fix bug in sharedmemory (#3229)

* fix bug in sharedmemory
* fix bug in sharedmemory

ppdet/data/data_feed.py

@@ -90,18 +90,15 @@ def create_reader(feed, max_iter=0, args_path=None, my_source=None):
     # named `DATASET_DIR` (e.g., coco, pascal), if not present either, download
     data_config = _prepare_data_config(feed, args_path)
 
-    bufsize = 10
-    use_process = False
-    if getattr(feed, 'bufsize', None) is not None:
-        bufsize = feed.bufsize
-    if getattr(feed, 'use_process', None) is not None:
-        use_process = feed.use_process
-
+    bufsize = getattr(feed, 'bufsize', 10)
+    use_process = getattr(feed, 'use_process', False)
+    memsize = getattr(feed, 'memsize', '3G')
     transform_config = {
         'WORKER_CONF': {
             'bufsize': bufsize,
             'worker_num': feed.num_workers,
-            'use_process': use_process
+            'use_process': use_process,
+            'memsize': memsize
         },
         'BATCH_SIZE': feed.batch_size,
         'DROP_LAST': feed.drop_last,
@@ -282,6 +279,10 @@ class DataFeed(object):
         shuffle (bool): if samples should be shuffled
         drop_last (bool): drop last batch if size is uneven
         num_workers (int): number of workers processes (or threads)
+        bufsize (int): size of queue used to buffer results from workers
+        use_process (bool): use process or thread as workers
+        memsize (str): size of shared memory used in result queue
+                        when 'use_process' is True, default to '3G'
     """
     __category__ = 'data'
 
@@ -299,6 +300,7 @@ class DataFeed(object):
                  num_workers=2,
                  bufsize=10,
                  use_process=False,
+                 memsize=None,
                  use_padded_im_info=False):
         super(DataFeed, self).__init__()
         self.fields = fields
@@ -313,6 +315,7 @@ class DataFeed(object):
         self.num_workers = num_workers
         self.bufsize = bufsize
         self.use_process = use_process
+        self.memsize = memsize
         self.dataset = dataset
         self.use_padded_im_info = use_padded_im_info
         if isinstance(dataset, dict):
@@ -337,7 +340,8 @@ class TrainFeed(DataFeed):
                  with_background=True,
                  num_workers=2,
                  bufsize=10,
-                 use_process=True):
+                 use_process=True,
+                 memsize=None):
         super(TrainFeed, self).__init__(
             dataset,
             fields,
@@ -351,7 +355,8 @@ class TrainFeed(DataFeed):
             with_background=with_background,
             num_workers=num_workers,
             bufsize=bufsize,
-            use_process=use_process, )
+            use_process=use_process,
+            memsize=memsize)
 
 
 @register
@@ -439,8 +444,10 @@ class FasterRCNNTrainFeed(DataFeed):
                  shuffle=True,
                  samples=-1,
                  drop_last=False,
+                 bufsize=10,
                  num_workers=2,
-                 use_process=False):
+                 use_process=False,
+                 memsize=None):
         # XXX this should be handled by the data loader, since `fields` is
         # given, just collect them
         sample_transforms.append(ArrangeRCNN())
@@ -454,8 +461,10 @@ class FasterRCNNTrainFeed(DataFeed):
             shuffle=shuffle,
             samples=samples,
             drop_last=drop_last,
+            bufsize=bufsize,
             num_workers=num_workers,
-            use_process=use_process)
+            use_process=use_process,
+            memsize=memsize)
         # XXX these modes should be unified
         self.mode = 'TRAIN'
 
@@ -722,7 +731,8 @@ class SSDTrainFeed(DataFeed):
                  drop_last=True,
                  num_workers=8,
                  bufsize=10,
-                 use_process=True):
+                 use_process=True,
+                 memsize=None):
         sample_transforms.append(ArrangeSSD())
         super(SSDTrainFeed, self).__init__(
             dataset,
@@ -736,7 +746,8 @@ class SSDTrainFeed(DataFeed):
             drop_last=drop_last,
             num_workers=num_workers,
             bufsize=bufsize,
-            use_process=use_process)
+            use_process=use_process,
+            memsize=None)
         self.mode = 'TRAIN'
 
