, Test Acc: 24.50%

[LuluW8071]

You need to post your code snippet as well can't determine anything from your output alone.

[moranaudbxhb]

I am having problems uploading but it should be here in few minutes

[moranaudbxhb]

[NEURAL NETWORK CLASSIFICATION.zip](https://githu@LuluW8071
.com/mrdbourke/pytorch-deep-learning/files/14546144/NEURAL.NETWORK.CLASSIFICATION.zip)
@LuluW8071 this is my notebook

[LuluW8071]

class BlobModel(nn.Module):
  def __init__(self, input_features,output_features, hidden_units=8):
        super().__init__()
        self.linear_layer_stack = nn.Sequential(
            nn.Linear(in_features=input_features, out_features=hidden_units),
            nn.Linear(in_features=hidden_units, out_features=hidden_units),
            nn.Linear(in_features=hidden_units, out_features=output_features))

  def forward(self,x):
        # You wrote only self.layer_stack instead of self.linear_layer_stack
        return self.linear_layer_stack(x)  

Now change forward passing y_logits=model_4.linear_layer_stack(X_blob_train) to y_logits=model_4(X_blob_train)

and perform similar actions to testing code test_logits=model_4(X_blob_test)

Do same for above training code for forward passing of example logits

If your accuracy is low define another model for non-linearity and training and testing loop is same

class BlobModel(nn.Module):
  def __init__(self, input_features,output_features, hidden_units=8):
        super().__init__()
        self.linear_layer_stack = nn.Sequential(
            nn.Linear(in_features=input_features, out_features=hidden_units),
            nn.ReLU(),
            nn.Linear(in_features=hidden_units, out_features=hidden_units),
            nn.ReLU(),
            nn.Linear(in_features=hidden_units, out_features=output_features))

  def forward(self,x):
        # You wrote only self.layer_stack instead of self.linear_layer_stack
        return self.linear_layer_stack(x) 

Also do note that sometimes passing X instead of x can cause conflict in training may be incorrect results while forward passing in model class

[LuluW8071]

With no ReLU()

With ReLU()

[moranaudbxhb]

@LuluW8071 sorry for taking time to get back to you . I have made those changes and it worked . I really appreciate your effort , that was so swift.