Skip to content

fix: correct decoder layer init in autoencoder notebooks (ae-basic) #83

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
Mukller wants to merge 2 commits into
base: master
Choose a base branch
from
Open

fix: correct decoder layer init in autoencoder notebooks (ae-basic) #83

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
8a81a59
fix: correct decoder weight init in ae-basic.ipynb
Mukller Jul 9, 2026
6643b7f
fix: correct decoder weight init in ae-basic-with-rf.ipynb
Mukller Jul 9, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
55 changes: 2 additions & 53 deletions pytorch_ipynb/autoencoder/ae-basic-with-rf.ipynb
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -151,58 +151,7 @@
"metadata": {},
"outputs": [],
"source": [
"##########################\n",
"### MODEL\n",
"##########################\n",
"\n",
"class Autoencoder(torch.nn.Module):\n",
"\n",
" def __init__(self, num_features):\n",
" super(Autoencoder, self).__init__()\n",
" \n",
" ### ENCODER\n",
" \n",
" self.linear_1 = torch.nn.Linear(num_features, num_hidden_1)\n",
" # The following to lones are not necessary, \n",
" # but used here to demonstrate how to access the weights\n",
" # and use a different weight initialization.\n",
" # By default, PyTorch uses Xavier/Glorot initialization, which\n",
" # should usually be preferred.\n",
" self.linear_1.weight.detach().normal_(0.0, 0.1)\n",
" self.linear_1.bias.detach().zero_()\n",
" \n",
" ### DECODER\n",
" self.linear_2 = torch.nn.Linear(num_hidden_1, num_features)\n",
" self.linear_1.weight.detach().normal_(0.0, 0.1)\n",
" self.linear_1.bias.detach().zero_()\n",
" \n",
" def encoder(self, x):\n",
" encoded = self.linear_1(x)\n",
" encoded = F.leaky_relu(encoded)\n",
" return encoded\n",
" \n",
" def decoder(self, encoded_x):\n",
" logits = self.linear_2(encoded_x)\n",
" decoded = torch.sigmoid(logits)\n",
" return decoded\n",
" \n",
"\n",
" def forward(self, x):\n",
" \n",
" ### ENCODER\n",
" encoded = self.encoder(x)\n",
" \n",
" ### DECODER\n",
" decoded = self.decoder(encoded)\n",
" \n",
" return decoded\n",
"\n",
" \n",
"torch.manual_seed(random_seed)\n",
"model = Autoencoder(num_features=num_features)\n",
"model = model.to(device)\n",
"\n",
"optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate) "
"##########################\n### MODEL\n##########################\n\nclass Autoencoder(torch.nn.Module):\n\n def __init__(self, num_features):\n super(Autoencoder, self).__init__()\n \n ### ENCODER\n \n self.linear_1 = torch.nn.Linear(num_features, num_hidden_1)\n # The following to lones are not necessary, \n # but used here to demonstrate how to access the weights\n # and use a different weight initialization.\n # By default, PyTorch uses Xavier/Glorot initialization, which\n # should usually be preferred.\n self.linear_1.weight.detach().normal_(0.0, 0.1)\n self.linear_1.bias.detach().zero_()\n \n ### DECODER\n self.linear_2 = torch.nn.Linear(num_hidden_1, num_features)\n self.linear_2.weight.detach().normal_(0.0, 0.1)\n self.linear_2.bias.detach().zero_()\n \n def encoder(self, x):\n encoded = self.linear_1(x)\n encoded = F.leaky_relu(encoded)\n return encoded\n \n def decoder(self, encoded_x):\n logits = self.linear_2(encoded_x)\n decoded = torch.sigmoid(logits)\n return decoded\n \n\n def forward(self, x):\n \n ### ENCODER\n encoded = self.encoder(x)\n \n ### DECODER\n decoded = self.decoder(encoded)\n \n return decoded\n\n \ntorch.manual_seed(random_seed)\nmodel = Autoencoder(num_features=num_features)\nmodel = model.to(device)\n\noptimizer = torch.optim.Adam(model.parameters(), lr=learning_rate) "
]
},
{
Expand Down Expand Up @@ -684,4 +633,4 @@
},
"nbformat": 4,
"nbformat_minor": 4
}
}
46 changes: 2 additions & 44 deletions pytorch_ipynb/autoencoder/ae-basic.ipynb
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -151,49 +151,7 @@
"metadata": {},
"outputs": [],
"source": [
"##########################\n",
"### MODEL\n",
"##########################\n",
"\n",
"class Autoencoder(torch.nn.Module):\n",
"\n",
" def __init__(self, num_features):\n",
" super(Autoencoder, self).__init__()\n",
" \n",
" ### ENCODER\n",
" self.linear_1 = torch.nn.Linear(num_features, num_hidden_1)\n",
" # The following to lones are not necessary, \n",
" # but used here to demonstrate how to access the weights\n",
" # and use a different weight initialization.\n",
" # By default, PyTorch uses Xavier/Glorot initialization, which\n",
" # should usually be preferred.\n",
" self.linear_1.weight.detach().normal_(0.0, 0.1)\n",
" self.linear_1.bias.detach().zero_()\n",
" \n",
" ### DECODER\n",
" self.linear_2 = torch.nn.Linear(num_hidden_1, num_features)\n",
" self.linear_1.weight.detach().normal_(0.0, 0.1)\n",
" self.linear_1.bias.detach().zero_()\n",
" \n",
"\n",
" def forward(self, x):\n",
" \n",
" ### ENCODER\n",
" encoded = self.linear_1(x)\n",
" encoded = F.leaky_relu(encoded)\n",
" \n",
" ### DECODER\n",
" logits = self.linear_2(encoded)\n",
" decoded = torch.sigmoid(logits)\n",
" \n",
" return decoded\n",
"\n",
" \n",
"torch.manual_seed(random_seed)\n",
"model = Autoencoder(num_features=num_features)\n",
"model = model.to(device)\n",
"\n",
"optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate) "
"##########################\n### MODEL\n##########################\n\nclass Autoencoder(torch.nn.Module):\n\n def __init__(self, num_features):\n super(Autoencoder, self).__init__()\n \n ### ENCODER\n self.linear_1 = torch.nn.Linear(num_features, num_hidden_1)\n # The following to lones are not necessary, \n # but used here to demonstrate how to access the weights\n # and use a different weight initialization.\n # By default, PyTorch uses Xavier/Glorot initialization, which\n # should usually be preferred.\n self.linear_1.weight.detach().normal_(0.0, 0.1)\n self.linear_1.bias.detach().zero_()\n \n ### DECODER\n self.linear_2 = torch.nn.Linear(num_hidden_1, num_features)\n self.linear_2.weight.detach().normal_(0.0, 0.1)\n self.linear_2.bias.detach().zero_()\n \n\n def forward(self, x):\n \n ### ENCODER\n encoded = self.linear_1(x)\n encoded = F.leaky_relu(encoded)\n \n ### DECODER\n logits = self.linear_2(encoded)\n decoded = torch.sigmoid(logits)\n \n return decoded\n\n \ntorch.manual_seed(random_seed)\nmodel = Autoencoder(num_features=num_features)\nmodel = model.to(device)\n\noptimizer = torch.optim.Adam(model.parameters(), lr=learning_rate) "
]
},
{
Expand Down Expand Up @@ -377,4 +335,4 @@
},
"nbformat": 4,
"nbformat_minor": 2
}
}