From 8e1ad36ed053125963850fbfe54ef37503caf34c Mon Sep 17 00:00:00 2001 From: Ekaterina Aidova Date: Wed, 6 Dec 2023 13:27:48 +0400 Subject: [PATCH] Fix causallm weights compression via quantizer (#484) --- optimum/intel/openvino/quantization.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/optimum/intel/openvino/quantization.py b/optimum/intel/openvino/quantization.py index 7a7da29040..f0230a7ee7 100644 --- a/optimum/intel/openvino/quantization.py +++ b/optimum/intel/openvino/quantization.py @@ -260,7 +260,8 @@ def _quantize_ovcausallm( save_directory.mkdir(parents=True, exist_ok=True) if weights_only: - self.model.model = nncf.compress_weights(self.model.model) + model = nncf.compress_weights(self.model._original_model) + self.model.model = model self.model.save_pretrained(save_directory) return