Update word2vec_basic.py (tensorflow#13531)

Use random.sample to simplify random selection of context words

Author: ZxYuan

tensorflow/examples/tutorials/word2vec/word2vec_basic.py

@@ -115,11 +115,9 @@ def generate_batch(batch_size, num_skips, skip_window):
   data_index += span
   for i in range(batch_size // num_skips):
     context_words = [w for w in range(span) if w != skip_window]
-    random.shuffle(context_words)
-    words_to_use = collections.deque(context_words)
-    for j in range(num_skips):
+    words_to_use = random.sample(context_words, num_skips)
+    for j, context_word in enumerate(words_to_use):
       batch[i * num_skips + j] = buffer[skip_window]
-      context_word = words_to_use.pop()
       labels[i * num_skips + j, 0] = buffer[context_word]
     if data_index == len(data):
       buffer[:] = data[:span]