Hendrik/xgboost
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Backport doc fixes that are compatible with 0.72 release

Browse Source 
* Clarify behavior of LIBSVM in XGBoost4J-Spark (#3524)
* Fix typo in faq.rst (#3521)
* Fix typo in parameter.rst, gblinear section (#3518)
* Clarify supported OSes for XGBoost4J published JARs (#3547)
* Update broken links (#3565)
* Grammar fixes and typos (#3568)
* Bring XGBoost4J Intro up-to-date (#3574)
This commit is contained in:
Nan Zhu
2018-07-28 17:34:39 -07:00
committed by Philip Cho
parent e19dded9a3
commit 4334b9cc91
6 changed files with 77 additions and 80 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
jvm-packages/README.md
View File

@@ -68,8 +68,12 @@ be found in the [examples package](https://github.com/dmlc/xgboost/tree/master/j
**NOTE on LIBSVM Format**:
* Use *1-based* ascending indexes for the LIBSVM format in distributed training mode
There is an inconsistent issue between XGBoost4J-Spark and other language bindings of XGBoost.
* Spark does the internal conversion, and does not accept formats that are 0-based
When users use Spark to load trainingset/testset in LibSVM format with the following code snippet:
* Whereas, use *0-based* indexes format when predicting in normal mode - for instance, while using the saved model in the Python package
```scala
spark.read.format("libsvm").load("trainingset_libsvm")
```
Spark assumes that the dataset is 1-based indexed. However, when you do prediction with other bindings of XGBoost (e.g. Python API of XGBoost), XGBoost assumes that the dataset is 0-based indexed. It creates a pitfall for the users who train model with Spark but predict with the dataset in the same format in other bindings of XGBoost.