Adding multiweight support to Quantized MobileNetV2 and MobileNetV3

torchvision/prototype/models/__init__.py

@@ -4,8 +4,7 @@
 from .googlenet import *
 from .inception import *
 from .mnasnet import *
-from .mobilenetv2 import *
-from .mobilenetv3 import *
+from .mobilenet import *
 from .regnet import *
 from .resnet import *
 from .shufflenetv2 import *

torchvision/prototype/models/mobilenet.py

@@ -0,0 +1,6 @@
+from .mobilenetv2 import *  # noqa: F401, F403
+from .mobilenetv3 import *  # noqa: F401, F403
+from .mobilenetv2 import __all__ as mv2_all
+from .mobilenetv3 import __all__ as mv3_all
+
+__all__ = mv2_all + mv3_all

torchvision/prototype/models/quantization/__init__.py

@@ -1,4 +1,5 @@
 from .googlenet import *
 from .inception import *
+from .mobilenet import *
 from .resnet import *
 from .shufflenetv2 import *

torchvision/prototype/models/quantization/mobilenet.py

@@ -0,0 +1,6 @@
+from .mobilenetv2 import *  # noqa: F401, F403
+from .mobilenetv3 import *  # noqa: F401, F403
+from .mobilenetv2 import __all__ as mv2_all
+from .mobilenetv3 import __all__ as mv3_all
+
+__all__ = mv2_all + mv3_all

torchvision/prototype/models/quantization/mobilenetv2.py

@@ -0,0 +1,80 @@
+import warnings
+from functools import partial
+from typing import Any, Optional, Union
+
+from torchvision.transforms.functional import InterpolationMode
+
+from ....models.quantization.mobilenetv2 import (
+    QuantizableInvertedResidual,
+    QuantizableMobileNetV2,
+    _replace_relu,
+    quantize_model,
+)
+from ...transforms.presets import ImageNetEval
+from .._api import Weights, WeightEntry
+from .._meta import _IMAGENET_CATEGORIES
+from ..mobilenetv2 import MobileNetV2Weights
+
+
+__all__ = [
+    "QuantizableMobileNetV2",
+    "QuantizedMobileNetV2Weights",
+    "mobilenet_v2",
+]
+
+
+class QuantizedMobileNetV2Weights(Weights):
+    ImageNet1K_QNNPACK_RefV1 = WeightEntry(
+        url="https://download.pytorch.org/models/quantized/mobilenet_v2_qnnpack_37f702c5.pth",
+        transforms=partial(ImageNetEval, crop_size=224),
+        meta={
+            "size": (224, 224),
+            "categories": _IMAGENET_CATEGORIES,
+            "interpolation": InterpolationMode.BILINEAR,
+            "backend": "qnnpack",
+            "quantization": "qat",
+            "recipe": "https://github.com/pytorch/vision/tree/main/references/classification#qat-mobilenetv2",
+            "unquantized": MobileNetV2Weights.ImageNet1K_RefV1,
+            "acc@1": 71.658,
+            "acc@5": 90.150,
+        },
+    )
+
+
+def mobilenet_v2(
+    weights: Optional[Union[QuantizedMobileNetV2Weights, MobileNetV2Weights]] = None,
+    progress: bool = True,
+    quantize: bool = False,
+    **kwargs: Any,
+) -> QuantizableMobileNetV2:
+    if "pretrained" in kwargs:
+        warnings.warn("The argument pretrained is deprecated, please use weights instead.")
+        if kwargs.pop("pretrained"):
+            weights = (
+                QuantizedMobileNetV2Weights.ImageNet1K_QNNPACK_RefV1
+                if quantize
+                else MobileNetV2Weights.ImageNet1K_RefV1
+            )
+        else:
+            weights = None
+
+    if quantize:
+        weights = QuantizedMobileNetV2Weights.verify(weights)
+    else:
+        weights = MobileNetV2Weights.verify(weights)
+
+    if weights is not None:
+        kwargs["num_classes"] = len(weights.meta["categories"])
+        if "backend" in weights.meta:
+            kwargs["backend"] = weights.meta["backend"]
+    backend = kwargs.pop("backend", "qnnpack")
+
+    model = QuantizableMobileNetV2(block=QuantizableInvertedResidual, **kwargs)
+    _replace_relu(model)
+    if quantize:
+        quantize_model(model, backend)
+
+    if weights is not None:
+        model.load_state_dict(weights.state_dict(progress=progress))
+
+    return model

