Move registry for typing

NVIDIA · Sep 11, 2024 · 9b1df32 · 9b1df32
1 parent 9a389f5
commit 9b1df32
Show file tree

Hide file tree

Showing 5 changed files with 49 additions and 34 deletions.
diff --git a/tripy/tests/test_function_registry.py b/tripy/tests/test_function_registry.py
@@ -318,7 +318,7 @@ def func(n: Sequence[Union[int, Sequence[int]]]) -> int:
 
     def test_tensor_literal(self, registry):
         @registry("test")
-        def func(n: "tripy.TensorLiteral.sig"):
+        def func(n: "tripy.types.tensor_literal"):
             return n
 
         assert registry["test"](1) == 1
@@ -472,7 +472,7 @@ def func(n: Sequence[Union[int, float]]) -> int:
 
     def test_error_tensor_literal_not_sequence(self, registry):
         @registry("test")
-        def func(n: "tripy.TensorLiteral.sig"):
+        def func(n: "tripy.types.tensor_literal"):
             return n
 
         with helper.raises(
@@ -484,19 +484,19 @@ def func(n: "tripy.TensorLiteral.sig"):
 
                 --> \x1b\[38;5;3m{__file__}\x1b\[0m:[0-9]+ in \x1b\[38;5;6mfunc\(\)\x1b\[0m
                       \|
-                  [0-9]+ \|         def func\(n: \"tripy\.TensorLiteral\.sig\"\):
+                  [0-9]+ \|         def func\(n: \"tripy\.types\.tensor_literal\"\):
                   [0-9]+ \|     \.\.\.
                       \|\s
 
-                Not a valid overload because: For parameter: 'n', expected an instance of type: 'typing\.Union\[numbers\.Number, typing\.Sequence\[ForwardRef\('tripy\.TensorLiteral\.sig'\)\]\]' but got argument of type: 'str'\.
+                Not a valid overload because: For parameter: 'n', expected an instance of type: 'typing\.Union\[numbers\.Number, typing\.Sequence\[ForwardRef\('tripy\.types\.tensor_literal'\)\]\]' but got argument of type: 'str'\.
             """
             ).strip(),
         ):
             registry["test"]("hi")
 
     def test_error_tensor_literal_not_sequence_of_numbers(self, registry):
         @registry("test")
-        def func(n: "tripy.TensorLiteral.sig"):
+        def func(n: "tripy.types.tensor_literal"):
             return n
 
         with helper.raises(
@@ -508,11 +508,11 @@ def func(n: "tripy.TensorLiteral.sig"):
 
                 --> \x1b\[38;5;3m{__file__}\x1b\[0m:[0-9]+ in \x1b\[38;5;6mfunc\(\)\x1b\[0m
                       \|
-                  [0-9]+ \|         def func\(n: \"tripy\.TensorLiteral\.sig\"\):
+                  [0-9]+ \|         def func\(n: \"tripy\.types\.tensor_literal\"\):
                   [0-9]+ \|     \.\.\.
                       \|\s
 
-                Not a valid overload because: For parameter: 'n', expected an instance of type: 'typing\.Union\[numbers\.Number, typing\.Sequence\[ForwardRef\('tripy\.TensorLiteral\.sig'\)\]\]' but got argument of type: 'List\[List\[str\]\]'
+                Not a valid overload because: For parameter: 'n', expected an instance of type: 'typing\.Union\[numbers\.Number, typing\.Sequence\[ForwardRef\('tripy\.types\.tensor_literal'\)\]\]' but got argument of type: 'List\[List\[str\]\]'
             """
             ).strip(),
         ):

