diff --git a/tools/who_what_benchmark/whowhatbench/model_loaders.py b/tools/who_what_benchmark/whowhatbench/model_loaders.py
index 0209b65484..6517aed628 100644
--- a/tools/who_what_benchmark/whowhatbench/model_loaders.py
+++ b/tools/who_what_benchmark/whowhatbench/model_loaders.py
@@ -210,6 +210,7 @@ def load_text_model(
                     use_cache=True,
                     device=device,
                     ov_config=ov_config,
+                    **kwargs
                 )
             except Exception:
                 config = AutoConfig.from_pretrained(model_id)
@@ -219,6 +220,7 @@ def load_text_model(
                     use_cache=True,
                     device=device,
                     ov_config=ov_config,
+                    **kwargs
                 )
 
     return model
@@ -270,18 +272,18 @@ def load_text2image_model(
         if 'adapters' in kwargs and kwargs['adapters'] is not None:
             raise ValueError("Adapters are not supported for OVPipelineForText2Image.")
 
+        model_kwargs = {"ov_config": ov_config, "safety_checker": None}
+        if kwargs.get('from_onnx'):
+            model_kwargs['from_onnx'] = kwargs['from_onnx']
         try:
-            model = TEXT2IMAGEPipeline.from_pretrained(
-                model_id, device=device, ov_config=ov_config, safety_checker=None,
-            )
+            model = TEXT2IMAGEPipeline.from_pretrained(model_id, device=device, **model_kwargs)
         except ValueError:
             model = TEXT2IMAGEPipeline.from_pretrained(
                 model_id,
                 trust_remote_code=True,
                 use_cache=True,
                 device=device,
-                ov_config=ov_config,
-                safety_checker=None,
+                **model_kwargs
             )
 
     return model
@@ -410,7 +412,7 @@ def load_image2image_genai_pipeline(model_dir, device="CPU", ov_config=None):
 
 
 def load_imagetext2image_model(
-    model_id, device="CPU", ov_config=None, use_hf=False, use_genai=False
+    model_id, device="CPU", ov_config=None, use_hf=False, use_genai=False, **kwargs
 ):
     if use_hf:
         from diffusers import AutoPipelineForImage2Image
@@ -424,18 +426,18 @@ def load_imagetext2image_model(
     else:
         logger.info("Using Optimum API")
         from optimum.intel.openvino import OVPipelineForImage2Image
+        model_kwargs = {"ov_config": ov_config, "safety_checker": None}
+        if kwargs.get('from_onnx'):
+            model_kwargs['from_onnx'] = kwargs['from_onnx']
         try:
-            model = OVPipelineForImage2Image.from_pretrained(
-                model_id, device=device, ov_config=ov_config, safety_checker=None,
-            )
+            model = OVPipelineForImage2Image.from_pretrained(model_id, device=device, **model_kwargs)
         except ValueError:
             model = OVPipelineForImage2Image.from_pretrained(
                 model_id,
                 trust_remote_code=True,
                 use_cache=True,
                 device=device,
-                ov_config=ov_config,
-                safety_checker=None,
+                **model_kwargs
             )
     return model
 
@@ -455,7 +457,7 @@ def load_inpainting_genai_pipeline(model_dir, device="CPU", ov_config=None):
 
 
 def load_inpainting_model(
-    model_id, device="CPU", ov_config=None, use_hf=False, use_genai=False
+    model_id, device="CPU", ov_config=None, use_hf=False, use_genai=False, **kwargs
 ):
     if use_hf:
         from diffusers import AutoPipelineForInpainting
@@ -469,10 +471,11 @@ def load_inpainting_model(
     else:
         logger.info("Using Optimum API")
         from optimum.intel.openvino import OVPipelineForInpainting
+        model_kwargs = {"ov_config": ov_config, "safety_checker": None}
+        if kwargs.get('from_onnx'):
+            model_kwargs['from_onnx'] = kwargs['from_onnx']
         try:
-            model = OVPipelineForInpainting.from_pretrained(
-                model_id, device=device, ov_config=ov_config, safety_checker=None,
-            )
+            model = OVPipelineForInpainting.from_pretrained(model_id, device=device, **model_kwargs)
         except ValueError as e:
             logger.error("Failed to load inpaiting pipeline. Details:\n", e)
             model = OVPipelineForInpainting.from_pretrained(
@@ -480,8 +483,7 @@ def load_inpainting_model(
                 trust_remote_code=True,
                 use_cache=True,
                 device=device,
-                ov_config=ov_config,
-                safety_checker=None,
+                **model_kwargs
             )
     return model
 
