[TorchFX] Constant folding (#3047)

### Changes

Constant folding is enabled by default in TorchFX backend

### Reason for changes

To align quantizers placement between OV and TorchFX

### Related tickets

#2766 

### Tests

* test_constant_folding
*  test_constant_folding_with_constraints
* test_models.py references are updated
* post_training_quantization/535/ - finished successfully

---------

Co-authored-by: Alexander Suslov <[email protected]>
Co-authored-by: Aamir Nazir <[email protected]>

nncf/experimental/torch/fx/constant_folding.py

@@ -0,0 +1,257 @@
+# Copyright (c) 2024 Intel Corporation
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#      http://www.apache.org/licenses/LICENSE-2.0
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import collections
+from typing import Any, Callable, Dict, List, Optional
+
+import torch.fx
+import torch.utils._pytree as pytree
+
+aten = torch.ops.aten
+
+
+def _replace_node_with_constant(
+    gm: torch.fx.GraphModule,
+    node: torch.fx.Node,
+    value: torch.Tensor,
+) -> None:
+    """
+    Replaces given node to a constant with given value.
+
+    :param gm: GraphModule instance.
+    :param node: A node to replace with a constant.
+    :param value: Value to use as a constant instead of the given node.
+    """
+    g = gm.graph
+
+    if not hasattr(gm, "_frozen_param_count"):
+        gm._frozen_param_count = 0  # type: ignore[assignment]
+    i = gm._frozen_param_count
+
+    while True:
+        qualname = f"_frozen_param{i}"
+        if not hasattr(gm, qualname):
+            break
+        i += 1
+
+    gm._frozen_param_count = i + 1
+
+    with g.inserting_before(node):
+        new_input_node = g.create_node("get_attr", qualname, (), {})
+        node.replace_all_uses_with(new_input_node)
+        new_input_node.meta.update(node.meta)
+        g.erase_node(node)
+
+    # needed to suppress `does not reference an nn.Module, nn.Parameter, or buffer` warning
+    gm.register_buffer(qualname, value)
+    setattr(gm, qualname, value)
+
+
+def _is_const_source(node: torch.fx.Node) -> bool:
+    """
+    Return True if node is a constant node.
+
+    :param node: Given node.
+    :return: True if node is a constant node.
+    """
+    return node.op == "get_attr"
+
+
+class ConstantFolder(torch.fx.Interpreter):
+    """
+    Finds constant subraphs and place it in node_replacement attribute.
+    """
+
+    def __init__(
+        self,
+        gm: torch.fx.GraphModule,
+    ) -> None:
+        super().__init__(gm)
+        self.node_replacements: Dict[torch.fx.Node, Any] = {}
+        self.replaced_uses: Dict[torch.fx.Node, int] = collections.Counter()
+        self.unknown_value = object()
+
+        # overwrite this to deallocate env values if their only remaining use
+        # is the output
+        self.user_to_last_uses = self.node_to_last_non_output_use()
+
+    def _deduce_value(self, node: torch.fx.Node) -> Any:
+        return super().run_node(node)
+
+    def is_impure(self, node: torch.fx.node.Node) -> bool:
+        def is_woq_int8_pattern(node: torch.fx.node.Node) -> bool:
+            return (
+                node.target == torch.ops.prims.convert_element_type.default  # type: ignore[return-value]
+                and isinstance(node.args[0], torch.fx.Node)
+                and "val" in node.args[0].meta
+                and node.args[0].meta["val"].dtype == torch.int8  # type: ignore[union-attr]
+                and node.args[1] == torch.bfloat16
+            )
+
+        if (
+            is_woq_int8_pattern(node)
+            or (
+                node.target == torch.ops.aten.permute.default
+                and len(node.users) == 1
+                and is_woq_int8_pattern(next(iter(node.users)))
+            )
+        ) and _is_const_source(node.args[0]):
+            # Case 1: int8_weight -> dq -> bf16_weight
+            # Case 2: int8_weight -> permute -> dq -> bf16_weight
+            return True
+
+        quant_registered = getattr(torch.ops.quantized_decomposed, "dequantize_per_channel", None) is not None
+        if quant_registered and node.target in [
+            torch.ops.quantized_decomposed.dequantize_per_channel.default,
+            torch.ops.quantized_decomposed.dequantize_per_tensor.default,
+            torch.ops.quantized_decomposed.dequantize_per_tensor.tensor,
+        ]:
+            # For the pattern fp32_weight -> q -> dq
+            # We only folding fp32_weight -> q
+            # int8_weight and leave dq in graph to be fused
+            return True
+        return False
+
+    def node_to_last_non_output_use(self) -> Dict[torch.fx.Node, List[torch.fx.Node]]:
+        last_non_output_use = collections.defaultdict(list)
+        seen_uses = set()
+        output_node = next(iter(reversed(self.module.graph.nodes)))
+
+        for node in reversed(self.module.graph.nodes):
+            if node.target == "output":
+                continue
+
+            def add_use(inp: torch.fx.Node) -> None:
+                if inp in seen_uses:
+                    return
+
+                seen_uses.add(inp)
+                last_non_output_use[node].append(inp)
+
+            # In-place is fine since we don't mutate
+            pytree.tree_map_only_(torch.fx.Node, add_use, (node.args, node.kwargs))
+
+            # if this node is only used in output, we want to gc it right away
+            if len(node.users) == 1 and output_node in node.users:
+                last_non_output_use[node].append(node)
+
+        return last_non_output_use
+
+    def run_node(self, node: torch.fx.Node) -> Any:
+        if node.target == "output":
+            # because we remove nodes from env on last non output use,
+            # re-define them now or we'll get error in interpreter
+            def set_env(arg: torch.fx.Node) -> None:
+                self.env[arg] = self.unknown_value
+
+            # In-place is fine since we don't mutate
+            pytree.tree_map_only_(torch.fx.Node, set_env, node.args)
+            return super().run_node(node)
+
+        args, kwargs = self.fetch_args_kwargs_from_env(node)
+        flattened_inputs = pytree.arg_tree_leaves(*args, **kwargs)
+
+        # We need to do this weird thing because in cases where flattened_inputs
+        # contains a ScriptObject, equality checking results in a type error if
+        # the types are different.
+        if any(
+            type(self.unknown_value) is type(input_) and self.unknown_value == input_ for input_ in flattened_inputs
+        ):
+            return self.unknown_value
+
+        if node.op == "call_function" and node.target == aten._efficientzerotensor.default:
+            return self.unknown_value
+
+        if node.op == "call_function" and node.name == "triton_kernel_wrapper_functional_proxy":
+            return self.unknown_value
+
+        # skip constructors, since inductor generates optimal code for them already
+        # and turning into tensor would result in an additional global memory read
+        if not _is_const_source(node) and not any(isinstance(e, torch.Tensor) for e in flattened_inputs):
+            return self.unknown_value
+
+        # All mutations should either be removed or on inputs which we did not make constant
+        if isinstance(node.target, torch._ops.OpOverload) and torch.Tag.nondeterministic_seeded in node.target.tags:
+            return self.unknown_value
+
+        out = self._deduce_value(node)
+        if out == self.unknown_value:
+            return self.unknown_value
+
+        if not _is_const_source(node) and isinstance(out, torch.Tensor):
+            if out.device.type == "meta":
+                return out
+
+            if self.is_impure(node):
+                return self.unknown_value
+
+            self.add_node_replacement(node, out)
+
+            flattened_node_inps = pytree.arg_tree_leaves(*node.args, **node.kwargs)
+
+            for n in flattened_node_inps:
+                if not isinstance(n, torch.fx.Node):
+                    continue
+
+                self.replaced_uses[n] += 1
+
+            for to_delete in self.user_to_last_uses.get(node, []):
+                if self.replaced_uses[to_delete] == len(to_delete.users):
+                    self.node_replacements.pop(to_delete, None)
+
+        return out
+
+    def add_node_replacement(self, node: torch.fx.Node, tensor: torch.Tensor) -> None:
+        self.node_replacements[node] = tensor
+
+    def run(self) -> Any:  # type: ignore[override]
+        env: Dict[torch.fx.Node, Any] = {}
+        self.insert_placerholder_values(env)
+        return super().run(initial_env=env)
+
+    def insert_placerholder_values(self, env: Dict[torch.fx.Node, Any]) -> None:
+        for n in self.module.graph.find_nodes(op="placeholder"):
+            env[n] = self.unknown_value  # type: ignore[assignment]
+
+
+def constant_fold(
+    gm: torch.fx.GraphModule,
+    constraint_fn: Optional[Callable[[torch.fx.Node], bool]] = None,
+) -> None:
+    """
+    Calcualtes constant subgraphs values and replaces them with a constant node inplace.
+
+    :param gm: Given graph model.
+    :param constraint_fn: Constraint function which takes a node and returs the constraint:
+        should the node be constant folded or not.
+    """
+    with torch.utils._python_dispatch._disable_current_modes():
+        cf = ConstantFolder(gm)
+        cf.run()
+
+        for node, constant in cf.node_replacements.items():
+            if constraint_fn is not None and not constraint_fn(node):
+                continue
+            _replace_node_with_constant(gm, node, constant)
+
+        erased_params = []
+        for node in gm.graph.find_nodes(op="get_attr"):
+            if len(node.users) == 0:
+                if hasattr(gm, node.target):
+                    delattr(gm, node.target)
+                erased_params.append(node)
+
+        for node in erased_params:
+            gm.graph.erase_node(node)
+
+        gm.graph.eliminate_dead_code()
+        gm.graph.lint()
+        gm.recompile()

