fix gamma by sigmoid

config/APAN_large.yml

@@ -20,7 +20,7 @@ memory:
 gnn:
   - arch: 'identity'
 train:
-  - epoch: 3
+  - epoch: 50
     batch_size: 3000
     lr: 0.0002
     dropout: 0.1

examples/test_all.sh

@@ -10,23 +10,22 @@ partitions="8"
 node_per="4"
 nnodes="2"
 node_rank="0"
-probability_params=("0.6" "1")
-sample_type_params=("boundery_recent_decay" "recent")
+probability_params=("0.1")
+sample_type_params=("boundery_recent_decay")
 #"boundery_recent_decay")
 #sample_type_params=("recent" "boundery_recent_decay") #"boundery_recent_uniform")
 #memory_type=("all_update" "p2p" "all_reduce" "historical" "local")
 memory_type=("historical")
 #"historical")
 #memory_type=("local" "all_update" "historical" "all_reduce")
-shared_memory_ssim=("0.3")
-#("0" "0.1" "0.3" "0.5" "0.7" "0.9" "1.3" "1.5" "1.7" "2")
+shared_memory_ssim=("0.1" "0.3" "0.5" "0.7" "0.9" "1.3" "1.5" "1.7" "2")
 #"historical")
 #memory_type=("local" "all_update" "historical" "all_reduce")
 #data_param=("WIKI" "REDDIT" "LASTFM" "WikiTalk")
 neighbor_num=( "10" "20")
 neighbor="10"
-topk_list=("0.1" "0.02" "0.04" "0.06" "0.08" "0.1" "0.2" "0.3")
-data_param=("GDELT") 
+topk_list=("0.1")
+data_param=("LASTFM" "StackOverflow" "GDELT") 
 #"GDELT")
 #data_param=("WIKI" "REDDIT" "LASTFM" "DGraphFin" "WikiTalk" "StackOverflow")
 #data_param=("WIKI" "REDDIT" "LASTFM" "WikiTalk" "StackOverflow")

examples/train_boundery.py

@@ -206,7 +206,7 @@ def main():
     if(args.dataname=='GDELT'):
         train_param['epoch'] = 10
     #if(args.probability > 0.005):
-    train_param['epoch'] = 1
+    #train_param['epoch'] = 1
     #torch.autograd.set_detect_anomaly(True)
 # 确保 CUDA 可用
     if torch.cuda.is_available():

starrygl/module/memorys.py

@@ -406,11 +406,11 @@ class AsyncMemeoryUpdater(torch.nn.Module):
                                            wait_submit=submit_to_queue,spread_mail=spread_mail,
                                            update_cross_mm=False,
                                            )
-        #self.mailbox.update_shared()
-        #self.mailbox.update_p2p_mem()
-        #self.mailbox.update_p2p_mail()
-        #self.mailbox.sychronize_shared()
-        #self.mailbox.handle_last_async()
+        self.mailbox.update_shared()
+        self.mailbox.update_p2p_mem()
+        self.mailbox.update_p2p_mail()
+        self.mailbox.sychronize_shared()
+        self.mailbox.handle_last_async()
         if nxt_fetch_func is not None:
             nxt_fetch_func()
 
@@ -482,6 +482,7 @@ class AsyncMemeoryUpdater(torch.nn.Module):
 
                     transition_dense = b.srcdata['his_mem'][mask] + self.filter.get_incretment(shared_ind)
                     #print(transition_dense.shape)
+                    #print('check historical {} {} {} {}'.format(b.srcdata['his_mem'][mask],transition_dense,torch.isnan(transition_dense).sum(),torch.isnan(b.srcdata['his_mem'][mask]).sum()))
                     if not (transition_dense.max().item() == 0):      
                         transition_dense -= transition_dense.min()
                         transition_dense /= transition_dense.max() 
@@ -494,7 +495,9 @@ class AsyncMemeoryUpdater(torch.nn.Module):
             #upd0[mask] = self.ceil_updater(his_mem, b.srcdata['his_mem'][mask])
             #updated_memory = torch.where(mask.unsqueeze(1),self.gamma*updated_memory0 + (1-self.gamma)*(b.srcdata['his_mem'])
             # ,updated_memory0)
-            updated_memory = torch.where(mask.unsqueeze(1),self.gamma*updated_memory0 + (1-self.gamma)*(upd0),updated_memory0)
+            updated_memory = torch.where(mask.unsqueeze(1),torch.sigmoid(self.gamma)*updated_memory0 + (1-torch.sigmoid(self.gamma))*(upd0),updated_memory0)
+            #print('check {} {} {} {}'.format(self.gamma,updated_memory,torch.isnan(updated_memory).sum(),torch.isnan(updated_memory0).sum()))
+            #print(updated_memory,isnan)
             with torch.no_grad():
                 if self.mode == 'historical':
                     change = updated_memory[mask] - b.srcdata['his_mem'][mask]