Merge pull request #741 from yinhaofeng/fix_tipc

fix_tipc

Author: frankwhzhang

models/rank/flen/README.md

@@ -71,7 +71,7 @@ os : windows/linux/macos
 python -u ../../../tools/trainer.py -m config.yaml # 全量数据运行config_bigdata.yaml 
 # 动态图预测
 python -u ../../../tools/infer.py -m config.yaml # 全量数据运行config_bigdata.yaml
-
+```
 
 ## 模型组网
 

models/recall/word2vec/dygraph_model.py

@@ -85,7 +85,6 @@ def create_metrics(self):
     # construct train forward phase  
     def train_forward(self, dy_model, metrics_list, batch_data, config):
         input_word, true_word, neg_word = self.create_feeds(batch_data, config)
-
         true_logits, neg_logits = dy_model.forward(
             [input_word, true_word, neg_word])
         loss = self.create_loss(true_logits, neg_logits, config)

models/recall/word2vec/static_model.py

@@ -95,6 +95,7 @@ def net(self, inputs, is_infer=False):
 
     def create_optimizer(self, strategy=None):
         optimizer = paddle.optimizer.SGD(learning_rate=self.learning_rate)
+        optimizer.minimize(self._cost)
         #            learning_rate=paddle.fluid.layers.exponential_decay(
         #                learning_rate=self.learning_rate,
         #                decay_steps=self.decay_steps,

models/recall/word2vec/word2vec_reader.py

@@ -16,7 +16,8 @@
 import numpy as np
 import io
 import six
-
+import time
+import random
 from paddle.io import IterableDataset
 
 
@@ -35,7 +36,7 @@ def __call__(self):
             self.idx = 0
 
         result = self.buffer[self.idx]
-        self.idx += 1
+        self.idx = self.idx + 1
         return result
 
 
@@ -52,7 +53,9 @@ def init(self):
         self.neg_num = self.config.get("hyper_parameters.neg_num")
         self.with_shuffle_batch = self.config.get(
             "hyper_parameters.with_shuffle_batch")
-        self.random_generator = NumpyRandomInt(1, self.window_size + 1)
+        #self.random_generator = NumpyRandomInt(1, self.window_size + 1)
+        np.random.seed(12345)
+        self.random_generator = np.random.randint(1, self.window_size + 1)
         self.batch_size = self.config.get("runner.batch_size")
 
         self.cs = None
@@ -78,7 +81,7 @@ def get_context_words(self, words, idx):
         idx: input word index
         window_size: window size
         """
-        target_window = self.random_generator()
+        target_window = self.random_generator
         # if (idx - target_window) > 0 else 0
         start_point = idx - target_window
         if start_point < 0:
@@ -102,11 +105,15 @@ def __iter__(self):
                                 np.array([int(target_id)]).astype('int64'))
                             output.append(
                                 np.array([int(context_id)]).astype('int64'))
-                            np.random.seed(12345)
-                            neg_array = self.cs.searchsorted(
-                                np.random.sample(self.neg_num))
+
+                            tmp = []
+                            random.seed(12345)
+                            for i in range(self.neg_num):
+                                tmp.append(random.random())
+                            neg_array = self.cs.searchsorted(tmp)
+
                             output.append(
-                                np.array([int(str(i))
+                                np.array([int(i)
                                           for i in neg_array]).astype('int64'))
                             yield output
 

test_tipc/prepare.sh

@@ -272,6 +272,7 @@ elif [ ${model_name} == "tisas" ]; then
         cp -r ./models/recall/tisas/data/sample_data/* ./test_tipc/data/train
         cp -r ./models/recall/tisas/data/sample_data/* ./test_tipc/data/infer
         echo "demo data ready"
+    fi
 
 elif [ ${model_name} == "dselect_k" ]; then
     mkdir -p ./test_tipc/data/train
@@ -300,5 +301,4 @@ elif [ ${model_name} == "dselect_k" ]; then
         cp -r ./models/multitask/dselect_k/data/* ./test_tipc/data/train
         cp -r ./datasets/Multi_MNIST_DselectK/test/* ./test_tipc/data/infer
     fi
-
 fi