!632 Detect a summary dir having summary files or not

Merge pull request !632 from LiHongzhang/fix_caching

mindinsight/datavisual/data_transform/data_manager.py

@@ -50,29 +50,18 @@ class _BasicTrainJob:
     Basic info about train job.
 
     Args:
-        train_id (str): Id of the train job.
         abs_summary_base_dir (str): The canonical path of summary base directory. It should be the return value of
             realpath().
-        abs_summary_dir (str): The canonical path of summary directory. It should be the return value of realpath().
-        create_time (DateTime): The create time of summary directory.
-        update_time (DateTime): The latest modify time of summary files directly in the summary directory.
-        profiler_dir (str): The relative path of profiler directory.
-        profiler_type (str): The profiler device type.
+        entry (dict): The summary dir entry listed by SummaryWatcher.
     """
-    def __init__(self, train_id, abs_summary_base_dir, abs_summary_dir, create_time, update_time, profiler_dir,
-                 profiler_type=""):
-        self._train_id = train_id
+    def __init__(self, abs_summary_base_dir, entry):
         self._abs_summary_base_dir = abs_summary_base_dir
-        self._abs_summary_dir = abs_summary_dir
-        self._create_time = create_time
-        self._update_time = update_time
-        self._profiler_dir = profiler_dir
-        self._profiler_type = profiler_type
+        self._entry = entry
 
     @property
     def abs_summary_dir(self):
         """Get summary directory path."""
-        return self._abs_summary_dir
+        return os.path.realpath(os.path.join(self._abs_summary_base_dir, self._entry['relative_path']))
 
     @property
     def summary_base_dir(self):
@@ -82,27 +71,46 @@ class _BasicTrainJob:
     @property
     def train_id(self):
         """Get train id."""
-        return self._train_id
+        return self._entry['relative_path']
 
     @property
     def profiler_dir(self):
         """Get profiler directory path."""
-        return self._profiler_dir
+        if self._entry['profiler'] is not None:
+            return self._entry['profiler']['directory']
+        return None
 
     @property
     def create_time(self):
         """Get create time."""
-        return self._create_time
+        return self._entry['create_time']
 
     @property
     def update_time(self):
         """Get update time."""
-        return self._update_time
+        return self._entry['update_time']
 
     @property
     def profiler_type(self):
         """Get profiler type"""
-        return self._profiler_type
+        if self._entry['profiler'] is not None:
+            return self._entry['profiler']['profiler_type']
+        return ''
+
+    @property
+    def summary_files(self):
+        """Get the summary files count in the summary dir."""
+        return self._entry['summary_files']
+
+    @property
+    def graph_files(self):
+        """Get the graph pb files count in the summary dir."""
+        return self._entry['graph_files']
+
+    @property
+    def lineage_files(self):
+        """Get the lineage files count in the summary dir."""
+        return self._entry['lineage_files']
 
 
 class CachedTrainJob:
@@ -377,18 +385,9 @@ class _BriefCacheManager(_BaseCacheManager):
 
         basic_train_jobs = []
         for info in summaries_info:
-            profiler = info['profiler']
             basic_train_jobs.append(_BasicTrainJob(
-                train_id=info['relative_path'],
                 abs_summary_base_dir=self._summary_base_dir,
-                abs_summary_dir=os.path.realpath(os.path.join(
-                    self._summary_base_dir,
-                    info['relative_path']
-                )),
-                create_time=info['create_time'],
-                update_time=info['update_time'],
-                profiler_dir=None if profiler is None else profiler['directory'],
-                profiler_type="" if profiler is None else profiler['profiler_type'],
+                entry=info
             ))
 
         with self._lock:

mindinsight/datavisual/data_transform/summary_watcher.py

@@ -26,6 +26,7 @@ from mindinsight.datavisual.utils.utils import contains_null_byte
 from mindinsight.datavisual.common.exceptions import MaxCountExceededError
 from mindinsight.utils.exceptions import FileSystemPermissionError
 
+LINEAGE_SUMMARY_SUFFIX = '_lineage'
 
 class SummaryWatcher:
     """SummaryWatcher class."""
@@ -99,18 +100,8 @@ class SummaryWatcher:
         for key, value in summary_dict.items():
             directory = {
                 'relative_path': key,
-                'profiler': None,
-                'create_time': value['ctime'],
-                'update_time': value['mtime'],
+                **value
             }
-            profiler = value.get('profiler')
-            if profiler is not None:
-                directory['profiler'] = {
-                    'directory': profiler['directory'],
-                    'create_time': profiler['ctime'],
-                    'update_time': profiler['mtime'],
-                    'profiler_type': profiler['profiler_type']
-                }
             directories.append(directory)
 
