Fix tests

nncf/torch/model_creation.py

@@ -64,7 +64,7 @@ def create_compressed_model(
     """
     The main function used to produce a model ready for compression fine-tuning from an original PyTorch
     model and a configuration object.
-    dummy_forward_fn
+
     :param model: The original model. Should have its parameters already loaded from a checkpoint or another
         source.
     :param config: A configuration object used to determine the exact compression modifications to be applied
@@ -124,8 +124,7 @@ def create_compressed_model(
 
     should_init = compression_state is None
 
-    input_info = get_input_info_from_config(config)
-    nncf_network = create_nncf_network(model, input_info, dummy_forward_fn, wrap_inputs_fn, wrap_outputs_fn)
+    nncf_network = create_nncf_network(model, config, dummy_forward_fn, wrap_inputs_fn, wrap_outputs_fn)
 
     if dump_graphs and is_main_process():
         nncf_network.nncf.get_graph().visualize_graph(osp.join(config.get("log_dir", "."), "original_graph.dot"))
@@ -186,27 +185,24 @@ def get_input_info_from_config(config: NNCFConfig) -> ModelInputInfo:
 
 
 def create_nncf_network_with_inputs_from_config(model: torch.nn.Module, config: NNCFConfig):
-    return create_nncf_network(model, input_info=get_input_info_from_config(config))
+    return create_nncf_network(model, config=config)
 
 
 def create_nncf_network(
     model: torch.nn.Module,
-    input_info: ModelInputInfo,
+    config: NNCFConfig,
     dummy_forward_fn: Callable[[Module], Any] = None,
     wrap_inputs_fn: Callable = None,
-    wrap_outputs_fn: Callable = None,
-    ignored_scopes: List[str] = None,
-    target_scopes: List[str] = None,
-    scopes_without_shape_matching: List[str] = None,
+    wrap_outputs_fn: Callable = None
 ) -> NNCFNetwork:
     """
     The main function used to produce a model ready for adding compression from an original PyTorch
     model and a configuration object.
 
     :param model: The original model. Should have its parameters already loaded from a checkpoint or another
         source.
-    :param input_info: A list of descriptors for each of the model's tensor inputs, used to build the input
-        parameters for the internal `.forward` calls made on the model object to build its graph representation
+    :param config: A configuration object used to determine the exact compression modifications to be applied
+        to the model
     :param dummy_forward_fn: if supplied, will be used instead of a *forward* function call to build
         the internal graph representation via tracing. Specifying this is useful when the original training pipeline
         has special formats of data loader output or has additional *forward* arguments other than input tensors.
@@ -228,12 +224,6 @@ def create_nncf_network(
         the same as were supplied in input, but each tensor in the original input. Must be specified if
         dummy_forward_fn is specified.
     :param wrap_outputs_fn: Same as `wrap_inputs_fn`, but for marking model outputs with
-    :param scopes_without_shape_matching: A list of scopes in the model in which the activation tensor shapes will
-        not be considered for purposes of scope matching - this helps handle RNN-like cases.
-    :param ignored_scopes: A list of scopes in the model for which NNCF handling should not be applied. Functions as
-    a "denylist". If left unspecified, nothing will be ignored.
-    :param target_scopes: A list of scopes in the model for which NNCF handling should be applied. Functions as
-    an "allowlist". If left unspecified, everything will be targeted.
 
     :return: A model wrapped by NNCFNetwork, which is ready for adding compression."""
 
@@ -251,6 +241,11 @@ def create_nncf_network(
         # model that are used on training stage only (e.g. AuxLogits of Inception-v3 model) or unused modules with
         # weights. As a consequence, no need to care about spoiling BN statistics, as they're disabled in eval mode.
 
+        input_info = get_input_info_from_config(config)
+        scopes_without_shape_matching = config.get("scopes_without_shape_matching", [])
+        ignored_scopes = config.get("ignored_scopes")
+        target_scopes = config.get("target_scopes")
+
         nncf_network = NNCFNetwork(
             model,
             input_info=input_info,

tests/torch/sparsity/movement/helpers/run_recipe.py

@@ -442,7 +442,7 @@ class ClipVisionRunRecipe(BaseMockRunRecipe):
     def model_input_info(self) -> FillerInputInfo:
         num_channels = self.model_config.num_channels
         image_size = self.model_config.image_size
-        return [FillerInputElement(shape=[1, num_channels, image_size, image_size], type_str="float")]
+        return FillerInputInfo([FillerInputElement(shape=[1, num_channels, image_size, image_size], type_str="float")])
 
     @staticmethod
     def get_nncf_modules_in_transformer_block_order(compressed_model: NNCFNetwork) -> List[DictInTransformerBlockOrder]: