Adding multiweight support to Quantized InceptionV3 (#4850)

datumbox · web-flow · commit 1cbd9cde35eb · 2021-11-03T18:19:28.000Z
* Moving builder to the bottom to use proper typing.

* Renaming weights.

* Adding quantizated inception builder.

* Correct meta info.

* Fix linter.

* Removing init_weights to avoid exposing it on the class.
diff --git a/torchvision/models/quantization/inception.py b/torchvision/models/quantization/inception.py
@@ -24,72 +24,6 @@
 }
 
 
-def inception_v3(
-    pretrained: bool = False,
-    progress: bool = True,
-    quantize: bool = False,
-    **kwargs: Any,
-) -> "QuantizableInception3":
-
-    r"""Inception v3 model architecture from
-    `"Rethinking the Inception Architecture for Computer Vision" <http://arxiv.org/abs/1512.00567>`_.
-
-    .. note::
-        **Important**: In contrast to the other models the inception_v3 expects tensors with a size of
-        N x 3 x 299 x 299, so ensure your images are sized accordingly.
-
-    Note that quantize = True returns a quantized model with 8 bit
-    weights. Quantized models only support inference and run on CPUs.
-    GPU inference is not yet supported
-
-    Args:
-        pretrained (bool): If True, returns a model pre-trained on ImageNet
-        progress (bool): If True, displays a progress bar of the download to stderr
-        quantize (bool): If True, return a quantized version of the model
-        aux_logits (bool): If True, add an auxiliary branch that can improve training.
-            Default: *True*
-        transform_input (bool): If True, preprocesses the input according to the method with which it
-            was trained on ImageNet. Default: *False*
-    """
-    if pretrained:
-        if "transform_input" not in kwargs:
-            kwargs["transform_input"] = True
-        if "aux_logits" in kwargs:
-            original_aux_logits = kwargs["aux_logits"]
-            kwargs["aux_logits"] = True
-        else:
-            original_aux_logits = False
-
-    model = QuantizableInception3(**kwargs)
-    _replace_relu(model)
-
-    if quantize:
-        # TODO use pretrained as a string to specify the backend
-        backend = "fbgemm"
-        quantize_model(model, backend)
-    else:
-        assert pretrained in [True, False]
-
-    if pretrained:
-        if quantize:
-            if not original_aux_logits:
-                model.aux_logits = False
-                model.AuxLogits = None
-            model_url = quant_model_urls["inception_v3_google_" + backend]
-        else:
-            model_url = inception_module.model_urls["inception_v3_google"]
-
-        state_dict = load_state_dict_from_url(model_url, progress=progress)
-
-        model.load_state_dict(state_dict)
-
-        if not quantize:
-            if not original_aux_logits:
-                model.aux_logits = False
-                model.AuxLogits = None
-    return model
-
-
 class QuantizableBasicConv2d(inception_module.BasicConv2d):
     def __init__(self, *args: Any, **kwargs: Any) -> None:
         super().__init__(*args, **kwargs)
@@ -237,3 +171,68 @@ def fuse_model(self) -> None:
         for m in self.modules():
             if type(m) is QuantizableBasicConv2d:
                 m.fuse_model()
+
+
+def inception_v3(
+    pretrained: bool = False,
+    progress: bool = True,
+    quantize: bool = False,
+    **kwargs: Any,
+) -> QuantizableInception3:
+    r"""Inception v3 model architecture from
+    `"Rethinking the Inception Architecture for Computer Vision" <http://arxiv.org/abs/1512.00567>`_.
+
+    .. note::
+        **Important**: In contrast to the other models the inception_v3 expects tensors with a size of
+        N x 3 x 299 x 299, so ensure your images are sized accordingly.
+
+    Note that quantize = True returns a quantized model with 8 bit
+    weights. Quantized models only support inference and run on CPUs.
+    GPU inference is not yet supported
+
+    Args:
+        pretrained (bool): If True, returns a model pre-trained on ImageNet
+        progress (bool): If True, displays a progress bar of the download to stderr
+        quantize (bool): If True, return a quantized version of the model
+        aux_logits (bool): If True, add an auxiliary branch that can improve training.
+            Default: *True*
+        transform_input (bool): If True, preprocesses the input according to the method with which it
+            was trained on ImageNet. Default: *False*
+    """
+    if pretrained:
+        if "transform_input" not in kwargs:
+            kwargs["transform_input"] = True
+        if "aux_logits" in kwargs:
+            original_aux_logits = kwargs["aux_logits"]
+            kwargs["aux_logits"] = True
+        else:
+            original_aux_logits = False
+
+    model = QuantizableInception3(**kwargs)
+    _replace_relu(model)
+
+    if quantize:
+        # TODO use pretrained as a string to specify the backend
+        backend = "fbgemm"
+        quantize_model(model, backend)
+    else:
+        assert pretrained in [True, False]
+
+    if pretrained:
+        if quantize:
+            if not original_aux_logits:
+                model.aux_logits = False
+                model.AuxLogits = None
+            model_url = quant_model_urls["inception_v3_google_" + backend]
+        else:
+            model_url = inception_module.model_urls["inception_v3_google"]
+
+        state_dict = load_state_dict_from_url(model_url, progress=progress)
+
+        model.load_state_dict(state_dict)
+
+        if not quantize:
+            if not original_aux_logits:
+                model.aux_logits = False
+                model.AuxLogits = None
+    return model
diff --git a/torchvision/prototype/models/inception.py b/torchvision/prototype/models/inception.py
@@ -10,10 +10,10 @@
 from ._meta import _IMAGENET_CATEGORIES
 