diff --git a/tripy/tripy/frontend/trace/ops/binary_elementwise.py b/tripy/tripy/frontend/trace/ops/binary_elementwise.py
@@ -184,7 +184,8 @@ def to_flat_ir(self, inputs, outputs):
     function_name="__radd__",
 )
 def __add__(
-    self: Union["tripy.Tensor", "tripy.TensorLiteral.sig"], other: Union["tripy.Tensor", "tripy.TensorLiteral.sig"]
+    self: Union["tripy.Tensor", "tripy.types.tensor_literal"],
+    other: Union["tripy.Tensor", "tripy.types.tensor_literal"],
 ) -> "tripy.Tensor":
     """
     Performs an elementwise sum.
@@ -217,7 +218,8 @@ def __add__(
     dtype_constraints={"self": "T1", "other": "T1", constraints.RETURN_VALUE: "T1"},
 )
 def __sub__(
-    self: Union["tripy.Tensor", "tripy.TensorLiteral.sig"], other: Union["tripy.Tensor", "tripy.TensorLiteral.sig"]
+    self: Union["tripy.Tensor", "tripy.types.tensor_literal"],
+    other: Union["tripy.Tensor", "tripy.types.tensor_literal"],
 ) -> "tripy.Tensor":
     """
     Performs an elementwise subtraction.
@@ -250,7 +252,7 @@ def __sub__(
     dtype_constraints={"other": "T1", constraints.RETURN_VALUE: "T1"},
 )
 def __rsub__(
-    self: "tripy.TensorLiteral.sig", other: Union["tripy.Tensor", "tripy.TensorLiteral.sig"]
+    self: "tripy.types.tensor_literal", other: Union["tripy.Tensor", "tripy.types.tensor_literal"]
 ) -> "tripy.Tensor":
     """
     Performs an elementwise subtraction.
@@ -283,7 +285,8 @@ def __rsub__(
     dtype_constraints={"self": "T1", "other": "T1", constraints.RETURN_VALUE: "T1"},
 )
 def __pow__(
-    self: Union["tripy.Tensor", "tripy.TensorLiteral.sig"], other: Union["tripy.Tensor", "tripy.TensorLiteral.sig"]
+    self: Union["tripy.Tensor", "tripy.types.tensor_literal"],
+    other: Union["tripy.Tensor", "tripy.types.tensor_literal"],
 ) -> "tripy.Tensor":
     """
     Performs an elementwise exponentiation.
@@ -354,7 +357,8 @@ def __rpow__(self: numbers.Number, other: Union["tripy.Tensor", Any]) -> "tripy.
     function_name="__rmul__",
 )
 def __mul__(
-    self: Union["tripy.Tensor", "tripy.TensorLiteral.sig"], other: Union["tripy.Tensor", "tripy.TensorLiteral.sig"]
+    self: Union["tripy.Tensor", "tripy.types.tensor_literal"],
+    other: Union["tripy.Tensor", "tripy.types.tensor_literal"],
 ) -> "tripy.Tensor":
     """
     Performs an elementwise multiplication.
@@ -387,7 +391,8 @@ def __mul__(
     dtype_constraints={"self": "T1", "other": "T1", constraints.RETURN_VALUE: "T1"},
 )
 def __truediv__(
-    self: Union["tripy.Tensor", "tripy.TensorLiteral.sig"], other: Union["tripy.Tensor", "tripy.TensorLiteral.sig"]
+    self: Union["tripy.Tensor", "tripy.types.tensor_literal"],
+    other: Union["tripy.Tensor", "tripy.types.tensor_literal"],
 ) -> "tripy.Tensor":
     """
     Performs an elementwise division.
@@ -419,7 +424,7 @@ def __truediv__(
     dtype_variables={"T1": ["float32", "float16", "bfloat16", "int8", "int32", "int64"]},
     dtype_constraints={"other": "T1", constraints.RETURN_VALUE: "T1"},
 )
-def __rtruediv__(self: numbers.Number, other: Union["tripy.Tensor", "tripy.TensorLiteral.sig"]) -> "tripy.Tensor":
+def __rtruediv__(self: numbers.Number, other: Union["tripy.Tensor", "tripy.types.tensor_literal"]) -> "tripy.Tensor":
     """
     Performs an elementwise division.
 
