From a74f32201083122a0be702255dbe5b00bc9a837f Mon Sep 17 00:00:00 2001
From: Mrtutu <zhangwei3.0@qq.com>
Date: Tue, 23 Jan 2024 00:34:42 +0800
Subject: [PATCH] fix kernel bug

---
 .../cluster_advice/kernel_cluster_advice.py     | 17 +++++++++++++++--
 .../common_func_advisor/constant.py             |  1 +
 2 files changed, 16 insertions(+), 2 deletions(-)

diff --git a/profiler/advisor/advisor_backend/cluster_advice/kernel_cluster_advice.py b/profiler/advisor/advisor_backend/cluster_advice/kernel_cluster_advice.py
index 587c40f949..644d46714e 100644
--- a/profiler/advisor/advisor_backend/cluster_advice/kernel_cluster_advice.py
+++ b/profiler/advisor/advisor_backend/cluster_advice/kernel_cluster_advice.py
@@ -20,7 +20,12 @@ class KernelClusterAdvice(ClusterAdviceBase):
         return self.calculate_data()
 
     def load_kernel_details_data(self):
-        data_map = PytorchDataPreprocessor(self.collection_path).get_data_map()
+        prof_dirs = self.get_prof_dirs(self.collection_path)
+        if not prof_dirs:
+            msg = "[ERROR] There is no profile in this collection path, terminate analysis."
+            raise RuntimeError(msg)
+
+        data_map = PytorchDataPreprocessor(prof_dirs).get_data_map()
         self.all_kernel_data = pd.DataFrame()
         for rank_id, profiling_dir_path in data_map.items():
             kernel_file = os.path.join(profiling_dir_path, Constant.SINGLE_OUTPUT, Constant.KERNEL_DETAILS_CSV)
@@ -36,7 +41,7 @@ class KernelClusterAdvice(ClusterAdviceBase):
                 df = df_temp[columns_to_keep]
                 df.insert(loc=0, column='rank id', value=rank_id)
                 # 将数据添加到最终的数据框中
-                self.all_kernel_data = self.all_kernel_data._append(df, ignore_index=True)
+                self.all_kernel_data = pd.concat([self.all_kernel_data, df], ignore_index=True)
 
     def calculate_data(self):
         # 存储所有合并后的数据
@@ -46,3 +51,11 @@ class KernelClusterAdvice(ClusterAdviceBase):
         view_data = self.all_kernel_data.groupby(group_col).agg(calculate_dict).reset_index()
         view_data.columns = [''.join(col) if col[1] == "" else '_'.join(col) for col in view_data.columns]
         return view_data
+
+    def get_prof_dirs(self, collection_path):
+        prof_dirs = []
+        for prof_dir in os.listdir(collection_path):
+            if prof_dir.endswith(Constant.PT_PROF_SUFFIX):
+                prof_dirs.append(os.path.join(collection_path, prof_dir))
+
+        return prof_dirs
\ No newline at end of file
diff --git a/profiler/advisor/advisor_backend/common_func_advisor/constant.py b/profiler/advisor/advisor_backend/common_func_advisor/constant.py
index 2267d69dbc..932edb0a7f 100644
--- a/profiler/advisor/advisor_backend/common_func_advisor/constant.py
+++ b/profiler/advisor/advisor_backend/common_func_advisor/constant.py
@@ -38,6 +38,7 @@ class Constant:
     OPTIM = "optimizer"
     OP_SCHE = "op_schedule"
 
+    PT_PROF_SUFFIX = "ascend_pt"
     COLLECTION_PATH = "collection_path"
     CLUSTER_ANALYSIS_OUTPUT = "cluster_analysis_output"
     CLUSTER_STEP_TIME_CSV = "cluster_step_trace_time.csv"
-- 
Gitee