nncf/experimental/torch/fx/transformations.py

@@ -22,6 +22,7 @@
 import nncf.torch
 from nncf.common.graph.graph import NNCFNode
 from nncf.common.graph.transformations.commands import TargetType
+from nncf.experimental.torch.fx.constant_folding import constant_fold
 from nncf.experimental.torch.fx.node_utils import get_graph_node_by_name
 from nncf.experimental.torch.fx.node_utils import get_tensor_constant_from_node
 from nncf.torch.graph.transformations.commands import PTTargetPoint
@@ -802,12 +803,26 @@ def apply_quantization_transformations(model: torch.fx.GraphModule) -> None:
     # to make it easier for algorithms to work
     # with the target graph BatchNorm operations
     # are being fused
+    fold_constant_except_qdq(model)
     fuse_conv_bn(model)
     separate_conv_and_bias(model)
     separate_linear_and_bias(model)
     shared_constants_unification_transformation(model)
 
 
+def fold_constant_except_qdq(model: torch.fx.GraphModule):
+    """
+    Performs constant folding avoiding quantize-dequantize pattern.
+
+    :param model: Model to perform constant folding on.
+    """
+
+    def constraint_fn(node: torch.fx.Node):
+        return node.op != "call_function" or node.target not in QUANTIZE_NODE_TARGETS + DEQUANTIZE_NODE_TARGETS
+
+    constant_fold(model, constraint_fn=constraint_fn)
+
+
 def revert_quantization_transformations(model: torch.fx.GraphModule) -> None:
     """
     Reverts quantization transformations from the model.