 
@@ -767,7 +778,8 @@ class SSDEvalFeed(DataFeed):
             drop_last=True,
             num_workers=8,
             bufsize=10,
-            use_process=False):
+            use_process=False,
+            memsize=None):
         sample_transforms.append(ArrangeEvalSSD())
         super(SSDEvalFeed, self).__init__(
             dataset,
@@ -781,7 +793,8 @@ class SSDEvalFeed(DataFeed):
             drop_last=drop_last,
             num_workers=num_workers,
             bufsize=bufsize,
-            use_process=use_process)
+            use_process=use_process,
+            memsize=memsize)
         self.mode = 'VAL'
 
 
@@ -809,7 +822,8 @@ class SSDTestFeed(DataFeed):
                  drop_last=False,
                  num_workers=8,
                  bufsize=10,
-                 use_process=False):
+                 use_process=False,
+                 memsize=None):
         sample_transforms.append(ArrangeTestSSD())
         if isinstance(dataset, dict):
             dataset = SimpleDataSet(**dataset)
@@ -825,7 +839,8 @@ class SSDTestFeed(DataFeed):
             drop_last=drop_last,
             num_workers=num_workers,
             bufsize=bufsize,
-            use_process=use_process)
+            use_process=use_process,
+            memsize=memsize)
         self.mode = 'TEST'
 
 
@@ -873,6 +888,7 @@ class YoloTrainFeed(DataFeed):
                  num_workers=8,
                  bufsize=128,
                  use_process=True,
+                 memsize=None,
                  num_max_boxes=50,
                  mixup_epoch=250):
         sample_transforms.append(ArrangeYOLO())
@@ -889,7 +905,8 @@ class YoloTrainFeed(DataFeed):
             with_background=with_background,
             num_workers=num_workers,
             bufsize=bufsize,
-            use_process=use_process)
+            use_process=use_process,
+            memsize=memsize)
         self.num_max_boxes = num_max_boxes
         self.mixup_epoch = mixup_epoch
         self.mode = 'TRAIN'
@@ -923,7 +940,8 @@ class YoloEvalFeed(DataFeed):
                  with_background=False,
                  num_workers=8,
                  num_max_boxes=50,
-                 use_process=False):
+                 use_process=False,
+                 memsize=None):
         sample_transforms.append(ArrangeEvalYOLO())
         super(YoloEvalFeed, self).__init__(
             dataset,
@@ -937,7 +955,8 @@ class YoloEvalFeed(DataFeed):
             drop_last=drop_last,
             with_background=with_background,
             num_workers=num_workers,
-            use_process=use_process)
+            use_process=use_process,
+            memsize=memsize)
         self.num_max_boxes = num_max_boxes
         self.mode = 'VAL'
         self.bufsize = 128
@@ -976,7 +995,8 @@ class YoloTestFeed(DataFeed):
                  with_background=False,
                  num_workers=8,
                  num_max_boxes=50,
-                 use_process=False):
+                 use_process=False,
+                 memsize=None):
         sample_transforms.append(ArrangeTestYOLO())
         if isinstance(dataset, dict):
             dataset = SimpleDataSet(**dataset)
@@ -992,7 +1012,8 @@ class YoloTestFeed(DataFeed):
             drop_last=drop_last,
             with_background=with_background,
             num_workers=num_workers,
-            use_process=use_process)
+            use_process=use_process,
+            memsize=memsize)
         self.mode = 'TEST'
         self.bufsize = 128
 

ppdet/data/tests/coco.yml

@@ -29,7 +29,11 @@ TRANSFORM:
         BATCH_SIZE: 1
         IS_PADDING: True
         DROP_LAST: False
-        
+        WORKER_CONF:
+            BUFSIZE: 100
+            WORKER_NUM: 4
+            USE_PROCESS: True
+            MEMSIZE: 2G
     VAL:
         OPS:
             - OP: DecodeImage
@@ -39,6 +43,6 @@ TRANSFORM:
             - OP: ArrangeSSD
         BATCH_SIZE: 1
         WORKER_CONF:
-        BUFSIZE: 200
-        WORKER_NUM: 8
-        USE_PROCESS: False
+            BUFSIZE: 100
+            WORKER_NUM: 4
+            USE_PROCESS: True

