ConstantLayerAttributes

nncf/experimental/torch2/function_hook/extractor.py

@@ -33,12 +33,6 @@
     om.PTDepthwiseConv3dSubtype,
 )
 
-CONV_TRANSPOSE_METATYPES = (
-    om.PTConvTranspose1dMetatype,
-    om.PTConvTranspose2dMetatype,
-    om.PTConvTranspose3dMetatype,
-)
-
 
 class ExtractedFunc(nn.Module):
     """
@@ -58,9 +52,18 @@ def forward(self, x: torch.Tensor) -> torch.Tensor:
         return self.fn(x, **self.kwargs)
 
 
-def apply_args_defaults(
+def apply_args_to_kwargs(
     args: Sequence[Any], kwargs: Dict[str, Any], indexed_args: List[Tuple[int, str]]
 ) -> Dict[str, Any]:
+    """
+    Applies the given arguments and keyword arguments to a dictionary of keyword arguments.
+
+    :param args: The positional arguments.
+    :param kwargs: The keyword arguments.
+    :param indexed_args: The list of pairs of indexes and names.
+
+    :return: A dictionary of keyword arguments with the applied arguments and keyword arguments.
+    """
     args_dict: Dict[str, Any] = dict()
     for idx, arg_name in indexed_args:
         if idx < len(args):
@@ -91,8 +94,10 @@ def extract_bn(model: nn.Module, graph: PTNNCFGraph, node: NNCFNode) -> Extracte
     running_mean = get_const_data_on_port(model, graph, node, 3)
     running_var = get_const_data_on_port(model, graph, node, 4)
 
-    bn_kwargs = apply_args_defaults(
-        layer_attr.op_args, layer_attr.op_kwargs, [(6, "momentum"), (7, "eps"), (8, "cudnn_enabled")]
+    bn_kwargs = apply_args_to_kwargs(
+        layer_attr.op_args,
+        layer_attr.op_kwargs,
+        [(6, "momentum"), (7, "eps"), (8, "cudnn_enabled")],
     )
     bn_kwargs["weight"] = weight
     bn_kwargs["bias"] = bias
@@ -120,7 +125,8 @@ def extract_conv(
     """
     weight_node = get_const_node(input_node, 1, graph)
     if weight_node is None:
-        raise nncf.InternalError(f"Weight node not found for {input_node}")
+        msg = "Weight node not found for {input_node}"
+        raise nncf.InternalError(msg)
     weight = get_const_data(weight_node, model)
 
     hook_storage = get_hook_storage(model)
@@ -138,8 +144,10 @@ def extract_conv(
         msg = f"Expected PT2OpLayerAttributes for input_node.layer_attributes, actual: {type(layer_attrs)}"
         raise nncf.InternalError(msg)
 
-    conv_kwargs = apply_args_defaults(
-        layer_attrs.op_args, layer_attrs.op_kwargs, [(3, "stride"), (4, "padding"), (5, "dilation"), (6, "groups")]
+    conv_kwargs = apply_args_to_kwargs(
+        layer_attrs.op_args,
+        layer_attrs.op_kwargs,
+        [(3, "stride"), (4, "padding"), (5, "dilation"), (6, "groups")],
     )
     conv_kwargs["weight"] = weight
     conv_kwargs["bias"] = bias
@@ -178,7 +186,8 @@ def extract_model(
     """
 
     if len(input_nodes) != 1 or len(output_nodes) != 1:
-        raise nncf.InternalError("input_nodes and output_nodes should contain only one node.")
+        msg = "input_nodes and output_nodes should contain only one node."
+        raise nncf.InternalError(msg)
 
     input_node = graph.get_node_by_name(input_nodes[0])
     output_node = graph.get_node_by_name(output_nodes[0])