@@ -451,7 +456,7 @@ def __rtruediv__(self: numbers.Number, other: Union["tripy.Tensor", "tripy.Tenso
     dtype_constraints={"lhs": "T1", "rhs": "T1", constraints.RETURN_VALUE: "T1"},
 )
 def maximum(
-    lhs: Union["tripy.Tensor", "tripy.TensorLiteral.sig"], rhs: Union["tripy.Tensor", "tripy.TensorLiteral.sig"]
+    lhs: Union["tripy.Tensor", "tripy.types.tensor_literal"], rhs: Union["tripy.Tensor", "tripy.types.tensor_literal"]
 ) -> "tripy.Tensor":
     """
     Performs an elementwise maximum.
@@ -484,7 +489,7 @@ def maximum(
     dtype_constraints={"lhs": "T1", "rhs": "T1", constraints.RETURN_VALUE: "T1"},
 )
 def minimum(
-    lhs: Union["tripy.Tensor", "tripy.TensorLiteral.sig"], rhs: Union["tripy.Tensor", "tripy.TensorLiteral.sig"]
+    lhs: Union["tripy.Tensor", "tripy.types.tensor_literal"], rhs: Union["tripy.Tensor", "tripy.types.tensor_literal"]
 ) -> "tripy.Tensor":
     """
     Performs an elementwise minimum.
@@ -520,7 +525,8 @@ def minimum(
     dtype_constraints={"self": "T1", "other": "T1", constraints.RETURN_VALUE: "T2"},
 )
 def __lt__(
-    self: Union["tripy.Tensor", "tripy.TensorLiteral.sig"], other: Union["tripy.Tensor", "tripy.TensorLiteral.sig"]
+    self: Union["tripy.Tensor", "tripy.types.tensor_literal"],
+    other: Union["tripy.Tensor", "tripy.types.tensor_literal"],
 ) -> "tripy.Tensor":
     """
     Performs a 'less than' comparison.
@@ -556,7 +562,8 @@ def __lt__(
     dtype_constraints={"self": "T1", "other": "T1", constraints.RETURN_VALUE: "T2"},
 )
 def __le__(
-    self: Union["tripy.Tensor", "tripy.TensorLiteral.sig"], other: Union["tripy.Tensor", "tripy.TensorLiteral.sig"]
+    self: Union["tripy.Tensor", "tripy.types.tensor_literal"],
+    other: Union["tripy.Tensor", "tripy.types.tensor_literal"],
 ) -> "tripy.Tensor":
     """
     Performs a 'less than or equal' comparison.
@@ -592,7 +599,8 @@ def __le__(
     dtype_constraints={"self": "T1", "other": "T1", constraints.RETURN_VALUE: "T2"},
 )
 def __eq__(
-    self: Union["tripy.Tensor", "tripy.TensorLiteral.sig"], other: Union["tripy.Tensor", "tripy.TensorLiteral.sig"]
+    self: Union["tripy.Tensor", "tripy.types.tensor_literal"],
+    other: Union["tripy.Tensor", "tripy.types.tensor_literal"],
 ) -> "tripy.Tensor":
     """
     Performs an 'equal' comparison.
@@ -627,7 +635,9 @@ def __eq__(
     },
     dtype_constraints={"self": "T1", "other": "T1", constraints.RETURN_VALUE: "T2"},
 )
-def __ne__(self: Union["tripy.Tensor", "tripy.TensorLiteral.sig"], other: Union["tripy.Tensor", Any]) -> "tripy.Tensor":
+def __ne__(
+    self: Union["tripy.Tensor", "tripy.types.tensor_literal"], other: Union["tripy.Tensor", Any]
+) -> "tripy.Tensor":
     """
     Performs a 'not equal' comparison.
 
