Maciej Szymkiewicz created SPARK-38139:
------------------------------------------
Summary: ml.recommendation.ALS doctests failures
Key: SPARK-38139
URL: https://issues.apache.org/jira/browse/SPARK-38139
Project: Spark
Issue Type: Bug
Components: ML, PySpark
Affects Versions: 3.3.0
Reporter: Maciej Szymkiewicz
In my dev setups, ml.recommendation:ALS test consistently converges to value
lower than expected and fails with:
{code:python}
File "/path/to/spark/python/pyspark/ml/recommendation.py", line 322, in
__main__.ALS
Failed example:
predictions[0]
Expected:
Row(user=0, item=2, newPrediction=0.69291...)
Got:
Row(user=0, item=2, newPrediction=0.6929099559783936)
{code}
In can correct for that, but it creates some noise, so if anyone else
experiences this, we could drop a digit from the results
{code}
diff --git a/python/pyspark/ml/recommendation.py
b/python/pyspark/ml/recommendation.py
index f0628fb922..b8e2a6097d 100644
--- a/python/pyspark/ml/recommendation.py
+++ b/python/pyspark/ml/recommendation.py
@@ -320,7 +320,7 @@ class ALS(JavaEstimator, _ALSParams, JavaMLWritable,
JavaMLReadable):
>>> test = spark.createDataFrame([(0, 2), (1, 0), (2, 0)], ["user",
"item"])
>>> predictions = sorted(model.transform(test).collect(), key=lambda r:
r[0])
>>> predictions[0]
- Row(user=0, item=2, newPrediction=0.69291...)
+ Row(user=0, item=2, newPrediction=0.6929...)
>>> predictions[1]
Row(user=1, item=0, newPrediction=3.47356...)
>>> predictions[2]
{code}
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