         # sort by update time in descending order and relative path in ascending order
@@ -215,15 +206,24 @@ class SummaryWatcher:
                     return
             if relative_path not in summary_dict:
                 summary_dict[relative_path] = {
-                    'ctime': ctime,
-                    'mtime': mtime,
+                    'create_time': ctime,
+                    'update_time': mtime,
+                    'summary_files': 0,
+                    'lineage_files': 0,
+                    'graph_files': 0,
                     'profiler': None,
                 }
-            elif summary_dict[relative_path]['ctime'] < ctime:
+            if summary_dict[relative_path]['create_time'] < ctime:
                 summary_dict[relative_path].update({
-                    'ctime': ctime,
-                    'mtime': mtime,
+                    'create_time': ctime,
+                    'update_time': mtime,
                 })
+            if not summary_pattern:
+                summary_dict[relative_path]['graph_files'] += 1
+            elif entry.name.endswith(LINEAGE_SUMMARY_SUFFIX):
+                summary_dict[relative_path]['lineage_files'] += 1
+            else:
+                summary_dict[relative_path]['summary_files'] += 1
         elif entry.is_dir():
             profiler_pattern = re.search(self.PROFILER_DIRECTORY_REGEX, entry.name)
             full_dir_path = os.path.join(summary_base_dir, relative_path, entry.name)
@@ -233,16 +233,22 @@ class SummaryWatcher:
 
             profiler = {
                 'directory': os.path.join('.', entry.name),
-                'ctime': ctime,
-                'mtime': mtime,
+                'create_time': ctime,
+                'update_time': mtime,
                 "profiler_type": profiler_type
             }
 
-            summary_dict[relative_path] = {
-                'ctime': ctime,
-                'mtime': mtime,
-                'profiler': profiler,
-            }
+            if relative_path in summary_dict:
+                summary_dict[relative_path]['profiler'] = profiler
+            else:
+                summary_dict[relative_path] = {
+                    'create_time': ctime,
+                    'summary_files': 0,
+                    'lineage_files': 0,
+                    'graph_files': 0,
+                    'update_time': mtime,
+                    'profiler': profiler
+                }
 
     def is_summary_directory(self, summary_base_dir, relative_path):
         """

mindinsight/datavisual/processors/train_task_manager.py

@@ -145,6 +145,9 @@ class TrainTaskManager(BaseProcessor):
             profiler_dir=basic_info.profiler_dir,
             cache_status=train_job.cache_status.value,
             profiler_type=basic_info.profiler_type,
+            summary_files=basic_info.summary_files,
+            graph_files=basic_info.graph_files,
+            lineage_files=basic_info.lineage_files
         )
 
         if train_job.cache_status == CacheStatus.CACHED:

mindinsight/ui/src/locales/en-us.json

@@ -51,6 +51,8 @@
     "dataTraceback": "Dataset Lineage",
     "comparePlate": "Comparison Dashboard",
     "disableProfilerTip": "Failed to view profiling because no profiler log is available.",
+    "disableDashboardTip": "Failed to view training dashboard because no summary log or pb files are available.",
+    "disableParameterTip": "Failed to view parameter details because no lineage log is available.",
     "openNewTab": "Open Link in New Tab",
     "paramDetails": "Parameter Details",
     "trainingParamDetails": "Training Parameter Details",

mindinsight/ui/src/locales/zh-cn.json

@@ -51,6 +51,8 @@
     "dataTraceback": "数据溯源",
     "comparePlate": "对比看板",
     "disableProfilerTip": "无profiler日志，无法查看性能分析",
+    "disableDashboardTip": "无summary日志或pb文件，无法查看训练看板",
+    "disableParameterTip": "无lineage日志，无法查看参数详情",
     "openNewTab": "打开新页签",
     "paramDetails": "参数详情",
     "trainingParamDetails": "训练参数详情",

mindinsight/ui/src/views/train-manage/summary-manage.vue

@@ -74,10 +74,16 @@ limitations under the License.
                              class-name="operate-container"
                              width="400">
               <template slot-scope="scope">
-                <span class="menu-item"
+                <span class="menu-item operate-btn"
+                      v-if="scope.row.viewDashboard"
                       @contextmenu.prevent="rightClick(scope.row, $event, 0)"
                       @click.stop="goToTrainDashboard(scope.row)">
                   {{$t('summaryManage.viewDashboard')}} </span>
+                <span class="menu-item operate-btn button-disable"
+                      v-else
+                      :title="$t('summaryManage.disableDashboardTip')">
+                  {{$t('summaryManage.viewDashboard')}}
+                </span>
                 <span class="menu-item operate-btn"
                       v-if="scope.row.viewProfiler"
                       @contextmenu.prevent="rightClick(scope.row, $event, 1)"
@@ -89,9 +95,14 @@ limitations under the License.
                   {{$t('summaryManage.viewProfiler')}}
                 </span>
                 <span class="menu-item operate-btn"
+                      v-if="scope.row.paramDetails"
                       @click.stop="showModelDialog(scope.row)">
                   {{$t('summaryManage.paramDetails')}} </span>
-
+                <span class="menu-item operate-btn button-disable"
+                      v-else
+                      :title="$t('summaryManage.disableParameterTip')">
+                  {{$t('summaryManage.paramDetails')}}
+                </span>
               </template>
             </el-table-column>
           </el-table>
@@ -267,6 +278,8 @@ export default {
                     i.relative_path = i.relative_path ? i.relative_path : '--';
                     i.update_time = i.update_time ? i.update_time : '--';
                     i.viewProfiler = i.profiler_dir && i.profiler_dir.length;
+                    i.viewDashboard = i.summary_files || i.graph_files || i.lineage_files;
+                    i.paramDetails = i.lineage_files;
                   });
                   this.currentFolder = res.data.name ? res.data.name : '--';
                   this.pagination.total = res.data.total;