no_cuda.patch

diff --git a/Train.py b/Train.py
index ca3ad4e..8c19423 100644
--- a/Train.py
+++ b/Train.py
@@ -81,17 +81,17 @@ train_dataset = SceneLoader(
 
 # Setting up the model, memory items and optimizers
 if args.model_dir is not None:
-    model = torch.load(args.model_dir)
+    model = torch.load(args.model_dir, map_location='cpu')
 else:
     model = convAE(args.c, args.time_step + 1, args.msize, args.fdim, args.mdim)
-model.cuda()
+model#.cuda()
 model.train()
 
 if args.m_items_dir is not None:
-    m_items = torch.load(args.m_items_dir)
+    m_items = torch.load(args.m_items_dir, map_location='cpu')
 else:
     m_items = F.normalize(torch.rand((args.msize, args.mdim), dtype=torch.float), dim=1)
-m_items.cuda()
+m_items#.cuda()
 
 params_encoder = list(model.encoder.parameters())
 params_decoder = list(model.decoder.parameters())
@@ -159,8 +159,8 @@ for epoch in range(args.epochs):
 
             # Sampling k samples for training and validation each
             try:
-                imgs = Variable(next(train_batch)).cuda()
-                imgs_val = Variable(next(train_batch)).cuda()
+                imgs = Variable(next(train_batch))#.cuda()
+                imgs_val = Variable(next(train_batch))#.cuda()
             except StopIteration as e:
                 if scene in train_dataset.scenes:
                     train_dataset.scenes.remove(scene)
diff --git a/model/memory_final_spatial_sumonly_weight_ranking_top1.py b/model/memory_final_spatial_sumonly_weight_ranking_top1.py
index e428c6e..5cb7d6c 100644
--- a/model/memory_final_spatial_sumonly_weight_ranking_top1.py
+++ b/model/memory_final_spatial_sumonly_weight_ranking_top1.py
@@ -9,7 +9,7 @@ def random_uniform(shape, low, high, cuda):
     x = torch.rand(*shape)
     result_cpu = (high - low) * x + low
     if cuda:
-        return result_cpu.cuda()
+        return result_cpu#.cuda()
     else:
         return result_cpu
 
@@ -48,7 +48,7 @@ def MemoryLoss(memory):
     m, d = memory.size()
     memory_t = torch.t(memory)
     similarity = (torch.matmul(memory, memory_t)) / 2 + 1 / 2   # 30X30
-    identity_mask = torch.eye(m).cuda()
+    identity_mask = torch.eye(m)#.cuda()
     sim = torch.abs(similarity - identity_mask)
 
     return torch.sum(sim) / (m * (m - 1))
@@ -90,8 +90,8 @@ class Memory(nn.Module):
 
         m, d = mem.size()
         if train:
-            query_update = torch.zeros((m, d)).cuda()
-            # random_update = torch.zeros((m,d)).cuda()
+            query_update = torch.zeros((m, d))#.cuda()
+            # random_update = torch.zeros((m,d))#.cuda()
             for i in range(m):
                 idx = torch.nonzero(max_indices.squeeze(1) == i)
                 a, _ = idx.size()
@@ -103,7 +103,7 @@ class Memory(nn.Module):
             return query_update
 
         else:
-            query_update = torch.zeros((m, d)).cuda()
+            query_update = torch.zeros((m, d))#.cuda()
             for i in range(m):
                 idx = torch.nonzero(max_indices.squeeze(1) == i)
                 a, _ = idx.size()