diff --git a/src/PTQ/PTQ.cpp b/src/PTQ/PTQ.cpp index 3a240971dd3465300c94885b8aed7d4fb0ef79b2..8a36e4f44fb3cdfe41981004907cc76996326ed0 100644 --- a/src/PTQ/PTQ.cpp +++ b/src/PTQ/PTQ.cpp @@ -1046,6 +1046,10 @@ void quantizeNormalizedNetwork(std::shared_ptr<GraphView> graphView, std::uint8_ std::shared_ptr<Node> scalingNode = getUniqueChild(node); // TODO : assert if scalingNode is a Scaling ... + // TODO : double check this ... + if (node->type() == "MatMul") + rescaling /= inputIsUnsigned ? unsignedMax : signedMax; + multiplyScalingFactor(scalingNode, rescaling) ; }