nncf/experimental/torch2/function_hook/nncf_graph/nncf_graph_builder.py

@@ -21,6 +21,7 @@
 import nncf.torch.graph.operator_metatypes as om
 from nncf.common.graph.graph import NNCFNode
 from nncf.common.graph.layer_attributes import BaseLayerAttributes
+from nncf.common.graph.layer_attributes import ConstantLayerAttributes
 from nncf.common.graph.layer_attributes import Dtype
 from nncf.experimental.torch2.function_hook.graph.build_graph_mode import build_graph
 from nncf.experimental.torch2.function_hook.graph.graph_utils import ConstMeta
@@ -157,7 +158,8 @@ def get_layer_attributes(
     if isinstance(meta, FunctionMeta):
         constant_port_ids = get_constant_port_ids(nx_graph, node)
         return PT2OpLayerAttributes(meta.func, meta.args, meta.kwargs, constant_port_ids)
-
+    if isinstance(meta, ConstMeta):
+        return ConstantLayerAttributes(meta.name_in_model, meta.shape)
     return None
 
 

nncf/torch/model_graph_manager.py

@@ -19,7 +19,6 @@
 from nncf.common.graph.graph import NNCFNode
 from nncf.common.graph.operator_metatypes import CONST_NOOP_METATYPES
 from nncf.common.graph.operator_metatypes import OperatorMetatype
-from nncf.experimental.common.check_feature import is_experimental_torch_tracing_enabled
 from nncf.torch.dynamic_graph.context import PreHookId
 from nncf.torch.external_hook import ExternalOpCallHook
 from nncf.torch.graph import operator_metatypes as om
@@ -126,10 +125,7 @@ def get_const_data(const_node: NNCFNode, model: nn.Module) -> torch.Tensor:
     :param model: The NNCFNetwork object.
     :return: A torch.Tensor object containing the constant value.
     """
-    if is_experimental_torch_tracing_enabled():
-        const_name = const_node.layer_name
-    else:
-        const_name = const_node.layer_attributes.name
+    const_name = const_node.layer_attributes.name
     module_name, const_attr_name = split_const_name(const_name)
     module = get_module_by_name(module_name, model)
     data = getattr(module, const_attr_name)
@@ -265,10 +261,7 @@ def set_const_data(data: torch.Tensor, const_node: NNCFNode, model: NNCFNetwork)
     :param const_node: The NNCF node representing the constant data.
     :param model: The NNCF network model.
     """
-    if is_experimental_torch_tracing_enabled():
-        const_name = const_node.layer_name
-    else:
-        const_name = const_node.layer_attributes.name
+    const_name = const_node.layer_attributes.name
     module_name, const_attr_name = split_const_name(const_name)
     module = get_module_by_name(module_name, model)
     const = getattr(module, const_attr_name)
@@ -289,15 +282,11 @@ def set_const_data_to_port_id(
     :param const_port_id: The input port id of the node that receives the constant.
     :param model: The NNCF network containing the module to be modified.
     """
-    # graph = model.nncf.get_graph()
     const_node = get_const_node(node, port_id, graph)
     if const_node is None:
         msg = f"No found node with constant for {node.node_name} on {port_id} port"
         raise nncf.InternalError(msg)
-    if is_experimental_torch_tracing_enabled():
-        const_name = const_node.layer_name
-    else:
-        const_name = const_node.layer_attributes.name
+    const_name = const_node.layer_attributes.name
     module_name, const_attr_name = split_const_name(const_name)
     module = get_module_by_name(module_name, model)
     const = getattr(module, const_attr_name)

tests/torch2/function_hook/quantization/test_fast_bias_correction.py

@@ -58,7 +58,8 @@ def check_bias(model: GraphModelWrapper, ref_bias: list):
             # TODO(AlexanderDokuchaev): return atol=0.0001 after fix 109189
             assert torch.all(torch.isclose(bias_value, ref_bias, atol=0.02)), f"{bias_value} != {ref_bias}"
             return
-        raise ValueError("Not found node with bias")
+        msg = "Not found node with bias"
+        raise ValueError(msg)
 
 
 @pytest.mark.cuda
@@ -87,4 +88,5 @@ def check_bias(model: GraphModelWrapper, ref_bias: list):
             # TODO(AlexanderDokuchaev): return atol=0.0001 after fix 109189
             assert torch.all(torch.isclose(bias_value, ref_bias, atol=0.02)), f"{bias_value} != {ref_bias}"
             return
-        raise ValueError("Not found node with bias")
+        msg = "Not found node with bias"
+        raise ValueError(msg)