[quant] Add quantizer class skeleton

Pull Request resolved: https://github.com/pytorch/pytorch/pull/79936

Approved by: https://github.com/jerryzh168

mypy.ini

@@ -61,6 +61,9 @@ ignore_missing_imports = True
 [mypy-torch.ao.quantization.experimental.apot_utils]
 ignore_missing_imports = True
 
+[mypy-torch.ao.quantization.experimental.quantizer]
+ignore_missing_imports = True
+
 #
 # Files with various errors. Mostly real errors, possibly some false
 # positives as well.

test/quantization/core/experimental/test_quantized_tensor.py

@@ -1,22 +1,13 @@
 # Owner(s): ["oncall: quantization"]
 
-import torch
+from torch.ao.quantization.experimental.quantizer import APoTQuantizer
-from torch.ao.quantization.experimental.APoT_tensor import TensorAPoT
 import unittest
 
 class TestQuantizedTensor(unittest.TestCase):
-    def test_quantize_APoT(self):
+    def test_int_repr(self):
-        t = torch.Tensor()
+        quantizer = APoTQuantizer()
         with self.assertRaises(NotImplementedError):
-            TensorAPoT.quantize_APoT(t)
+            quantizer.int_repr()
-
-    def test_dequantize(self):
-        with self.assertRaises(NotImplementedError):
-            TensorAPoT.dequantize(self)
-
-    def test_q_apot_alpha(self):
-        with self.assertRaises(NotImplementedError):
-            TensorAPoT.q_apot_alpha(self)
 
 if __name__ == '__main__':
     unittest.main()

test/quantization/core/experimental/test_quantizer.py

@@ -0,0 +1,22 @@
+# Owner(s): ["oncall: quantization"]
+
+import torch
+from torch.ao.quantization.experimental.quantizer import APoTQuantizer
+import unittest
+
+class TestQuantizer(unittest.TestCase):
+    def test_quantize_APoT(self):
+        t = torch.Tensor()
+        with self.assertRaises(NotImplementedError):
+            APoTQuantizer.quantize_APoT(t)
+
+    def test_dequantize(self):
+        with self.assertRaises(NotImplementedError):
+            APoTQuantizer.dequantize(self)
+
+    def test_q_apot_alpha(self):
+        with self.assertRaises(NotImplementedError):
+            APoTQuantizer.q_apot_alpha(self)
+
+if __name__ == '__main__':
+    unittest.main()

torch/ao/quantization/experimental/APoT_tensor.py

@@ -1,14 +1,12 @@
 import torch
-from torch import Tensor
+from torch.ao.quantization.experimental.quantizer import APoTQuantizer
 
 # class to store APoT quantized tensor
 class TensorAPoT(torch.Tensor):
-    @staticmethod
+    quantizer: APoTQuantizer
-    def quantize_APoT(tensor2quantize: Tensor) -> Tensor:
+
+    def __init__(self, quantizer):
         raise NotImplementedError
 
-    def dequantize(self) -> Tensor:
+    def int_repr(self):
-        raise NotImplementedError
-
-    def q_apot_alpha(self) -> float:
         raise NotImplementedError

torch/ao/quantization/experimental/quantizer.py

@@ -0,0 +1,15 @@
+from torch import Tensor
+
+# class to store APoT quantizer
+# implements quantize and dequantize
+# and stores all quantization parameters
+class APoTQuantizer():
+    @staticmethod
+    def quantize_APoT(tensor2quantize: Tensor) -> Tensor:
+        raise NotImplementedError
+
+    def dequantize(self) -> Tensor:
+        raise NotImplementedError
+
+    def q_apot_alpha(self) -> float:
+        raise NotImplementedError