update

goodman1204 · Mar 18, 2021 · d2c716a · d2c716a
1 parent fe18ba7
commit d2c716a
Show file tree

Hide file tree

Showing 47 changed files with 103,086 additions and 16 deletions.
diff --git a/can.py b/can.py
@@ -107,6 +107,7 @@ def training(args):
     mean_precision=[]
     mean_recall = []
     mean_entropy = []
+    mean_time= []
 
 
     # adj_norm = drop_edge(adj_norm,Y)
@@ -152,6 +153,7 @@ def training(args):
         loss_list=None
         pretrain_flag = False
 
+        start_time = time.time()
         for epoch in range(args.epochs):
             t = time.time()
 
@@ -199,6 +201,8 @@ def training(args):
                   "time=", "{:.5f}".format(time.time() - t))
 
         print("Optimization Finished!")
+        end_time = time.time()
+        print("total time spend:", end_time- start_time)
 
         pre, mu_c=clustering_latent_space(z.cpu().detach().numpy(),tru)
         plot_tsne(args.dataset,args.model,epoch,z.cpu(),torch.tensor(mu_c),Y,pre)
@@ -226,6 +230,7 @@ def training(args):
         mean_precision.append(round(precision,4))
         mean_recall.append(round(recall,4))
         mean_entropy.append(round(entropy,4))
+        mean_time.append(round(end_time-start_time,4))
 
 
         # np.save(embedding_node_mean_result_file, mu_u.data.numpy())
@@ -240,7 +245,8 @@ def training(args):
         # print('Test edge AP score: ' + str(ap_score))
         # print('Test attr ROC score: ' + str(roc_score_a))
         # print('Test attr AP score: ' + str(ap_score_a))
-    metrics_list=[mean_h,mean_c,mean_v,mean_ari,mean_ami,mean_nmi,mean_purity,mean_accuracy,mean_f1,mean_precision,mean_recall,mean_entropy]
+    # metrics_list=[mean_h,mean_c,mean_v,mean_ari,mean_ami,mean_nmi,mean_purity,mean_accuracy,mean_f1,mean_precision,mean_recall,mean_entropy]
+    metrics_list=[mean_h,mean_c,mean_v,mean_ari,mean_ami,mean_nmi,mean_purity,mean_accuracy,mean_f1,mean_precision,mean_recall,mean_entropy,mean_time]
     save_results(args,metrics_list)
 
     ###### Report Final Results ######
@@ -281,7 +287,7 @@ def parse_args():
 if __name__ == '__main__':
     args = parse_args()
     if args.cuda:
-        torch.cuda.set_device(4)
+        torch.cuda.set_device(1)
         # torch.cuda.manual_seed(args.seed)
     # random.seed(args.seed)
     # np.random.seed(args.seed)

diff --git a/daegce.py b/daegce.py
@@ -102,6 +102,7 @@ def training(args):
     mean_precision=[]
     mean_recall = []
     mean_entropy = []
+    mean_time= []
 
 
     if args.cuda>=0:
@@ -140,6 +141,7 @@ def training(args):
 
         loss_list=None
         pretrain_flag = False
+        start_time = time.time()
         for epoch in range(args.epochs):
             t = time.time()
             model.train()
@@ -202,9 +204,12 @@ def training(args):
                   # "val_edge_ap=", "{:.5f}".format(ap_curr),
                   # "val_attr_roc=", "{:.5f}".format(roc_curr_a),
                   # "val_attr_ap=", "{:.5f}".format(ap_curr_a),
-                  "time=", "{:.5f}".format(time.time() - t))
+                  "time=", "{:.5f}".format(time.time() - t),
+                  "total time=", "{:.5f}".format(time.time() - start_time))
 
         print("Optimization Finished!")
+        end_time = time.time()
+        print("total time spend:", end_time- start_time)
 
         pre,gamma_c = model.predict_soft_assignment(z)
 
@@ -234,11 +239,13 @@ def training(args):
         mean_precision.append(round(precision,4))
         mean_recall.append(round(recall,4))
         mean_entropy.append(round(entropy,4))
+        mean_time.append(round(end_time-start_time,4))
 
         plot_tsne(args.dataset,args.model,epoch,z.cpu(),model.mu_c.cpu(),tru,pre)
 
 
-    metrics_list=[mean_h,mean_c,mean_v,mean_ari,mean_ami,mean_nmi,mean_purity,mean_accuracy,mean_f1,mean_precision,mean_recall,mean_entropy]
+    # metrics_list=[mean_h,mean_c,mean_v,mean_ari,mean_ami,mean_nmi,mean_purity,mean_accuracy,mean_f1,mean_precision,mean_recall,mean_entropy]
+    metrics_list=[mean_h,mean_c,mean_v,mean_ari,mean_ami,mean_nmi,mean_purity,mean_accuracy,mean_f1,mean_precision,mean_recall,mean_entropy,mean_time]
     save_results(args, metrics_list)
     ###### Report Final Results ######
     print('Homogeneity:{}\t mean:{}\t std:{}\n'.format(mean_h,round(np.mean(mean_h),4),round(np.std(mean_h),4)))

diff --git a/gcn_ae_flickr_log b/gcn_ae_flickr_log
diff --git a/gcn_can_flickr_log b/gcn_can_flickr_log
@@ -0,0 +1,18 @@
+nohup: ignoring input
+Using Flickr dataset
+dataset:Flickr, node_num:7575,edge_num:479476,attribute_num:12047
+imported graph edge number (without selfloop):239738.0
+cluster number:9
+node size:7575, feature size:12047
+graph edge number after mask:239738.0
+graph edge number after normalize adjacent matrix:239738.0
+Traceback (most recent call last):
+  File "can.py", line 289, in <module>
+    training(args)
+  File "can.py", line 159, in training
+    loss_list,[mu_u, logvar_u, mu_a, logvar_a,z] = model.loss(features_training,adj_norm,labels = (adj_label, features_label), n_nodes = n_nodes, n_features = n_features,norm = (norm_u, norm_a), pos_weight = (pos_weight_u, pos_weight_a))
+  File "/home/syang/project/github_code/node-clustering/model.py", line 716, in loss
+    cost_a = norm_a * F.binary_cross_entropy_with_logits(pred_x, labels_sub_a, pos_weight = pos_weight_a)
+  File "/home/syang/anaconda3/lib/python3.7/site-packages/torch/nn/functional.py", line 2435, in binary_cross_entropy_with_logits
+    return torch.binary_cross_entropy_with_logits(input, target, weight, pos_weight, reduction_enum)
+RuntimeError: CUDA out of memory. Tried to allocate 350.00 MiB (GPU 0; 31.75 GiB total capacity; 3.39 GiB already allocated; 34.50 MiB free; 3.41 GiB reserved in total by PyTorch)