Comments

Author: daniil-lyakhov

nncf/quantization/algorithms/min_max/torch_backend.py

@@ -51,7 +51,7 @@
 from nncf.torch.quantization.layers import BaseQuantizer
 from nncf.torch.quantization.layers import PTQuantizerSpec
 from nncf.torch.quantization.layers import get_scale_shape
-from nncf.torch.utils import get_weight_nodes_in_inference_grpah
+from nncf.torch.utils import get_weight_nodes_in_inference_graph
 
 
 class PTMinMaxAlgoBackend(MinMaxAlgoBackend):
@@ -342,7 +342,7 @@ def get_ignored_names_by_layer_attributes(nncf_graph: NNCFGraph) -> Set[str]:
         return set()
 
     def get_weight_nodes(self, inference_nncf_graph: NNCFGraph) -> List[NNCFNode]:
-        return get_weight_nodes_in_inference_grpah(inference_nncf_graph, self.mat_mul_metatypes)
+        return get_weight_nodes_in_inference_graph(inference_nncf_graph, self.mat_mul_metatypes)
 
     def is_matmul_with_constant(self, node: NNCFNode, nncf_graph: NNCFGraph) -> bool:
         return node.metatype in self.mat_mul_metatypes and len(get_weight_tensor_port_ids(node, nncf_graph)) > 0

nncf/quantization/algorithms/min_max/torch_fx_backend.py

@@ -49,7 +49,7 @@
 from nncf.torch.quantization.layers import PTQuantizerSpec
 from nncf.torch.quantization.layers import get_scale_shape
 from nncf.torch.quantization.strip import convert_to_torch_fakequantizer
-from nncf.torch.utils import get_weight_nodes_in_inference_grpah
+from nncf.torch.utils import get_weight_nodes_in_inference_graph
 
 
 class FXMinMaxAlgoBackend(MinMaxAlgoBackend):
@@ -306,7 +306,7 @@ def get_ignored_names_by_layer_attributes(nncf_graph: NNCFGraph) -> Set[str]:
         return set()
 
     def get_weight_nodes(self, inference_nncf_graph: NNCFGraph) -> List[NNCFNode]:
-        return get_weight_nodes_in_inference_grpah(inference_nncf_graph, self.mat_mul_metatypes)
+        return get_weight_nodes_in_inference_graph(inference_nncf_graph, self.mat_mul_metatypes)
 
     def is_matmul_with_constant(self, node: NNCFNode, nncf_graph: NNCFGraph) -> bool:
         return node.metatype in self.mat_mul_metatypes and len(get_weight_tensor_port_ids(node, nncf_graph)) > 0

nncf/torch/utils.py

@@ -11,7 +11,7 @@
 import random
 from collections import OrderedDict
 from contextlib import contextmanager
-from typing import Any, Dict, Generator, List
+from typing import Any, Dict, Generator, List, Type
 
 import numpy as np
 import torch
@@ -472,7 +472,7 @@ def get_model_dtype(model: torch.nn.Module) -> torch.dtype:
     return dtype
 
 
-def get_weight_nodes_in_inference_grpah(
+def get_weight_nodes_in_inference_graph(
     inference_nncf_graph: NNCFGraph, mat_mul_metatypes: List[Type[om.PTOperatorMetatype]]
 ) -> List[NNCFNode]:
     """
@@ -510,6 +510,7 @@ def is_matmul_with_constant_in_inference_graph(node: NNCFNode, inference_nncf_gr
 
     # Inference graph does not contain constants, so
     # any missed input edge means it is a constant branch.
-    return node.metatype in [om.PTMatMulMetatype, om.PTAddmmMetatype] and len(
-        inference_nncf_graph.get_input_edges(node)
-    ) < len(node.metatype.weight_port_ids)
+    is_matmul_metatype = node.metatype in [om.PTMatMulMetatype, om.PTAddmmMetatype]
+    inputs_missed = 1 <= len(inference_nncf_graph.get_input_edges(node)) < len(node.metatype.weight_port_ids)
+
+    return is_matmul_metatype and inputs_missed