@@ -662,7 +672,8 @@ def __ne__(self: Union["tripy.Tensor", "tripy.TensorLiteral.sig"], other: Union[
     dtype_constraints={"self": "T1", "other": "T1", constraints.RETURN_VALUE: "T2"},
 )
 def __ge__(
-    self: Union["tripy.Tensor", "tripy.TensorLiteral.sig"], other: Union["tripy.Tensor", "tripy.TensorLiteral.sig"]
+    self: Union["tripy.Tensor", "tripy.types.tensor_literal"],
+    other: Union["tripy.Tensor", "tripy.types.tensor_literal"],
 ) -> "tripy.Tensor":
     """
     Performs a 'greater than or equal' comparison.
@@ -698,7 +709,8 @@ def __ge__(
     dtype_constraints={"self": "T1", "other": "T1", constraints.RETURN_VALUE: "T2"},
 )
 def __gt__(
-    self: Union["tripy.Tensor", "tripy.TensorLiteral.sig"], other: Union["tripy.Tensor", "tripy.TensorLiteral.sig"]
+    self: Union["tripy.Tensor", "tripy.types.tensor_literal"],
+    other: Union["tripy.Tensor", "tripy.types.tensor_literal"],
 ) -> "tripy.Tensor":
     """
     Performs a 'greater than' comparison.

diff --git a/tripy/tripy/frontend/trace/ops/dequantize.py b/tripy/tripy/frontend/trace/ops/dequantize.py
@@ -111,7 +111,7 @@ def to_flat_ir(self, inputs, outputs):
 @frontend_utils.convert_inputs_to_tensors(exclude=["dtype", "dim"])
 def dequantize(
     input: "tripy.Tensor",
-    scale: Union["tripy.Tensor", "tripy.TensorLiteral.sig"],
+    scale: Union["tripy.Tensor", "tripy.types.tensor_literal"],
     dtype: datatype.dtype,
     dim: Union[int, Any] = None,
 ) -> "tripy.Tensor":

diff --git a/tripy/tripy/frontend/trace/ops/quantize.py b/tripy/tripy/frontend/trace/ops/quantize.py
@@ -135,7 +135,7 @@ def to_flat_ir(self, inputs, outputs):
 )
 def quantize(
     input: "tripy.Tensor",
-    scale: Union["tripy.Tensor", "tripy.TensorLiteral.sig"],
+    scale: Union["tripy.Tensor", "tripy.types.tensor_literal"],
     dtype: datatype.dtype,
     dim: Union[int, Any] = None,
 ) -> "tripy.Tensor":

diff --git a/tripy/tripy/common/types.py → tripy/tripy/types.py b/tripy/tripy/common/types.py → tripy/tripy/types.py
@@ -1,5 +1,5 @@
 #
-# SPDX-FileCopyrightText: Copyright (c) 1993-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-FileCopyrightText: Copyright (c) 2024-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
@@ -16,17 +16,20 @@
 #
 
 import numbers
+import sys
 from typing import Union, Sequence
 
 from tripy import export
 
+export.public_api()(sys.modules[__name__])
 
-@export.public_api(document_under="types.rst")
-class TensorLiteral:
-    """
-    The `sig` member of this class can be used as a type annotation for tensor literals.
-    A tensor literal can be a Python number or a sequence of tensor literals
-    (i.e., a sequence of numbers of any depth).
-    """
-
-    sig = Union[numbers.Number, Sequence["tripy.TensorLiteral.sig"]]
+tensor_literal = export.public_api(
+    document_under="types.rst",
+    module=sys.modules[__name__],
+    symbol="tensor_literal",
+)(Union[numbers.Number, Sequence["tripy.types.tensor_literal"]])
+"""
+Denotes the recursive type annotation for tensor literals.
+A tensor literal can be a Python number or a sequence of tensor literals
+(i.e., a sequence of numbers of any depth).
+"""