Github user jkbradley commented on a diff in the pull request:
https://github.com/apache/spark/pull/5748#discussion_r31455507
--- Diff:
mllib/src/main/scala/org/apache/spark/mllib/feature/Word2Vec.scala ---
@@ -426,38 +422,40 @@ class Word2Vec extends Serializable with Logging {
/**
* :: Experimental ::
* Word2Vec model
+ *
+ * @param wordIndex: Maps each word to an index, which can retrieve the
corresponding
+ * vector from wordVectors (see below).
+ * @param wordVectors: Array of length numWords * vectorSize, vector
corresponding
+ * to the word mapped with index i can be retrieved by
the slice
+ * (i * vectorSize, i * vectorSize + vectorSize)
*/
@Experimental
class Word2VecModel private[mllib] (
- model: Map[String, Array[Float]]) extends Serializable with Saveable {
-
- // wordList: Ordered list of words obtained from model.
- private val wordList: Array[String] = model.keys.toArray
+ wordIndex: Map[String, Int],
+ wordVectors: Array[Float]) extends Serializable with Saveable {
- // wordIndex: Maps each word to an index, which can retrieve the
corresponding
- // vector from wordVectors (see below).
- private val wordIndex: Map[String, Int] = wordList.zip(0 until
model.size).toMap
-
- // vectorSize: Dimension of each word's vector.
- private val vectorSize = model.head._2.size
private val numWords = wordIndex.size
+ // vectorSize: Dimension of each word's vector.
+ private val vectorSize = wordVectors.length / numWords
+
+ // wordList: Ordered list of words obtained from wordIndex.
+ private val wordList: Array[String] = wordIndex.keys.toArray
--- End diff --
This should sort by ```wordIndex._2``` to make sure the order matches
wordVectors
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
