New parameters for office dataset training

core/train.py

@@ -144,10 +144,10 @@ def train_dann(model, params, src_data_loader, tgt_data_loader, tgt_data_loader_
             "lr": 0.0001
         }, {
             "params": model.classifier.parameters(),
-            "lr": 0.0001
+            "lr": 0.001
         }, {
             "params": model.discriminator.parameters(),
-            "lr": 0.0001
+            "lr": 0.001
         }]
         optimizer = optim.SGD(parameter_list, momentum=0.9, weight_decay=1e-4)
 
@@ -248,6 +248,12 @@ def train_dann(model, params, src_data_loader, tgt_data_loader, tgt_data_loader_
         loggi.info('\n============ Summary ============= \n')
         loggi.info('Accuracy of the %s dataset: %f' % (params.src_dataset, bestAccS))
         loggi.info('Accuracy of the %s dataset: %f' % (params.tgt_dataset, bestAcc))
+    loggi.info('Saving the final weights')
+    # save final model
+    save_model(model, params.model_root, params.src_dataset + '-' + params.tgt_dataset + "-dann-final.pt")
+    loggi.info('\n============ Summary ============= \n')
+    loggi.info('Accuracy of the %s dataset: %f' % (params.src_dataset, bestAccS))
+    loggi.info('Accuracy of the %s dataset: %f' % (params.tgt_dataset, bestAcc))
 
     return model
 

experiments/office.py

@@ -28,12 +28,16 @@ class Config(object):
     batch_size = 32
 
     # params for source dataset
-    src_dataset = "amazon31"
+    # src_dataset = "amazon31"
+    # src_dataset = "dslr31"
+    src_dataset = "webcam31"
     src_model_trained = True
     src_classifier_restore = os.path.join(model_root, src_dataset + '-source-classifier-final.pt')
 
     # params for target dataset
-    tgt_dataset = "webcam31"
+    # tgt_dataset = "webcam31"
+    # tgt_dataset = "dslr31"
+    tgt_dataset = "amazon31"
     tgt_model_trained = True
     dann_restore = os.path.join(model_root, src_dataset + '-' + tgt_dataset + '-dann-final.pt')