From 14f969702520c4a2ad9bc7c1c3c6d41de68900ba Mon Sep 17 00:00:00 2001 From: Muhammad Haseeb Tariq Date: Sun, 3 Jul 2016 19:49:41 +0200 Subject: [PATCH] Inconsistency in libsvm formats (#1325) * Inconsistency in libsvm formats * note on libsvm formats --- jvm-packages/README.md | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/jvm-packages/README.md b/jvm-packages/README.md index a9aded6f8..7bbca4b13 100644 --- a/jvm-packages/README.md +++ b/jvm-packages/README.md @@ -16,6 +16,10 @@ and power xgboost into JVM ecosystem. You can find more about XGBoost on [Documentation](https://xgboost.readthedocs.org/en/latest/jvm/index.html) and [Resource Page](../demo/README.md). ## Hello World +*NOTE*: Use **1-based** ascending indexes for libsvm format in distributed training mode: +- Spark does the internal conversion, and do not accept formats that are 0-based +- WHEREAS - use **0-based** indexes format when predicting (for instance using the saved model in the Python package) in normal mode + ### XGBoost Scala ```scala import ml.dmlc.xgboost4j.scala.DMatrix