 
-__all__ = ["Inception3", "InceptionOutputs", "_InceptionOutputs", "Inception3Weights", "inception_v3"]
+__all__ = ["Inception3", "InceptionOutputs", "_InceptionOutputs", "InceptionV3Weights", "inception_v3"]
 
 
-class Inception3Weights(Weights):
+class InceptionV3Weights(Weights):
     ImageNet1K_TFV1 = WeightEntry(
         url="https://download.pytorch.org/models/inception_v3_google-0cc3c7bd.pth",
         transforms=partial(ImageNetEval, crop_size=299, resize_size=342),
@@ -28,11 +28,11 @@ class Inception3Weights(Weights):
     )
 
 
-def inception_v3(weights: Optional[Inception3Weights] = None, progress: bool = True, **kwargs: Any) -> Inception3:
+def inception_v3(weights: Optional[InceptionV3Weights] = None, progress: bool = True, **kwargs: Any) -> Inception3:
     if "pretrained" in kwargs:
         warnings.warn("The argument pretrained is deprecated, please use weights instead.")
-        weights = Inception3Weights.ImageNet1K_TFV1 if kwargs.pop("pretrained") else None
-    weights = Inception3Weights.verify(weights)
+        weights = InceptionV3Weights.ImageNet1K_TFV1 if kwargs.pop("pretrained") else None
+    weights = InceptionV3Weights.verify(weights)
 
     original_aux_logits = kwargs.get("aux_logits", True)
     if weights is not None:
diff --git a/torchvision/prototype/models/quantization/__init__.py b/torchvision/prototype/models/quantization/__init__.py
@@ -1,2 +1,3 @@
 from .googlenet import *
+from .inception import *
 from .resnet import *
diff --git a/torchvision/prototype/models/quantization/inception.py b/torchvision/prototype/models/quantization/inception.py
@@ -0,0 +1,87 @@
+import warnings
+from functools import partial
+from typing import Any, Optional, Union
+
+from torchvision.transforms.functional import InterpolationMode
+
+from ....models.quantization.inception import (
+    QuantizableInception3,
+    _replace_relu,
+    quantize_model,
+)
+from ...transforms.presets import ImageNetEval
+from .._api import Weights, WeightEntry
+from .._meta import _IMAGENET_CATEGORIES
+from ..inception import InceptionV3Weights
+
+
+__all__ = [
+    "QuantizableInception3",
+    "QuantizedInceptionV3Weights",
+    "inception_v3",
+]
+
+
+class QuantizedInceptionV3Weights(Weights):
+    ImageNet1K_FBGEMM_TFV1 = WeightEntry(
+        url="https://download.pytorch.org/models/quantized/inception_v3_google_fbgemm-71447a44.pth",
+        transforms=partial(ImageNetEval, crop_size=299, resize_size=342),
+        meta={
+            "size": (299, 299),
+            "categories": _IMAGENET_CATEGORIES,
+            "interpolation": InterpolationMode.BILINEAR,
+            "backend": "fbgemm",
+            "quantization": "ptq",
+            "recipe": "https://github.com/pytorch/vision/tree/main/references/classification#post-training-quantized-models",
+            "unquantized": InceptionV3Weights.ImageNet1K_TFV1,
+            "acc@1": 77.176,
+            "acc@5": 93.354,
+        },
+    )
+
+
+def inception_v3(
+    weights: Optional[Union[QuantizedInceptionV3Weights, InceptionV3Weights]] = None,
+    progress: bool = True,
+    quantize: bool = False,
+    **kwargs: Any,
+) -> QuantizableInception3:
+    if "pretrained" in kwargs:
+        warnings.warn("The argument pretrained is deprecated, please use weights instead.")
+        if kwargs.pop("pretrained"):
+            weights = (
+                QuantizedInceptionV3Weights.ImageNet1K_FBGEMM_TFV1 if quantize else InceptionV3Weights.ImageNet1K_TFV1
+            )
+        else:
+            weights = None
+
+    if quantize:
+        weights = QuantizedInceptionV3Weights.verify(weights)
+    else:
+        weights = InceptionV3Weights.verify(weights)
+
+    original_aux_logits = kwargs.get("aux_logits", False)
+    if weights is not None:
+        if "transform_input" not in kwargs:
+            kwargs["transform_input"] = True
+        kwargs["aux_logits"] = True
+        kwargs["num_classes"] = len(weights.meta["categories"])
+        if "backend" in weights.meta:
+            kwargs["backend"] = weights.meta["backend"]
+    backend = kwargs.pop("backend", "fbgemm")
+
+    model = QuantizableInception3(**kwargs)
+    _replace_relu(model)
+    if quantize:
+        quantize_model(model, backend)
+
+    if weights is not None:
+        if quantize and not original_aux_logits:
+            model.aux_logits = False
+            model.AuxLogits = None
+        model.load_state_dict(weights.state_dict(progress=progress))
+        if not quantize and not original_aux_logits:
+            model.aux_logits = False
+            model.AuxLogits = None
+
+    return model

Original file line number	Diff line number	Diff line change
`@@ -1,2 +1,3 @@`
`1`	`1`	`from .googlenet import *`
	`2`	`+from .inception import *`
`2`	`3`	`from .resnet import *`