[fbsync] Adding resnext101 64x4d model (#5935)

datumbox · facebook-github-bot · commit 309483c3de3a · 2022-05-11T01:58:30.000-07:00
Summary:
* Add resnext101_64x4d model definition

* