2% sample actually or 4k

Baunsgaard · Oct 23, 2024 · 097f649 · 097f649
1 parent ecbc907
commit 097f649
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/src/main/java/org/apache/sysds/runtime/transform/encode/CompressedEncode.java b/src/main/java/org/apache/sysds/runtime/transform/encode/CompressedEncode.java
@@ -533,7 +533,7 @@ private <T> void  estimateRCDMapSize(ColumnEncoderComposite c) {
 			return;
 		}
 		// 2% sample or max 3000
-		int sampleSize = Math.min(in.getNumRows() / 2, 3000);
+		int sampleSize = Math.min(in.getNumRows() / 50, 4096);
 		// Find the frequencies of distinct values in the sample
 		Map<T, Integer> distinctFreq = new HashMap<>();
 		Array<T> col = (Array<T>) in.getColumn(c._colID-1);