fix style

huggingface · Feb 15, 2024 · f1c9d6f · f1c9d6f
1 parent 70468a6
commit f1c9d6f
Show file tree

Hide file tree

Showing 3 changed files with 8 additions and 7 deletions.
diff --git a/optimum/intel/openvino/modeling_decoder.py b/optimum/intel/openvino/modeling_decoder.py
@@ -242,7 +242,6 @@ def _from_transformers(
         quantization_config: Optional[Union[OVWeightQuantizationConfig, Dict]] = None,
         **kwargs,
     ):
-
         if config.model_type.replace("_", "-") not in _SUPPORTED_ARCHITECTURES:
             logger.warning(
                 f"This architecture : {config.model_type} was not validated, only :{', '.join(_SUPPORTED_ARCHITECTURES)} architectures were "

diff --git a/optimum/intel/openvino/quantization.py b/optimum/intel/openvino/quantization.py
@@ -588,9 +588,8 @@ def _int4_weight_only_quantization(
     if model.export_feature != "text-generation":
         raise ValueError("Only `OVModelForCausalLM` are supported for now")
 
-
     quantization_config = quantization_config or _check_default_4bit_configs(model.config)
-    
+
     # Data-free weight-only quantization to asymmetric INT4
     if quantization_config is None:
         quantization_config = OVWeightQuantizationConfig(bits=4, sym=False)
@@ -632,7 +631,7 @@ def _weight_only_quantization(model: OVBaseModel, quantization_config: Union[OVW
         mode = CompressWeightsMode.INT8_SYM if config.sym else CompressWeightsMode.INT8_ASYM
     else:
         mode = CompressWeightsMode.INT4_SYM if config.sym else CompressWeightsMode.INT4_ASYM
-        
+
     model.model = nncf.compress_weights(
         ov_model,
         mode=mode,
@@ -644,4 +643,3 @@ def _weight_only_quantization(model: OVBaseModel, quantization_config: Union[OVW
         ignored_scope=ignored_scope,
         dataset=dataset,
     )
-
diff --git a/tests/openvino/test_quantization.py b/tests/openvino/test_quantization.py
@@ -349,7 +349,9 @@ def test_ovmodel_load_with_compressed_weights(self, model_cls, model_type):
     def test_ovmodel_4bit_auto_compression(self, model_cls, model_type, expected_ov_int8, expected_ov_int4):
         with tempfile.TemporaryDirectory() as tmp_dir:
             model_id = MODEL_NAMES[model_type]
-            model = model_cls.from_pretrained(model_id, export=True, quantization_config=OVWeightQuantizationConfig(bits=4))
+            model = model_cls.from_pretrained(
+                model_id, export=True, quantization_config=OVWeightQuantizationConfig(bits=4)
+            )
             tokenizer = AutoTokenizer.from_pretrained(model_id)
             if tokenizer.pad_token is None:
                 tokenizer.pad_token = tokenizer.eos_token
@@ -404,7 +406,9 @@ def transform_fn(data, tokenizer):
         model = model_cls.from_pretrained(
             model_id,
             export=True,
-            quantization_config=OVWeightQuantizationConfig(bits=4, sym=True, group_size=-1, ratio=0.8, dataset=quantization_dataset),
+            quantization_config=OVWeightQuantizationConfig(
+                bits=4, sym=True, group_size=-1, ratio=0.8, dataset=quantization_dataset
+            ),
         )
 
         _, num_int8, num_int4 = get_num_quantized_nodes(model)