add is_sparse

fluid/PaddleRec/word2vec/network_conf.py

@@ -64,21 +64,11 @@ def skip_gram_word2vec(dict_size,
             non_leaf_num=non_leaf_num,
             ptable=ptable,
             pcode=pcode,
-            is_costum=True)
+            is_costum=True,
+            is_sparse=is_sparse)
 
         return cost
 
-    def get_loss(loss1, loss2):
-        loss_op1 = fluid.layers.elementwise_sub(
-            fluid.layers.fill_constant_batch_size_like(input=loss1, shape=[-1, 1], value=margin,
-                                                       dtype='float32'), cos_q_pt)
-        loss_op2 = fluid.layers.elementwise_add(loss_op1, cos_q_nt)
-        loss_op3 = fluid.layers.elementwise_max(
-            fluid.layers.fill_constant_batch_size_like(input=loss_op2, shape=[-1, 1], value=0.0,
-                                                       dtype='float32'), loss_op2)
-        avg_cost = fluid.layers.mean(loss_op3)
-        return avg_cost
-
     datas = []
 
     input_word = fluid.layers.data(name="input_word", shape=[1], dtype='int64')
@@ -87,14 +77,12 @@ def skip_gram_word2vec(dict_size,
     datas.append(predict_word)
 
     if with_hsigmoid:
-        if max_code_length:
-            ptable = fluid.layers.data(
-                name='ptable', shape=[max_code_length], dtype='int64')
-            pcode = fluid.layers.data(
-                name='pcode', shape=[max_code_length], dtype='int64')
-        else:
-            ptable = fluid.layers.data(name='ptable', shape=[40], dtype='int64')
-            pcode = fluid.layers.data(name='pcode', shape=[40], dtype='int64')
+        ptable = fluid.layers.data(name='ptable',
+                                   shape=[max_code_length if max_code_length else 40],
+                                   dtype='int64')
+        pcode = fluid.layers.data(name='pcode',
+                                  shape=[max_code_length if max_code_length else 40],
+                                  dtype='int64')
         datas.append(ptable)
         datas.append(pcode)
 
@@ -116,13 +104,13 @@ def skip_gram_word2vec(dict_size,
 
     if with_nce:
         cost_nce = nce_layer(emb, words[1], embedding_size, dict_size, 5, "uniform",
-                         word_frequencys, None)
+                             word_frequencys, None)
         cost = cost_nce
     if with_hsigmoid:
         cost_hs = hsigmoid_layer(emb, words[1], words[2], words[3], dict_size)
         cost = cost_hs
     if with_nce and with_hsigmoid:
-        cost = fluid.layers.elementwise_add(cost_nce, cost)
+        cost = fluid.layers.elementwise_add(cost_nce, cost_hs)
 
     avg_cost = fluid.layers.reduce_mean(cost)
 

fluid/PaddleRec/word2vec/train.py

@@ -87,7 +87,7 @@ def parse_args():
         '--with_nce',
         action='store_true',
         required=False,
-        default=True,
+        default=False,
         help='using negtive sampling, (default: True)')
 
     parser.add_argument(
@@ -188,6 +188,9 @@ def train_loop(args, train_program, reader, py_reader, loss, trainer_id):
 def train():
     args = parse_args()
 
+    if not args.with_nce and not args.with_hs:
+        logger.error("with_nce or with_hs must choose one")
+
     if not os.path.isdir(args.model_output_dir):
         os.mkdir(args.model_output_dir)