gguf-tools: Q4_K dequantization uses the wrong scale for half of each block
Reported · high · Wrong-result bug
The high-nibble weights of each Q4_K block use the scale and min of the earlier sub-block. The code dequantizes half of every Q4_K tensor with wrong values and gives no error.
Found: ; reported: ; status since: . Reporter: Maksim Borisov
Page for this report · Public report · Fix or proposed fix (see status)