Update.

[pytorch.git] / tiny_vae.py

diff --git a/tiny_vae.py b/tiny_vae.py
index d33cc4b..cba42e1 100755 (executable)
--- a/tiny_vae.py
+++ b/tiny_vae.py
@@ -28,9 +28,9 @@ parser = argparse.ArgumentParser(
     description="Very simple implementation of a VAE for teaching."
 )
 
-parser.add_argument("--nb_epochs", type=int, default=25)
+parser.add_argument("--nb_epochs", type=int, default=100)
 
-parser.add_argument("--learning_rate", type=float, default=1e-3)
+parser.add_argument("--learning_rate", type=float, default=2e-4)
 
 parser.add_argument("--batch_size", type=int, default=100)
 
@@ -44,11 +44,6 @@ parser.add_argument("--nb_channels", type=int, default=128)
 
 parser.add_argument("--no_dkl", action="store_true")
 
-# With that option, do not follow the setup of the original VAE paper
-# of forcing the variance of X|Z to 1 during training and to 0 for
-# sampling, but optimize and use the variance.
-parser.add_argument("--no_hacks", action="store_true")
-
 args = parser.parse_args()
 
 log_file = open(args.log_filename, "w")
@@ -70,12 +65,14 @@ def log_string(s):
 ######################################################################
 
 
-def sample_gaussian(mu, log_var):
+def sample_gaussian(param):
+    mu, log_var = param
     std = log_var.mul(0.5).exp()
     return torch.randn(mu.size(), device=mu.device) * std + mu
 
 
-def log_p_gaussian(x, mu, log_var):
+def log_p_gaussian(x, param):
+    mu, log_var = param
     var = log_var.exp()
     return (
         (-0.5 * ((x - mu).pow(2) / var) - 0.5 * log_var - 0.5 * math.log(2 * math.pi))
@@ -84,9 +81,9 @@ def log_p_gaussian(x, mu, log_var):
     )
 
 
-def dkl_gaussians(mean_a, log_var_a, mean_b, log_var_b):
-    mean_a, log_var_a = mean_a.flatten(1), log_var_a.flatten(1)
-    mean_b, log_var_b = mean_b.flatten(1), log_var_b.flatten(1)
+def dkl_gaussians(param_a, param_b):
+    mean_a, log_var_a = param_a[0].flatten(1), param_a[1].flatten(1)
+    mean_b, log_var_b = param_b[0].flatten(1), param_b[1].flatten(1)
     var_a = log_var_a.exp()
     var_b = log_var_b.exp()
     return 0.5 * (
@@ -144,8 +141,7 @@ class ImageGivenLatentNet(nn.Module):
     def forward(self, z):
         output = self.model(z.view(z.size(0), -1, 1, 1))
         mu, log_var = output[:, 0:1], output[:, 1:2]
-        if not args.no_hacks:
-            log_var[...] = 0
+        log_var.flatten(1)[...] = log_var.flatten(1)[:, :1]
         return mu, log_var
 
 
@@ -187,28 +183,26 @@ test_input.sub_(train_mu).div_(train_std)
 
 ######################################################################
 
-mean_p_Z = train_input.new_zeros(1, args.latent_dim)
-log_var_p_Z = mean_p_Z
+zeros = train_input.new_zeros(1, args.latent_dim)
+
+param_p_Z = zeros, zeros
 
 for epoch in range(args.nb_epochs):
     acc_loss = 0
 
     for x in train_input.split(args.batch_size):
-        mean_q_Z_given_x, log_var_q_Z_given_x = model_q_Z_given_x(x)
-        z = sample_gaussian(mean_q_Z_given_x, log_var_q_Z_given_x)
-        mean_p_X_given_z, log_var_p_X_given_z = model_p_X_given_z(z)
+        param_q_Z_given_x = model_q_Z_given_x(x)
+        z = sample_gaussian(param_q_Z_given_x)
+        param_p_X_given_z = model_p_X_given_z(z)
+        log_p_x_given_z = log_p_gaussian(x, param_p_X_given_z)
 
         if args.no_dkl:
-            log_q_z_given_x = log_p_gaussian(z, mean_q_Z_given_x, log_var_q_Z_given_x)
-            log_p_x_z = log_p_gaussian(
-                x, mean_p_X_given_z, log_var_p_X_given_z
-            ) + log_p_gaussian(z, mean_p_Z, log_var_p_Z)
+            log_q_z_given_x = log_p_gaussian(z, param_q_Z_given_x)
+            log_p_z = log_p_gaussian(z, param_p_Z)
+            log_p_x_z = log_p_x_given_z + log_p_x_z
             loss = -(log_p_x_z - log_q_z_given_x).mean()
         else:
-            log_p_x_given_z = log_p_gaussian(x, mean_p_X_given_z, log_var_p_X_given_z)
-            dkl_q_Z_given_x_from_p_Z = dkl_gaussians(
-                mean_q_Z_given_x, log_var_q_Z_given_x, mean_p_Z, log_var_p_Z
-            )
+            dkl_q_Z_given_x_from_p_Z = dkl_gaussians(param_q_Z_given_x, param_p_Z)
             loss = (-log_p_x_given_z + dkl_q_Z_given_x_from_p_Z).mean()
 
         optimizer.zero_grad()
@@ -235,23 +229,21 @@ save_image(x, "input.png")
 
 # Save the same images after encoding / decoding
 
-mean_q_Z_given_x, log_var_q_Z_given_x = model_q_Z_given_x(x)
-z = sample_gaussian(mean_q_Z_given_x, log_var_q_Z_given_x)
-mean_p_X_given_z, log_var_p_X_given_z = model_p_X_given_z(z)
-if args.no_hacks:
-    x = sample_gaussian(mean_p_X_given_z, log_var_p_X_given_z)
-else:
-    x = mean_p_X_given_z
+param_q_Z_given_x = model_q_Z_given_x(x)
+z = sample_gaussian(param_q_Z_given_x)
+param_p_X_given_z = model_p_X_given_z(z)
+x = sample_gaussian(param_p_X_given_z)
 save_image(x, "output.png")
+save_image(param_p_X_given_z[0], "output_mean.png")
 
 # Generate a bunch of images
 
-z = sample_gaussian(mean_p_Z.expand(x.size(0), -1), log_var_p_Z.expand(x.size(0), -1))
-mean_p_X_given_z, log_var_p_X_given_z = model_p_X_given_z(z)
-if args.no_hacks:
-    x = sample_gaussian(mean_p_X_given_z, log_var_p_X_given_z)
-else:
-    x = mean_p_X_given_z
+z = sample_gaussian(
+    (param_p_Z[0].expand(x.size(0), -1), param_p_Z[1].expand(x.size(0), -1))
+)
+param_p_X_given_z = model_p_X_given_z(z)
+x = sample_gaussian(param_p_X_given_z)
 save_image(x, "synth.png")
+save_image(param_p_X_given_z[0], "synth_mean.png")
 
 ######################################################################