Update quantize.cpp

The new imatrix GGUF format stores per-matrice token counts instead of per-tensor chunk counts which makes it possible to fix NaN's for low bits per wight quants

Author: nicoboss

tools/quantize/quantize.cpp

@@ -289,7 +289,7 @@ static int load_imatrix(const std::string & imatrix_file, std::vector<std::strin
             const float count = ((const float *) counts->data)[j];
             if (count > 0.0f) {
                 for (int64_t i = 0; i < ne0; ++i) {
-                    e[j*ne0 + i] = ((const float *) sums->data)[j*ne0 + i] / count;
+                    e[j*ne0 + i] = (((const float *) sums->data)[j*ne0 + i] + 1.0f) / (count + 1.0f);
                 }
             } else {
                 // Partial imatrix data, this tensor never got any input during calibration