fix update state in eval

bee5db57 · zhljJoan · a1d8044f · bee5db57 · bee5db57 · bee5db57
Commit bee5db57 authored Mar 15, 2025 by zhljJoan
Hide whitespace changes
Inline Side-by-side

Showing with 125 additions and 6 deletions

examples/draw_historical.py
+118 -0

examples/his.png
+0 -0

examples/train_boundery.py
+0 -1

starrygl/module/utils.py
+7 -5

No files found.
--- a/examples/draw_historical.py
+++ b/examples/draw_historical.py
+import matplotlib.pyplot as plt
+import re
+import os
+# 数据
+dict = {'LASTFM': {
+    1:0.8963 ,
+    0.1:0.9301,
+    0.05:0.9293 ,
+    0.01:0.9208,
+    0:0.8134 
+},
+'WikiTalk' : {
+    1:0.9756,0.1: 	0.9752 	,0.05:0.9798 	,0.01:0.9797 	,0:0.9674 
+}
+}
+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
+    if not os.path.exists(file_name):
+        return 0
+    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):
+    if not os.path.exists(file_name):
+        return 0
+    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[0]))
+    return sum(ap)/4
+# 创建新的图像和子图
+fig, axs = plt.subplots(1,4, figsize=(18, 4))
+# 设置全局标题
+#fig.suptitle('Test AP vs topK', fontsize=24, fontweight='bold')
+# 绘制LASTFM折线图
+color = ['b','g','r','c']
+for i,data in enumerate(['LASTFM','WikiTalk','StackOverflow','GDELT']):
+    model = 'TGN' if data == 'LASTFM' else 'TGN_large'
+    neighbor = 10 if model == 'TGN' else 20 
+    his = [0.1,0.3,0.5,0.7,0.9,1.3,1.5]#1.7,2]
+    ap = [ readap(f"all_12345_new_gamma_set_para/{data}/{model}/8-ours-0.1-historical-{p}-boundery_recent_decay-{0.1}-{neighbor}.out") for p in his]
+    #ap = [dict[data][p] if ap_ == 0 and data in ['LASTFM','WikiTalk'] else ap_ for ap_,p in zip(ap_new,p_list)]
+    print(ap)
+    #type='recent' if p == 1 else "decay"
+    axs[i].plot(his, ap, marker='o', linestyle='-',   linewidth=3,color = color[i],label=f'{data}')
+    #axs[i].legend()
+    axs[i].set_title(f'{data}', fontsize=20, fontweight='bold')
+    axs[i].set_xlabel(r'$\theta$', fontsize=16)
+    axs[i].set_ylabel('AP', fontsize=16)
+    axs[i].grid(True)
+plt.tight_layout()
+plt.subplots_adjust(top=0.92)
+print('save')
+plt.savefig('his.png')
--- a/examples/his.png
+++ b/examples/his.png
--- a/examples/train_boundery.py
+++ b/examples/train_boundery.py
@@ -496,7 +496,6 @@ def main():
            ones = torch.ones(metadata['dst_neg_index'].shape[0],device = model.device,dtype=torch.float)
            #with autocast():
            pred_pos, pred_neg = model(mfgs,metadata,neg_samples=args.neg_samples,async_param = param)
-            ada_param.update_gnn_aggregate_time(ada_param.last_start_event_gnn_aggregate)
            #print(time_count.elapsed_event(t2))
            loss = creterion(pred_pos, torch.ones_like(pred_pos)) 
            if args.local_neg_sample is False:

--- a/starrygl/module/utils.py
+++ b/starrygl/module/utils.py
@@ -80,7 +80,7 @@ class AdaParameter:
    start_event.record()
    return start_event
  def update_fetch_time(self,start_event):
-    if start_event is None:
+    if start_event is None or self.training:
      return
    end_event = torch.cuda.Event(enable_timing=True)
    end_event.record()
@@ -93,7 +93,7 @@ class AdaParameter:
  def update_memory_sync_time(self,start_event):
-    if start_event is None:
+    if start_event is None or self.training == False:
      return
    end_event = torch.cuda.Event(enable_timing=True)
    end_event.record()
@@ -104,7 +104,7 @@ class AdaParameter:
    self.end_event_memory_sync = (self.last_start_event_memory_sync,end_event)
  def update_memory_update_time(self,start_event):
-    if start_event is None:
+    if start_event is None or self.training == False:
      return
    end_event = torch.cuda.Event(enable_timing=True)
    end_event.record()
@@ -116,7 +116,7 @@ class AdaParameter:
  def update_gnn_aggregate_time(self,start_event):
-    if start_event is None:
+    if start_event is None or self.training == False:
      return
    end_event = torch.cuda.Event(enable_timing=True)
    end_event.record()
@@ -140,6 +140,8 @@ class AdaParameter:
    #print('beta is {} alpha is {}\n'.format(self.beta,self.alpha))
    #if self.count_fetch == 0 or self.count_memory_sync == 0 or self.count_memory_update == 0 or self.count_gnn_aggregate == 0:
    #  return
+    if self.training == False:
+      return
    if self.end_event_fetch is None or self.end_event_memory_sync is None or self.end_event_memory_update is None or self.end_event_gnn_aggregate is None:
      return
    self.end_event_fetch[1].synchronize()
@@ -160,7 +162,7 @@ class AdaParameter:
    average_memory_sync_time = self.average_memory_sync/self.count_memory_sync
    average_memory_update_time = self.average_memory_update/self.count_memory_update
    self.alpha = self.alpha - math.log(average_memory_update_time*(1+self.wait_threshold)) + math.log(average_memory_sync_time)
-    print(self.alpha)                                                                               
+    #print(self.alpha)                                                                               
    self.beta = max(min(self.beta, self.max_beta),self.min_beta)
    self.alpha = max(min(self.alpha, self.max_alpha),self.min_alpha)