From 13c169142662ec9c3f6040f8374e1d05f8aa71cf Mon Sep 17 00:00:00 2001
From: joe <zjiuzhou@gmail.com>
Date: Wed, 30 Jan 2019 10:39:56 +0800
Subject: [PATCH] =?UTF-8?q?=E4=BF=AE=E6=94=B9=E9=AA=8C=E8=AF=81=E9=9B=86?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 README.md                  |  2 +-
 data/{test.csv => dev.csv} |  0
 similarity.py              | 20 ++++++++++----------
 3 files changed, 11 insertions(+), 11 deletions(-)
 rename data/{test.csv => dev.csv} (100%)

diff --git a/README.md b/README.md
index ca1ceef..e95fad6 100644
--- a/README.md
+++ b/README.md
@@ -1,6 +1,6 @@
 # bert-utils
 
-本文基于Google开源的[BERT]()代码进行了进一步的简化，方便生成句向量与做文本分类
+本文基于Google开源的[BERT](https://github.com/google-research/bert)代码进行了进一步的简化，方便生成句向量与做文本分类
 
 1、下载BERT中文模型 
 
diff --git a/data/test.csv b/data/dev.csv
similarity index 100%
rename from data/test.csv
rename to data/dev.csv
diff --git a/similarity.py b/similarity.py
index 4d6e792..0472070 100644
--- a/similarity.py
+++ b/similarity.py
@@ -664,13 +664,13 @@ class BertSim:
 
 if __name__ == '__main__':
     sim = BertSim()
-    # sim.set_mode(tf.estimator.ModeKeys.TRAIN)
-    # sim.train()
-    # sim.set_mode(tf.estimator.ModeKeys.EVAL)
-    # sim.eval()
-    sim.set_mode(tf.estimator.ModeKeys.PREDICT)
-    while True:
-        sentence1 = input('sentence1: ')
-        sentence2 = input('sentence2: ')
-        predict = sim.predict(sentence1, sentence2)
-        print(f'similarity：{predict[0][1]}')
+    sim.set_mode(tf.estimator.ModeKeys.TRAIN)
+    sim.train()
+    sim.set_mode(tf.estimator.ModeKeys.EVAL)
+    sim.eval()
+    # sim.set_mode(tf.estimator.ModeKeys.PREDICT)
+    # while True:
+    #     sentence1 = input('sentence1: ')
+    #     sentence2 = input('sentence2: ')
+    #     predict = sim.predict(sentence1, sentence2)
+    #     print(f'similarity：{predict[0][1]}')