ppdet/data/tests/rcnn_dataset.yml

@@ -26,5 +26,7 @@ TRANSFORM:
         IS_PADDING: True
         DROP_LAST: False
         WORKER_CONF:
-        BUFSIZE: 10
-        WORKER_NUM: 2
+            BUFSIZE: 100
+            WORKER_NUM: 4
+            MEMSIZE: 2G
+            USE_PROCESS: True

ppdet/data/transform/parallel_map.py

@@ -49,8 +49,15 @@ class ParallelMappedDataset(ProxiedDataset):
         super(ParallelMappedDataset, self).__init__(source)
         worker_args = {k.lower(): v for k, v in worker_args.items()}
 
-        args = {'bufsize': 100, 'worker_num': 8}
+        args = {'bufsize': 100, 'worker_num': 8,
+            'use_process': False, 'memsize': '3G'}
         args.update(worker_args)
+        if args['use_process'] and type(args['memsize']) is str:
+            assert args['memsize'][-1].lower() == 'g', \
+                "invalid param for memsize[%s], should be ended with 'G' or 'g'" % (args['memsize'])
+            gb = args['memsize'][:-1]
+            args['memsize'] = int(gb) * 1024 ** 3
+
         self._worker_args = args
         self._started = False
         self._source = source
@@ -60,9 +67,7 @@ class ParallelMappedDataset(ProxiedDataset):
 
     def _setup(self):
         """setup input/output queues and workers """
-        use_process = False
-        if 'use_process' in self._worker_args:
-            use_process = self._worker_args['use_process']
+        use_process = self._worker_args.get('use_process', False)
         if use_process and sys.platform == "win32":
             logger.info("Use multi-thread reader instead of "
                         "multi-process reader on Windows.")
@@ -73,6 +78,9 @@ class ParallelMappedDataset(ProxiedDataset):
             from .shared_queue import SharedQueue as Queue
             from multiprocessing import Process as Worker
             from multiprocessing import Event
+            memsize = self._worker_args['memsize']
+            self._inq = Queue(bufsize, memsize=memsize)
+            self._outq = Queue(bufsize, memsize=memsize)
         else:
             if six.PY3:
                 from queue import Queue
@@ -80,11 +88,10 @@ class ParallelMappedDataset(ProxiedDataset):
                 from Queue import Queue
             from threading import Thread as Worker
             from threading import Event
-
             self._inq = Queue(bufsize)
             self._outq = Queue(bufsize)
-        consumer_num = self._worker_args['worker_num']
 
+        consumer_num = self._worker_args['worker_num']
         id = str(uuid.uuid4())[-3:]
         self._producer = threading.Thread(
             target=self._produce,

ppdet/data/transform/shared_queue/sharedmemory.py

@@ -318,16 +318,14 @@ class PageAllocator(object):
         while True:
             # maybe flags already has some '0' pages,
             # so just check 'page_num - len(flags)' pages
-            flags += self.get_page_status(
-                pos, page_num - len(flags), ret_flag=True)
+            flags = self.get_page_status(
+                pos, page_num, ret_flag=True)
 
             if flags.count('0') == page_num:
                 break
 
             # not found enough pages, so shift to next few pages
             free_pos = flags.rfind('1') + 1
-            flags = flags[free_pos:]
-
             pos += free_pos
             end = pos + page_num
             if end > pages:
@@ -355,9 +353,6 @@ class PageAllocator(object):
         self.set_page_status(pos, page_num, '1')
         used += page_num
         self.set_alloc_info(end, used)
-
-        assert self.get_page_status(pos, page_num) == (page_num, 1), \
-            'faild to validate the page status'
         return pos
 
     def free_page(self, start, page_num):
@@ -530,7 +525,7 @@ class SharedMemoryMgr(object):
             logger.info('destroy [%s]' % (self))
 
         if not self._released and not self._allocator.empty():
-            logger.warn('not empty when delete this SharedMemoryMgr[%s]' %
+            logger.debug('not empty when delete this SharedMemoryMgr[%s]' %
                         (self))
         else:
             self._released = True