fix gamma by sigmoid

examples/draw_topk.py

+import matplotlib.pyplot as plt
+import re
+# 数据
+k_values = [0.02, 0.04, 0.06, 0.08, 0.1, 0.2, 0.3]
+dict = {'LASTFM': [0.903726, 0.921197, 0.931237, 0.926789, 0.930719, 0.929332, 0.915848],
+'WikiTalk' : [0.981577, 0.980716, 0.979996, 0.979597, 0.979248, 0.975468, 0.972785],
+'StackOverflow' : [0.974805, 0.978219, 0.97924, 0.979436, 0.979456, 0.976746, 0.972544]}
+def read(file_name):
+
+# 定义正则表达式模式
+    pattern = re.compile(r"time_count\.time_forward=(\d+\.\d+).*?time_count\.time_backward=(\d+\.\d+).*?time_count\.time_memory_updater=(\d+\.\d+).*?time_count\.time_embedding=(\d+\.\d+).*?time_count\.time_local_update=(\d+\.\d+).*?time_count\.time_memory_sync=(\d+\.\d+).*?time_count\.time_sample_and_build=(\d+\.\d+).*?time_count\.time_memory_fetch=(\d+\.\d+)")
+    pattern = re.compile(r"time_count\.time_forward=(\d+\.\d+).*?time_count\.time_backward=(\d+\.\d+).*?time_count\.time_memory_updater=(\d+\.\d+).*?time_count\.time_embedding=(\d+\.\d+).*?time_count\.time_local_update=(\d+).*?time_count\.time_memory_sync=(\d+\.\d+).*?time_count\.time_sample_and_build=(\d+\.\d+).*?time_count\.time_memory_fetch=(\d+\.\d+)")
+
+    # 读取日志文件
+    with open(file_name, 'r') as file:
+        log_content = file.read()
+    #print(log_content)
+
+    # 查找所有匹配的行
+    matches = pattern.findall(log_content)
+    # 初始化累加器
+    forward_times = []
+    backward_times = []
+    memory_updater_times = []
+    embedding_times = []
+    local_update_times = []
+    memory_sync_times = []
+    sample_and_build_times = []
+    memory_fetch_times = []
+
+    # 提取数据并累加
+    for match in matches:
+
+        forward_times.append(float(match[0]))
+        backward_times.append(float(match[1]))
+        memory_updater_times.append(float(match[2]))
+        embedding_times.append(float(match[3]))
+        local_update_times.append(float(match[4]))
+        memory_sync_times.append(float(match[5]))
+        sample_and_build_times.append(float(match[6]))
+        memory_fetch_times.append(float(match[7]))
+
+    # 计算平均值
+    def calculate_average(times):
+        return sum(times) / len(times) if times else 0
+
+    average_forward = calculate_average(forward_times)
+    average_backward = calculate_average(backward_times)
+    average_memory_updater = calculate_average(memory_updater_times)
+    average_embedding = calculate_average(embedding_times)
+    average_local_update = calculate_average(local_update_times)
+    average_memory_sync = calculate_average(memory_sync_times)
+    average_sample_and_build = calculate_average    (sample_and_build_times)
+    average_memory_fetch = calculate_average(memory_fetch_times)
+    averages = {
+        'average_forward': average_forward,
+        'average_backward': average_backward,
+        'average_memory_updater': average_memory_updater,
+        'average_embedding': average_embedding,
+        'average_local_update': average_local_update,
+        'average_memory_sync': average_memory_sync,
+        'average_sample_and_build': average_sample_and_build,
+        'average_memory_fetch':average_memory_fetch,
+    }
+    return averages['average_memory_fetch']/1000
+
+def readap(file_name):
+    pattern = re.compile(r"best test AP:([\d.]+) test auc([\d.]+)")
+
+    with open(file_name, 'r') as file:
+        log_content = file.read()
+    ap = []
+    matches = pattern.findall(log_content)
+    for match in matches:
+        print(match)
+        ap.append(float(match[1]))
+    return sum(ap)/4
+# 创建新的图像和子图
+fig, axs = plt.subplots(1,3, figsize=(15, 4))
+
+# 设置全局标题
+#fig.suptitle('Test AP vs topK', fontsize=24, fontweight='bold')
+
+# 绘制LASTFM折线图
+for i,data in enumerate(['LASTFM','StackOverflow']):
+
+
+    model = 'TGN' if data == 'LASTFM' else 'TGN_large'
+    neighbor = 10 if model == 'TGN' else 20 
+    for p in [0.1,0.6,1]:
+        #if p > 0.1:
+            #if p != 1:
+        ap = [ 27.19218282 if data == 'StackOverflow' and p ==0.1 and topk==0.1 else read(f"all_12357/{data}/{model}/8-ours-{topk}-historical-{0.3}-boundery_recent_decay-{p}-{neighbor}.out") for topk in k_values]
+            #else:
+            #    ap = [readap(f"all_12357/{data}/{model}/8-ours-{topk}-historical-{0.3}-recent-{neighbor}.out") for topk in k_values]
+        #else:
+        #    ap = dict[data]
+        print(p,ap)
+        type='recent' if p == 1 else "decay"
+        axs[i].plot(k_values, ap, marker='o', linestyle='-',   linewidth=3,label=f'p={p}({type})')
+    axs[i].legend()
+    axs[i].set_title(f'{data}', fontsize=20, fontweight='bold')
+    axs[i].set_xlabel('Fetch Time(s)', fontsize=16)
+    axs[i].set_ylabel('Test AP', fontsize=16)
+    axs[i].grid(True)
+
+
+
+
+
+plt.tight_layout()
+plt.subplots_adjust(top=0.92)
+plt.savefig('topk_time.png')

examples/test_all.sh

@@ -2,13 +2,13 @@
 #跑了4卡的TaoBao
 # 定义数组变量
 seed=$1
-addr="192.168.1.107"
+addr="192.168.1.105"
 partition_params=("ours")
 #"metis" "ldg" "random")
 #("ours" "metis" "ldg" "random")
-partitions="8"
+partitions="4"
 node_per="4"
-nnodes="2"
+nnodes="1"
 node_rank="0"
 probability_params=("0.1")
 sample_type_params=("boundery_recent_decay")
@@ -17,9 +17,9 @@ sample_type_params=("boundery_recent_decay")
 memory_type=("historical")
 #"historical")
 #memory_type=("local" "all_update" "historical" "all_reduce")
-shared_memory_ssim=("0")
+shared_memory_ssim=("0.3")
 #data_param=("WIKI" "REDDIT" "LASTFM" "WikiTalk")
-data_param=("WikiTalk") 
+data_param=("LASTFM" "WikiTalk" "StackOverflow" "GDELT") 
 #"GDELT")
 #data_param=("WIKI" "REDDIT" "LASTFM" "DGraphFin" "WikiTalk" "StackOverflow")
 #data_param=("WIKI" "REDDIT" "LASTFM" "WikiTalk" "StackOverflow")

starrygl/module/memorys.py

@@ -493,7 +493,7 @@ 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)
             with torch.no_grad():
                 if self.mode == 'historical':
                     change = updated_memory[mask] - b.srcdata['his_mem'][mask]