huaxingao commented on a change in pull request #28974:
URL: https://github.com/apache/spark/pull/28974#discussion_r449915124
##########
File path:
mllib/src/main/scala/org/apache/spark/ml/classification/LinearSVC.scala
##########
@@ -156,24 +156,9 @@ class LinearSVC @Since("2.2.0") (
def setAggregationDepth(value: Int): this.type = set(aggregationDepth, value)
setDefault(aggregationDepth -> 2)
- /**
- * Set block size for stacking input data in matrices.
- * If blockSize == 1, then stacking will be skipped, and each vector is
treated individually;
- * If blockSize > 1, then vectors will be stacked to blocks, and
high-level BLAS routines
- * will be used if possible (for example, GEMV instead of DOT, GEMM instead
of GEMV).
- * Recommended size is between 10 and 1000. An appropriate choice of the
block size depends
- * on the sparsity and dim of input datasets, the underlying BLAS
implementation (for example,
- * f2jBLAS, OpenBLAS, intel MKL) and its configuration (for example, number
of threads).
- * Note that existing BLAS implementations are mainly optimized for dense
matrices, if the
- * input dataset is sparse, stacking may bring no performance gain, the
worse is possible
- * performance regression.
- * Default is 1.
- *
- * @group expertSetParam
- */
@Since("3.1.0")
- def setBlockSize(value: Int): this.type = set(blockSize, value)
- setDefault(blockSize -> 1)
+ def setMaxBlockMemoryInMB(value: Int): this.type = set(maxBlockMemoryInMB,
value)
Review comment:
nit: add a scala doc?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