torchvision/prototype/models/quantization/mobilenetv3.py

@@ -0,0 +1,100 @@
+import warnings
+from functools import partial
+from typing import Any, List, Optional, Union
+
+import torch
+from torchvision.transforms.functional import InterpolationMode
+
+from ....models.quantization.mobilenetv3 import (
+    InvertedResidualConfig,
+    QuantizableInvertedResidual,
+    QuantizableMobileNetV3,
+    _replace_relu,
+)
+from ...transforms.presets import ImageNetEval
+from .._api import Weights, WeightEntry
+from .._meta import _IMAGENET_CATEGORIES
+from ..mobilenetv3 import MobileNetV3LargeWeights, _mobilenet_v3_conf
+
+
+__all__ = [
+    "QuantizableMobileNetV3",
+    "QuantizedMobileNetV3LargeWeights",
+    "mobilenet_v3_large",
+]
+
+
+def _mobilenet_v3_model(
+    inverted_residual_setting: List[InvertedResidualConfig],
+    last_channel: int,
+    weights: Optional[Weights],
+    progress: bool,
+    quantize: bool,
+    **kwargs: Any,
+) -> QuantizableMobileNetV3:
+    if weights is not None:
+        kwargs["num_classes"] = len(weights.meta["categories"])
+        if "backend" in weights.meta:
+            kwargs["backend"] = weights.meta["backend"]
+    backend = kwargs.pop("backend", "qnnpack")
+
+    model = QuantizableMobileNetV3(inverted_residual_setting, last_channel, block=QuantizableInvertedResidual, **kwargs)
+    _replace_relu(model)
+
+    if quantize:
+        model.fuse_model()
+        model.qconfig = torch.quantization.get_default_qat_qconfig(backend)
+        torch.quantization.prepare_qat(model, inplace=True)
+
+    if weights is not None:
+        model.load_state_dict(weights.state_dict(progress=progress))
+
+    if quantize:
+        torch.quantization.convert(model, inplace=True)
+        model.eval()
+
+    return model
+
+
+class QuantizedMobileNetV3LargeWeights(Weights):
+    ImageNet1K_QNNPACK_RefV1 = WeightEntry(
+        url="https://download.pytorch.org/models/quantized/mobilenet_v3_large_qnnpack-5bcacf28.pth",
+        transforms=partial(ImageNetEval, crop_size=224),
+        meta={
+            "size": (224, 224),
+            "categories": _IMAGENET_CATEGORIES,
+            "interpolation": InterpolationMode.BILINEAR,
+            "backend": "qnnpack",
+            "quantization": "qat",
+            "recipe": "https://github.com/pytorch/vision/tree/main/references/classification#qat-mobilenetv3",
+            "unquantized": MobileNetV3LargeWeights.ImageNet1K_RefV1,
+            "acc@1": 73.004,
+            "acc@5": 90.858,
+        },
+    )
+
+
+def mobilenet_v3_large(
+    weights: Optional[Union[QuantizedMobileNetV3LargeWeights, MobileNetV3LargeWeights]] = None,
+    progress: bool = True,
+    quantize: bool = False,
+    **kwargs: Any,
+) -> QuantizableMobileNetV3:
+    if "pretrained" in kwargs:
+        warnings.warn("The argument pretrained is deprecated, please use weights instead.")
+        if kwargs.pop("pretrained"):
+            weights = (
+                QuantizedMobileNetV3LargeWeights.ImageNet1K_QNNPACK_RefV1
+                if quantize
+                else MobileNetV3LargeWeights.ImageNet1K_RefV1
+            )
+        else:
+            weights = None
+
+    if quantize:
+        weights = QuantizedMobileNetV3LargeWeights.verify(weights)
+    else:
+        weights = MobileNetV3LargeWeights.verify(weights)
+
+    inverted_residual_setting, last_channel = _mobilenet_v3_conf("mobilenet_v3_large", **kwargs)
+    return _mobilenet_v3_model(inverted_residual_setting, last_channel, weights, progress, quantize, **kwargs)