Upload folder using huggingface_hub

decoder.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73c42a41e9a9bf115ea236aa5dfe7690d07123270c362217d0adc4d18ce2a4e8
+size 802465833

decoder.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc47dc8c28c257501ee1e843d76772fca1b803c25d196b122d65b549967c22c9
+size 2406851328

decoder_earlystop_55.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db2cfe59b08c6b05178d213901e71c07e8d843318caaf198ccba1661283372f9
+size 802465833

decoder_earlystop_82.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9656b7cb181a8a6cc3af9fe9381a5137fe6249e2e03cf27abfce79bc7062db9c
+size 802465833

models.py

@@ -0,0 +1,74 @@
+import torch
+import torch.nn as nn
+
+class Decoder(nn.Module):
+    def __init__(self, input_dim, hidden_dim, gamma=0.1):
+        super().__init__()
+        self.input_dim = input_dim
+        self.hidden_dim = hidden_dim
+        self.gamma = gamma
+        self.float()
+        
+        #should be 512, 1024
+        self.fc = nn.Sequential(
+            nn.Linear(input_dim, hidden_dim),
+            nn.BatchNorm1d(hidden_dim),
+            nn.ReLU(),
+            nn.Linear(hidden_dim, hidden_dim * 2),
+            nn.BatchNorm1d(hidden_dim * 2),
+            nn.ReLU(),
+            nn.Linear(hidden_dim * 2, hidden_dim * 4),
+            nn.BatchNorm1d(hidden_dim * 4),
+            nn.ReLU(),
+            nn.Linear(hidden_dim * 4, hidden_dim * 8),
+            nn.BatchNorm1d(hidden_dim * 8),
+            nn.ReLU(),
+            nn.Linear(hidden_dim * 8, hidden_dim * 4 * 4),
+            nn.BatchNorm1d(hidden_dim * 4 * 4),
+            nn.ReLU()
+        )
+
+        self.decoder = nn.Sequential(
+            nn.ConvTranspose2d(1024, 768, kernel_size=4, stride=2, padding=1),
+            nn.BatchNorm2d(768),
+            nn.ReLU(),
+            nn.ConvTranspose2d(768, 512, kernel_size=4, stride=2, padding=1),
+            nn.BatchNorm2d(512),
+            nn.ReLU(),
+            nn.ConvTranspose2d(512, 256, kernel_size=4, stride=2, padding=1),
+            nn.BatchNorm2d(256),
+            nn.ReLU(),
+            nn.ConvTranspose2d(256, 128, kernel_size=4, stride=2, padding=1),
+            nn.BatchNorm2d(128),
+            nn.ReLU(),
+            nn.ConvTranspose2d(128, 64, kernel_size=4, stride=2, padding=1),
+            nn.BatchNorm2d(64),
+            nn.ReLU(),
+            nn.ConvTranspose2d(64, 32, kernel_size=4, stride=2, padding=1),
+            nn.BatchNorm2d(32),
+            nn.ReLU(),
+            nn.Conv2d(32, 3, kernel_size=3, padding=1),
+            nn.Sigmoid()
+        )
+    
+    def forward(self, z):
+        batch_size = z.shape[0]
+        # adding noise to inputs 
+        gamma = 0.05
+        z = z + self.gamma * torch.randn_like(z)
+        z = self.fc(z)
+        z = z.view(batch_size, 1024, 4, 4)
+        return self.decoder(z)
+        
+    def get_loss(self, emb, x):
+        x_hat = self.forward(emb)
+        l = nn.MSELoss(reduction="mean")
+        loss = l(x_hat, x)
+        return loss
+    
+    @torch.no_grad()
+    def sample(self, samples, device):
+        samples = samples.to(device)
+        x_hat = self.forward(samples)
+
+        return x_hat