Update quant.py (#114)

ModelTC · Sep 27, 2024 · e54df96 · e54df96
1 parent e68484e
commit e54df96
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/llmc/compression/quantization/quant.py b/llmc/compression/quantization/quant.py
@@ -176,7 +176,7 @@ def get_qparams(self, tensor_range, device):
             scales = abs_max / qmax
             zeros = torch.tensor(0.0)
         else:
-            scales = (max_val - min_val).clamp(min=1e-5) / qmax
+            scales = (max_val - min_val).clamp(min=1e-5) / (qmax - qmin)
             zeros = (qmin - torch.round(min_val / scales)).clamp(qmin, qmax)
             if not self.round_zp:
                 zeros = qmin - (min_val / scales)