split long sentence, drop sentence close by -1

fluid/DeepASR/data_utils/data_reader.py

@@ -59,10 +59,19 @@ class SampleInfoBucket(object):
         label_bin_paths (list|tuple): Files containing the binary label data.
         label_desc_paths (list|tuple): Files containing the description of
                                        samples' label data.
+        split_perturb(int): split long sentence' perturb sub-sentence length value. 
+        split_sentence_threshold(int): sentence length large than split_sentence_threshold trigger split operator.
+        split_sub_sentence_len(int): sub-sentence length is equal to (split_sub_sentence_len + rand() % split_perturb).
     """
 
-    def __init__(self, feature_bin_paths, feature_desc_paths, label_bin_paths,
-                 label_desc_paths):
+    def __init__(self,
+                 feature_bin_paths,
+                 feature_desc_paths,
+                 label_bin_paths,
+                 label_desc_paths,
+                 split_perturb=50,
+                 split_sentence_threshold=512,
+                 split_sub_sentence_len=256):
         block_num = len(label_bin_paths)
         assert len(label_desc_paths) == block_num
         assert len(feature_bin_paths) == block_num
@@ -73,6 +82,9 @@ class SampleInfoBucket(object):
         self._feature_desc_paths = feature_desc_paths
         self._label_bin_paths = label_bin_paths
         self._label_desc_paths = label_desc_paths
+        self._split_perturb = split_perturb
+        self._split_sentence_threshold = split_sentence_threshold
+        self._split_sub_sentence_len = split_sub_sentence_len
 
     def generate_sample_info_list(self):
         sample_info_list = []
@@ -100,11 +112,40 @@ class SampleInfoBucket(object):
                 label_size = int(label_desc_split[3])
                 label_frame_num = int(label_desc_split[4])
 
-                sample_info_list.append(
-                    SampleInfo(feature_bin_path, feature_start, feature_size,
-                               feature_frame_num, feature_dim, label_bin_path,
-                               label_start, label_size, label_frame_num))
-
+                if self._split_sentence_threshold == -1 or self._split_perturb == -1 or self._split_sub_sentence_len == -1 or self._split_sentence_threshold >= feature_frame_num:
+                    sample_info_list.append(
+                        SampleInfo(feature_bin_path, feature_start,
+                                   feature_size, feature_frame_num, feature_dim,
+                                   label_bin_path, label_start, label_size,
+                                   label_frame_num))
+                #split sentence 
+                else:
+                    cur_frame_pos = 0
+                    cur_frame_len = 0
+                    remain_frame_num = feature_frame_num
+                    while True:
+                        if remain_frame_num > self._split_sentence_threshold:
+                            cur_frame_len = self._split_sub_sentence_len + random.randint(
+                                0, self._split_perturb)
+                            if cur_frame_len > remain_frame_num:
+                                cur_frame_len = remain_frame_num
+                        else:
+                            cur_frame_len = remain_frame_num
+
+                        sample_info_list.append(
+                            SampleInfo(
+                                feature_bin_path, feature_start + cur_frame_pos
+                                * feature_dim * 4, cur_frame_len * feature_dim *
+                                4, cur_frame_len, feature_dim, label_bin_path,
+                                label_start + cur_frame_pos * 4, cur_frame_len *
+                                4, cur_frame_len))
+
+                        remain_frame_num -= cur_frame_len
+                        cur_frame_pos += cur_frame_len
+                        if remain_frame_num <= 0:
+                            break
+
+        print("generate_sample_info_list size ", len(sample_info_list))
         return sample_info_list
 
 
@@ -143,7 +184,7 @@ class DataReader(object):
                  sample_buffer_size=1024,
                  sample_info_buffer_size=1024,
                  batch_buffer_size=1024,
-                 shuffle_block_num=1,
+                 shuffle_block_num=10,
                  random_seed=0):
         self._feature_file_list = feature_file_list
         self._label_file_list = label_file_list
@@ -260,7 +301,8 @@ class DataReader(object):
                     time.sleep(0.001)
 
                 # drop long sentence
-                if self._drop_frame_len >= sample_data[0].shape[0]:
+                if self._drop_frame_len == -1 or self._drop_frame_len >= sample_data[
+                        0].shape[0]:
                     sample_queue.put(sample_data)
 
                 out_order[0] += 1
@@ -281,7 +323,6 @@ class DataReader(object):
             w.start()
 
         finished_process_num = 0
-
         while finished_process_num < self._process_num:
             sample = sample_queue.get()
             if isinstance(sample, EpochEndSignal):

fluid/DeepASR/stacked_dynamic_lstm.py

@@ -173,7 +173,7 @@ def train(args):
         trans_splice.TransSplice()
     ]
 
-    data_reader = reader.DataReader(args.feature_lst, args.label_lst)
+    data_reader = reader.DataReader(args.feature_lst, args.label_lst, -1)
     data_reader.set_transformers(ltrans)
 
     res_feature = fluid.LoDTensor()
@@ -198,7 +198,8 @@ def train(args):
                 fetch_list=[avg_cost] + accuracy.metrics,
                 return_numpy=False)
             train_acc = accuracy.eval(exe)
-            print("acc:", lodtensor_to_ndarray(loss))
+            print("pass_id", pass_id, "batch_id", batch_id, "acc:",
+                  lodtensor_to_ndarray(loss))
 
         pass_end_time = time.time()
         time_consumed = pass_end_time - pass_start_time