do not compress model twise when compress during export

huggingface · Dec 25, 2023 · eecf181 · eecf181
1 parent 3bc3d01
commit eecf181
Show file tree

Hide file tree

Showing 4 changed files with 4 additions and 4 deletions.
diff --git a/optimum/intel/openvino/modeling_base.py b/optimum/intel/openvino/modeling_base.py
@@ -302,7 +302,7 @@ def _from_transformers(
         )
 
         config.save_pretrained(save_dir_path)
-        return cls._from_pretrained(model_id=save_dir_path, config=config, load_in_8bit=load_in_8bit, **kwargs)
+        return cls._from_pretrained(model_id=save_dir_path, config=config, load_in_8bit=False, **kwargs)
 
     @classmethod
     def _to_load(

diff --git a/optimum/intel/openvino/modeling_base_seq2seq.py b/optimum/intel/openvino/modeling_base_seq2seq.py
@@ -270,7 +270,7 @@ def _from_transformers(
 
         config.save_pretrained(save_dir_path)
         return cls._from_pretrained(
-            model_id=save_dir_path, config=config, use_cache=use_cache, load_in_8bit=load_in_8bit, **kwargs
+            model_id=save_dir_path, config=config, use_cache=use_cache, load_in_8bit=False, **kwargs
         )
 
     def _reshape(self, model: openvino.runtime.Model, batch_size: int, sequence_length: int, is_decoder=True):

diff --git a/optimum/intel/openvino/modeling_decoder.py b/optimum/intel/openvino/modeling_decoder.py
@@ -249,7 +249,7 @@ def _from_transformers(
         config.is_encoder_decoder = False
         config.save_pretrained(save_dir_path)
         return cls._from_pretrained(
-            model_id=save_dir_path, config=config, use_cache=use_cache, load_in_8bit=load_in_8bit, **kwargs
+            model_id=save_dir_path, config=config, use_cache=use_cache, load_in_8bit=False, **kwargs
         )
 
     def _reshape(

diff --git a/optimum/intel/openvino/modeling_diffusion.py b/optimum/intel/openvino/modeling_diffusion.py
@@ -325,7 +325,7 @@ def _from_transformers(
             tokenizer_2=tokenizer_2,
             scheduler=scheduler,
             feature_extractor=feature_extractor,
-            load_in_8bit=load_in_8bit,
+            load_in_8bit=False,
             **kwargs,
         )