@@ -530,8 +532,7 @@ def load_embedding_model(model_id, device="CPU", ov_config=None, use_hf=False, u
             model = OVModelForFeatureExtraction.from_pretrained(
                 model_id, device=device, ov_config=ov_config, safety_checker=None,
             )
-        except ValueError as e:
-            logger.error("Failed to load embedding pipeline. Details:\n", e)
+        except ValueError:
             model = OVModelForFeatureExtraction.from_pretrained(
                 model_id,
                 trust_remote_code=True,
@@ -596,8 +597,7 @@ def load_reranking_model(model_id, device="CPU", ov_config=None, use_hf=False, u
             model = model_cls.from_pretrained(
                 model_id, device=device, ov_config=ov_config, safety_checker=None,
             )
-        except ValueError as e:
-            logger.error("Failed to load reranking pipeline, an attempt will be made again with updated parameters. Details:\n", e)
+        except ValueError:
             model = model_cls.from_pretrained(
                 model_id,
                 trust_remote_code=True,
@@ -631,9 +631,9 @@ def load_model(
     elif model_type == "visual-text":
         return load_visual_text_model(model_id, device, ov_options, use_hf, use_genai, **kwargs)
     elif model_type == "image-to-image":
-        return load_imagetext2image_model(model_id, device, ov_options, use_hf, use_genai)
+        return load_imagetext2image_model(model_id, device, ov_options, use_hf, use_genai, **kwargs)
     elif model_type == "image-inpainting":
-        return load_inpainting_model(model_id, device, ov_options, use_hf, use_genai)
+        return load_inpainting_model(model_id, device, ov_options, use_hf, use_genai, **kwargs)
     elif model_type == "text-embedding":
         return load_embedding_model(model_id, device, ov_options, use_hf, use_genai, **kwargs)
     elif model_type == "text-reranking":
diff --git a/tools/who_what_benchmark/whowhatbench/wwb.py b/tools/who_what_benchmark/whowhatbench/wwb.py
index 73b975b439..7855df6490 100644
--- a/tools/who_what_benchmark/whowhatbench/wwb.py
+++ b/tools/who_what_benchmark/whowhatbench/wwb.py
@@ -711,7 +711,6 @@ def main():
         kwargs["cb_config"] = read_cb_config(args.cb_config)
     if args.from_onnx:
         kwargs["from_onnx"] = args.from_onnx
-        kwargs["use_cache"] = False
     if args.gguf_file:
         kwargs["gguf_file"] = args.gguf_file
     if args.adapters is not None:
@@ -721,9 +720,10 @@ def main():
         else:
             kwargs["alphas"] = [1.0] * len(args.adapters)
     kwargs["empty_adapters"] = args.empty_adapters
-    kwargs["embeds_pooling"] = args.embeds_pooling_type
-    kwargs["embeds_normalize"] = args.embeds_normalize
-    kwargs["embeds_padding_side"] = args.embeds_padding_side
+    if args.model_type == "text-embedding":
+        kwargs["embeds_pooling"] = args.embeds_pooling_type
+        kwargs["embeds_normalize"] = args.embeds_normalize
+        kwargs["embeds_padding_side"] = args.embeds_padding_side
 
     if args.gt_data and os.path.exists(args.gt_data):
         evaluator = create_evaluator(None, args)