Relax tolerance for FP comparison in test harness

Tests like RNN needs slightly more relaxed tolerance than 4ULP.
Use 1e-5 from corresponding TFLite tests.

Bug: 63905942
Test: VtsHalNeuralnetworksV1_0TargetTest on Angler
Change-Id: I721711e13575f56e956d87e3ad537467f85e93f4

parent 502fdebb

neuralnetworks/1.0/vts/functional/GeneratedTestHarness.cpp

+1 −1

Original line number	Diff line number	Diff line
		@@ -172,7 +172,7 @@ void Execute(const sp<IDevice>& device, std::function<Model(void)> create_model,
		auto& test_float = test_float_operands[index];
		for (unsigned int i = 0; i < golden_float.size(); i++) {
		SCOPED_TRACE(i);
		EXPECT_FLOAT_EQ(golden_float[i], test_float[i]);
		EXPECT_NEAR(golden_float[i], test_float[i], 1.e-5);
		}
		});