[SOT] use better Singleton

python/paddle/jit/sot/infer_meta.py

@@ -103,8 +103,7 @@ def __hash__(self):
         return hash((tuple(self.shape), self.dtype, self.stop_gradient))
 
 
-@Singleton
-class VariableCreator:
+class VariableCreator(metaclass=Singleton):
     """
     We use the static graph Variable to infer the meta information of Tensor.
     This singleton class is used to create Variable for infer meta.
@@ -305,8 +304,7 @@ def ast_infer_meta(static_function, *args, **kwargs):
     return out
 
 
-@Singleton
-class SpecialInferMeta:
+class SpecialInferMeta(metaclass=Singleton):
     """
     There are some functions that cannot be inferred directly through static graph,
     and need to be implemented manually. This class is used to implement infer meta
@@ -340,8 +338,7 @@ def infermeta_grad(
         return inputs
 
 
-@Singleton
-class InferMetaCache(Cache):
+class InferMetaCache(Cache, metaclass=Singleton):
     def key_fn(
         self, func, *args, **kwargs
     ):  # args & kwargs have transformed to MetaInfo
@@ -362,8 +359,7 @@ def value_fn(self, func, *args, **kwargs):
         return infer_meta(func, *args, **kwargs)
 
 
-@Singleton
-class LayerInferMetaCache(Cache):
+class LayerInferMetaCache(Cache, metaclass=Singleton):
     def key_fn(self, layer, *args, **kwargs):
         params = [
             MetaInfo.from_tensor(x)

python/paddle/jit/sot/opcode_translator/breakpoint.py

@@ -36,8 +36,7 @@ def __hash__(self):
         return hash((self.file, self.line, self.co_name, self.offset))
 
 
-@Singleton
-class BreakpointManager:
+class BreakpointManager(metaclass=Singleton):
     def __init__(self):
         self.breakpoints = set()
         self.executors = OpcodeExecutorBase.call_stack

python/paddle/jit/sot/opcode_translator/executor/executor_cache.py

@@ -42,8 +42,7 @@
 dummy_guard.lambda_expr = "lambda frame: True"
 
 
-@Singleton
-class OpcodeExecutorCache:
+class OpcodeExecutorCache(metaclass=Singleton):
     """
     A singleton class that implements a cache for translated instructions.
     This cache is used to store previously translated instructions along with their corresponding guard functions.

python/paddle/jit/sot/opcode_translator/instruction_utils/stack_analyse.py

@@ -15,8 +15,7 @@
 from ...utils import Singleton
 
 
-@Singleton
-class StackAnalyser:
+class StackAnalyser(metaclass=Singleton):
     def stack_effect(self, instr):
         if "BINARY" in instr.opname or "INPLACE" in instr.opname:
             return 2, 1

python/paddle/jit/sot/symbolic/compile_cache.py

@@ -154,8 +154,7 @@ def __call__(self, *args, **kwargs):
             return outputs
 
 
-@Singleton
-class CompileSIRCache(Cache):
+class CompileSIRCache(Cache, metaclass=Singleton):
     """
     Cache the compiled function of SIR
     """

python/paddle/jit/sot/symbolic/statement_ir.py

@@ -283,8 +283,7 @@ def __repr__(self):
         return self.__str__()
 
 
-@Singleton
-class StatementIRFactory:
+class StatementIRFactory(metaclass=Singleton):
     """
     It is used to create a StatementIR.
     """
@@ -320,8 +319,7 @@ def clear(self):
             del self.cache[key]
 
 
-@Singleton
-class SIRRuntimeCache:
+class SIRRuntimeCache(metaclass=Singleton):
     """
     It is used to cache the runtime information of the StatementIR.
     """

python/paddle/jit/sot/utils/call_ast_utils.py

@@ -40,8 +40,7 @@ def _is_wrapped(f):
     return func
 
 
-@Singleton
-class StaticFunctionManager:
+class StaticFunctionManager(metaclass=Singleton):
     def __init__(self):
         self.code_map = {}
 

python/paddle/jit/sot/utils/utils.py

@@ -23,7 +23,7 @@
 from collections import OrderedDict
 from contextlib import contextmanager
 from enum import Enum
-from typing import Any, Generic, TypeVar
+from typing import Any, TypeVar
 from weakref import WeakValueDictionary
 
 import numpy as np
@@ -48,15 +48,13 @@
 ConstTypes = (int, float, str, bool, type(None))
 
 
-class Singleton(Generic[T]):
-    def __init__(self, cls: type[T]):
-        self._cls = cls
-        self._instance = {}
+class Singleton(type):
+    _instances: dict[Any, Any] = {}
 
-    def __call__(self) -> T:
-        if self._cls not in self._instance:
-            self._instance[self._cls] = self._cls()
-        return self._instance[self._cls]
+    def __call__(cls, *args: Any, **kwargs: Any):
+        if cls not in cls._instances:
+            cls._instances[cls] = super().__call__(*args, **kwargs)
+        return cls._instances[cls]
 
 
 class NameGenerator:
@@ -107,8 +105,7 @@ def current_tmp_name_records():
     return _tmp_name_records
 
 
-@Singleton
-class ResumeFnNameFactory:
+class ResumeFnNameFactory(metaclass=Singleton):
     def __init__(self) -> None:
         self.gen = NameGenerator('resume_')
 
@@ -316,8 +313,7 @@ def get_unbound_method(obj, name):
     return getattr(obj.__class__, name)
 
 
-@Singleton
-class GraphLogger:
+class GraphLogger(metaclass=Singleton):
     graph_num: int
     op_num: int
     graphs: list[Program]
@@ -376,8 +372,7 @@ def print_info(self):
         print(self)
 
 
-@Singleton
-class SotUndefinedVar:
+class SotUndefinedVar(metaclass=Singleton):
     pass
 
 
@@ -461,8 +456,7 @@ def need_dynamic_info(self):
         return len(self.dyn_time_costs) < self.REQUIRED_DYN_INFOS
 
 
-@Singleton
-class StepInfoManager:
+class StepInfoManager(metaclass=Singleton):
     def __init__(self):
         self.step_record = {}
         self.current_code = None