Repository: https://github.com/nntrainer/nntrainer
Commit: a7ea056e79ab8e14447ea305c1b634e233343258
Run date: 2026-09-03
Host: macOS ARM64

UINT4 reproducer
----------------
UINT4 scale=0.25 zero_point=8 q=[4,8,12,10]
C++ exception with description "[FloatTensor::copyData] Error: Unsupported data type"
Test status: FAILED
Process status: 1

UINT8 automatic-qparams reproducer
----------------------------------
UINT8 scale=1 zero_point=0 q=[0,0,128]
UINT8 dequantized=[0,0,128]
The difference between output.getValue(0, 0, 0, w) and input_data[w] is 127.5,
which exceeds generated_scale, where generated_scale evaluates to 1.
Test status: FAILED
Process status: 1

Existing explicit-parameter controls
------------------------------------
nntrainer_Quantizer.per_tensor_affine_04_p: PASSED
nntrainer_Quantizer.per_tensor_affine_05_p: PASSED
Control status: 2/2 passed

