Disable Float32_2Bits_Asymmetric_256x256 test (#27046)

vraspar · web-flow · commit 5e7e7a300b23 · 2026-01-21T00:03:57.000Z
### Description
This test seems to be flaky and fails `Linux QNN CI Pipeline`. Disabling
this test until I figure out the root cause for the inaccuracy


### Motivation and Context
diff --git a/onnxruntime/test/contrib_ops/matmul_2bits_test.cc b/onnxruntime/test/contrib_ops/matmul_2bits_test.cc
@@ -371,7 +371,8 @@ TEST(MatMulNBitsLutGemm, Float32_2Bits_Symmetric_256x256) {
   TestMatMul2BitsLutGemm<float>(1, 256, 256, 32, false);
 }
 
-TEST(MatMulNBitsLutGemm, Float32_2Bits_Asymmetric_256x256) {
+// TODO: Re-enable once LUT GEMM asymmetric quantization accuracy issue is resolved
+TEST(MatMulNBitsLutGemm, DISABLED_Float32_2Bits_Asymmetric_256x256) {
   TestMatMul2BitsLutGemm<float>(1, 256, 256, 32, true);
 }
 

Original file line number	Diff line number	Diff line change
`@@ -371,7 +371,8 @@ TEST(MatMulNBitsLutGemm, Float32_2Bits_Symmetric_256x256) {`
`371`	`371`	`TestMatMul2BitsLutGemm<float>(1, 256, 256, 32, false);`
`372`	`372`	`}`
`373`	`373`
`374`		`-TEST(MatMulNBitsLutGemm, Float32_2Bits_Asymmetric_256x256) {`
	`374`	`+// TODO: Re-enable once LUT GEMM asymmetric quantization accuracy issue is resolved`
	`375`	`+TEST(MatMulNBitsLutGemm, DISABLED_Float32_2Bits_Asymmetric_256x256) {`
`375`	`376`	`TestMatMul2BitsLutGemm<float>(1, 256, 256, 32, true);`
`376`	`377`	`}`
`377`	`378`