much faster set-intersection based version

tensorflow · Nov 1, 2023 · f8da6c5 · f8da6c5
1 parent 99aad3b
commit f8da6c5
Showing 1 changed file with 7 additions and 11 deletions.
diff --git a/docs/tutorials/word2vec.ipynb b/docs/tutorials/word2vec.ipynb
@@ -728,13 +728,12 @@
         "          window_size=window_size,\n",
         "          negative_samples=0)\n",
         "\n",
-        "    # Generate positive context windows for each target word in the sequence.\n",
-        "    window = defaultdict(list)\n",
-        "    for i in range(window_size, len(sequence)-window_size):\n",
-        "      window[sequence[i]].append(sequence[i-window_size:1+i+window_size])\n",
-        "\n",
         "    # Iterate over each positive skip-gram pair to produce training examples\n",
         "    # with a positive context word and negative samples.\n",
+        "    window = defaultdict(set)\n",
+        "    for target_word, context_word in positive_skip_grams:\n",
+        "      window[target_word].add(context_word)\n",
+        "\n",
         "    for target_word, context_word in positive_skip_grams:\n",
         "      context_class = tf.expand_dims(\n",
         "          tf.constant([context_word], dtype=\"int64\"), 1)\n",
@@ -747,12 +746,9 @@
         "          seed=seed,\n",
         "          name=\"negative_sampling\")\n",
         "\n",
-        "      # Discard iteration if negative samples overlap with positive context.\n",
-        "      for target in window[target_word]:\n",
-        "        if not any(t in target for t in negative_sampling_candidates):\n",
-        "          break  # All candidates are true negatives: use this skip_gram.\n",
-        "      else:\n",
-        "        continue # Discard this skip_gram.\n",
+        "      # Discard this negative sample if it intersects with the positive context.\n",
+        "      if window[target_word].intersection(negative_sampling_candidates.numpy()):\n",
+        "        continue\n",
         "\n",
         "      # Build context and label vectors (for one target word).\n",
         "      context = tf.concat([tf.squeeze(context_class,1), negative_sampling_candidates], 0)\n",