visualization reconstructions

.idea/Pytorch_CNN_mixt_representation.iml

@@ -2,7 +2,7 @@
 <module type="PYTHON_MODULE" version="4">
   <component name="NewModuleRootManager">
     <content url="file://$MODULE_DIR$" />
-    <orderEntry type="jdk" jdkName="Python 3.7" jdkType="Python SDK" />
+    <orderEntry type="jdk" jdkName="Python 3.8 (Pytorch_CNN_mixt_representation)" jdkType="Python SDK" />
     <orderEntry type="sourceFolder" forTests="false" />
   </component>
   <component name="PyDocumentationSettings">

.idea/misc.xml

@@ -3,5 +3,5 @@
   <component name="JavaScriptSettings">
     <option name="languageLevel" value="ES6" />
   </component>
-  <component name="ProjectRootManager" version="2" project-jdk-name="Python 3.7" project-jdk-type="Python SDK" />
+  <component name="ProjectRootManager" version="2" project-jdk-name="Python 3.8 (Pytorch_CNN_mixt_representation)" project-jdk-type="Python SDK" />
 </project>
\ No newline at end of file

Experiments/experiments.py

@@ -8,44 +8,31 @@ import torch
 
 
 def viz_reconstruction(model, path, expe_name, batch):
-<<<<<<< HEAD
-=======
-
->>>>>>> 0c34b372fa08007c42e406c24e2b23ddea1753b3
     file_path = os.path.join(path, expe_name, 'checkpoints', 'last')
     checkpoint = torch.load(file_path, map_location=torch.device('cpu'))
     model.load_state_dict(checkpoint['model_states']['model'])
+    nb_epochs = checkpoint['iter']
 
     viz_chairs = Viz(model)
     viz_chairs.save_images = False
 
     recon_grid, _ = viz_chairs.reconstructions(batch, size=(8, 8))
-    fig = plt.figure(figsize=(10, 10))
+    plt.figure(figsize=(10, 10))
     recon_grid = recon_grid.permute(1, 2, 0)
-<<<<<<< HEAD
-    plt.title(expe_name)
-=======
->>>>>>> 0c34b372fa08007c42e406c24e2b23ddea1753b3
+    plt.title('model: {}, nb_epochs trained: {}'.format(expe_name, nb_epochs))
     plt.imshow(recon_grid.numpy())
     plt.savefig('../reconstruction_im/charis_' + expe_name + '.png')
     plt.show()
 
 
-<<<<<<< HEAD
-def plot_loss(expe_name=None, path=None, save=False):
-=======
 def plot_loss(expe_name=None, save=False, path=None):
 
->>>>>>> 0c34b372fa08007c42e406c24e2b23ddea1753b3
     file_path = os.path.join(path, expe_name, 'checkpoints', 'last')
     checkpoint = torch.load(file_path, map_location=torch.device('cpu'))
     losses = checkpoint['loss']
     title = 'losses model:' + expe_name
 
-<<<<<<< HEAD
     plt.title(title)
-=======
->>>>>>> 0c34b372fa08007c42e406c24e2b23ddea1753b3
     plt.plot(losses)
     plt.xlabel('Epochs')
     plt.ylabel('loss')
@@ -64,7 +51,6 @@ _, dataloader_chairs = get_chairs_dataloader(batch_size=32)
 for batch_chairs, labels_chairs in dataloader_chairs:
     break
 
-# torch.load('data/batch_chairs.pt')
 
 path_to_model_folder_chairs = '../trained_models/rendered_chairs/'
 list_expe = ['VAE_bs_64', 'VAE_bs_256', 'beta_VAE_bs_64', 'beta_VAE_bs_256', 'VAE_bs_64_ls_10_lr_1e_3',
@@ -73,22 +59,20 @@ list_expe = ['VAE_bs_64', 'VAE_bs_256', 'beta_VAE_bs_64', 'beta_VAE_bs_256', 'VA
 list_expe_ls_5 = ['VAE_bs_64_ls_5', 'beta_VAE_bs_64_ls_5']
 list_expe_ls_15 = ['VAE_bs_64_ls_15', 'beta_VAE_bs_64_ls_15']
 list_expe_ls_20 = ['VAE_bs_64_ls_20', 'beta_VAE_bs_64_ls_20']
+list_expe_ls_30 = ['VAE_bs_64_ls_30']
+list_expe_ls_40 = ['VAE_bs_64_ls_40']
+list_expe_ls_50 = ['VAE_bs_64_ls_50']
 
-<<<<<<< HEAD
 list_expe_ls_10_64_64_128_128 = ['VAE_bs_64_conv_64_64_128_128']
-list_expe_ls_10_128_128_256_256 = ['VAE_bs_64_conv_128_128_256_256']
 
 img_size = (3, 64, 64)
-
 path = '../trained_models/rendered_chairs'
+
+"""
 for i in list_expe_ls_5:
     plot_loss(i, path=path)
-
 """
-=======
-img_size = (3, 64, 64)
 
->>>>>>> 0c34b372fa08007c42e406c24e2b23ddea1753b3
 latent_spec = {"cont": 10}
 model = VAE(img_size, latent_spec=latent_spec)
 for i in list_expe:
@@ -108,18 +92,24 @@ latent_spec = {"cont": 20}
 model = VAE(img_size, latent_spec=latent_spec)
 for i in list_expe_ls_20:
     viz_reconstruction(model, path_to_model_folder_chairs, i, batch_chairs)
-<<<<<<< HEAD
+
+latent_spec = {"cont": 30}
+model = VAE(img_size, latent_spec=latent_spec)
+for i in list_expe_ls_30:
+    viz_reconstruction(model, path_to_model_folder_chairs, i, batch_chairs)
+
+latent_spec = {"cont": 40}
+model = VAE(img_size, latent_spec=latent_spec)
+for i in list_expe_ls_40:
+    viz_reconstruction(model, path_to_model_folder_chairs, i, batch_chairs)
+
+latent_spec = {"cont": 50}
+model = VAE(img_size, latent_spec=latent_spec)
+for i in list_expe_ls_50:
+    viz_reconstruction(model, path_to_model_folder_chairs, i, batch_chairs)
 
 latent_spec = {"cont": 10}
 model = VAE(img_size, latent_spec=latent_spec, nb_filter_conv1=64, nb_filter_conv2=64, nb_filter_conv3=128,
             nb_filter_conv4=128)
 for i in list_expe_ls_10_64_64_128_128:
     viz_reconstruction(model, path_to_model_folder_chairs, i, batch_chairs)
-
-model = VAE(img_size, latent_spec=latent_spec, nb_filter_conv1=128, nb_filter_conv2=128, nb_filter_conv3=256,
-            nb_filter_conv4=256)
-for i in list_expe_ls_10_128_128_256_256:
-    viz_reconstruction(model, path_to_model_folder_chairs, i, batch_chairs)
-"""
-=======
->>>>>>> 0c34b372fa08007c42e406c24e2b23ddea1753b3

main.py

@@ -77,10 +77,7 @@ def main(args):
 
     # Define trainer
     trainer = Trainer(model, device, optimizer, criterion, save_step=args.save_step, ckpt_dir=args.ckpt_dir,
-<<<<<<< HEAD
                       load_model_checkpoint=args.load_model_checkpoint,
-=======
->>>>>>> 0c34b372fa08007c42e406c24e2b23ddea1753b3
                       ckpt_name=args.ckpt_name,
                       expe_name=args.experiment_name,
                       dataset=args.dataset,