add stream shuffle

pgl/utils/data/dataloader.py

@@ -14,7 +14,7 @@
 # limitations under the License.
 """dataloader
 """
-
+import warnings
 import numpy as np
 from collections import namedtuple
 
@@ -64,15 +64,15 @@ class Dataloader(object):
                 print(batch_data)
     """
 
-    def __init__(
-            self,
+    def __init__(self,
                  dataset,
                  batch_size=1,
                  drop_last=False,
                  shuffle=False,
                  num_workers=1,
                  collate_fn=None,
-            buf_size=1000, ):
+                 buf_size=1000,
+                 stream_shuffle_size=0):
 
         self.dataset = dataset
         self.batch_size = batch_size
@@ -81,6 +81,22 @@ class Dataloader(object):
         self.collate_fn = collate_fn
         self.buf_size = buf_size
         self.drop_last = drop_last
+        self.stream_shuffle_size = stream_shuffle_size
+
+        if self.shuffle and isinstance(self.dataset, StreamDataset):
+            warn_msg = "[shuffle] should not be True with StreamDataset. " \
+                    "It will be ignored. " \
+                    "You might want to set [stream_shuffle_size] with StreamDataset."
+            warnings.warn(warn_msg)
+
+        if self.stream_shuffle_size > 0 and self.batch_size >= stream_shuffle_size:
+            raise ValueError("stream_shuffle_size must be larger than batch_size," \
+                    "but got [stream_shuffle_size=%s] smaller than [batch_size=%s]" \
+                    % (self.stream_shuffle_size, self.batch_size))
+
+        if self.num_workers < 1:
+            raise ValueError("num_workers(default: 1) should be larger than 0, " \
+                        "but got [num_workers=%s] < 1." % self.num_workers)
 
     def __len__(self):
         if not isinstance(self.dataset, StreamDataset):
@@ -129,6 +145,8 @@ class _DataLoaderIter(object):
         self.collate_fn = dataloader.collate_fn
         self.num_workers = dataloader.num_workers
         self.drop_last = dataloader.drop_last
+        self.batch_size = dataloader.batch_size
+        self.stream_shuffle_size = dataloader.stream_shuffle_size
         self.fid = fid
         self.count = 0
 
@@ -167,16 +185,62 @@ class _DataLoaderIter(object):
 
             # make sure do not repeat in multiprocessing 
             self.count += 1
-            #  if self.count % self.num_workers != self.fid:
-            #  continue
 
             if self.collate_fn is not None:
                 yield self.collate_fn(batch_data)
             else:
                 yield batch_data
 
+    def _stream_shuffle_data_generator(self):
+        def _stream_shuffle_index_generator():
+            shuffle_size = [i for i in range(self.stream_shuffle_size)]
+            while True:
+                yield shuffle_size
+
+        def _data_generator():
+            dataset = iter(self.dataset)
+            for shuffle_size in _stream_shuffle_index_generator():
+                shuffle_size_data = []
+                for idx in shuffle_size:
+                    try:
+                        shuffle_size_data.append(next(dataset))
+                    except StopIteration:
+                        break
+
+                if len(shuffle_size_data) == 0:
+                    break
+
+                yield shuffle_size_data
+
+        def _batch_data_generator():
+            batch_data = []
+            for shuffle_size_data in _data_generator():
+                np.random.shuffle(shuffle_size_data)
+
+                for d in shuffle_size_data:
+                    batch_data.append(d)
+                    if len(batch_data) == self.batch_size:
+                        yield batch_data
+                        batch_data = []
+
+            if not self.drop_last and len(batch_data) > 0:
+                yield batch_data
+
+        self._worker_info = WorkerInfo(
+            num_workers=self.num_workers, fid=self.fid)
+        self.dataset._set_worker_info(self._worker_info)
+
+        for batch_data in _batch_data_generator():
+            if self.collate_fn is not None:
+                yield self.collate_fn(batch_data)
+            else:
+                yield batch_data
+
     def __iter__(self):
         if isinstance(self.dataset, StreamDataset):
+            if self.stream_shuffle_size > 0:
+                data_generator = self._stream_shuffle_data_generator
+            else:
                 data_generator = self._streamdata_generator
         else:
             data_generator = self._data_generator