From 90bceaafdc0b8845462301ac3301082e7faf5e59 Mon Sep 17 00:00:00 2001
From: Yu Ji <maple.jiyu@hotmail.com>
Date: Mon, 29 Jul 2019 06:04:56 +0000
Subject: [PATCH] Replace PyReader with PipeReader

---
 PaddleRec/word2vec/net.py   | 13 +++----------
 PaddleRec/word2vec/train.py | 18 ++++++------------
 2 files changed, 9 insertions(+), 22 deletions(-)

diff --git a/PaddleRec/word2vec/net.py b/PaddleRec/word2vec/net.py
index ab2abbc7..21903f4a 100644
--- a/PaddleRec/word2vec/net.py
+++ b/PaddleRec/word2vec/net.py
@@ -21,21 +21,14 @@ import paddle.fluid as fluid
 
 
 def skip_gram_word2vec(dict_size, embedding_size, is_sparse=False, neg_num=5):
-
-    datas = []
     input_word = fluid.layers.data(name="input_word", shape=[1], dtype='int64')
     true_word = fluid.layers.data(name='true_label', shape=[1], dtype='int64')
     neg_word = fluid.layers.data(
         name="neg_label", shape=[neg_num], dtype='int64')
 
-    datas.append(input_word)
-    datas.append(true_word)
-    datas.append(neg_word)
-
-    py_reader = fluid.layers.create_py_reader_by_data(
-        capacity=64, feed_list=datas, name='py_reader', use_double_buffer=True)
+    words = [input_word, true_word, neg_word]
+    pipe_reader = fluid.reader.PipeReader(feed_list=words)
 
-    words = fluid.layers.read_file(py_reader)
     init_width = 0.5 / embedding_size
     input_emb = fluid.layers.embedding(
         input=words[0],
@@ -107,7 +100,7 @@ def skip_gram_word2vec(dict_size, embedding_size, is_sparse=False, neg_num=5):
         fluid.layers.reduce_sum(
             neg_xent, dim=1))
     avg_cost = fluid.layers.reduce_mean(cost)
-    return avg_cost, py_reader
+    return avg_cost, pipe_reader
 
 
 def infer_network(vocab_size, emb_size):
diff --git a/PaddleRec/word2vec/train.py b/PaddleRec/word2vec/train.py
index 430ec132..648ecdf2 100644
--- a/PaddleRec/word2vec/train.py
+++ b/PaddleRec/word2vec/train.py
@@ -91,21 +91,15 @@ def convert_python_to_tensor(weight, batch_size, sample_reader):
             if len(result[0]) == batch_size:
                 tensor_result = []
                 for tensor in result:
-                    t = fluid.Tensor()
                     dat = np.array(tensor, dtype='int64')
                     if len(dat.shape) > 2:
                         dat = dat.reshape((dat.shape[0], dat.shape[2]))
                     elif len(dat.shape) == 1:
                         dat = dat.reshape((-1, 1))
-                    t.set(dat, fluid.CPUPlace())
-                    tensor_result.append(t)
-                tt = fluid.Tensor()
+                    tensor_result.append(dat)
                 neg_array = cs.searchsorted(np.random.sample(args.nce_num))
-                neg_array = np.tile(neg_array, batch_size)
-                tt.set(
-                    neg_array.reshape((batch_size, args.nce_num)),
-                    fluid.CPUPlace())
-                tensor_result.append(tt)
+                neg_array = np.tile(neg_array, batch_size).reshape((batch_size, args.nce_num))
+                tensor_result.append(neg_array)
                 yield tensor_result
                 result = [[], []]
 
@@ -115,7 +109,7 @@ def convert_python_to_tensor(weight, batch_size, sample_reader):
 def train_loop(args, train_program, reader, py_reader, loss, trainer_id,
                weight):
 
-    py_reader.decorate_tensor_provider(
+    py_reader.decorate_batch_generator(
         convert_python_to_tensor(weight, args.batch_size, reader.train()))
 
     place = fluid.CPUPlace()
@@ -153,9 +147,9 @@ def train_loop(args, train_program, reader, py_reader, loss, trainer_id,
 
                 if batch_id % args.print_batch == 0:
                     logger.info(
-                        "TRAIN --> pass: {} batch: {} loss: {} reader queue:{}".
+                        "TRAIN --> pass: {} batch: {} loss: {}".
                         format(pass_id, batch_id,
-                               loss_val.mean(), py_reader.queue.size()))
+                               loss_val.mean()))
                 if args.with_speed:
                     if batch_id % 500 == 0 and batch_id != 0:
                         elapsed = (time.time() - start)
-- 
GitLab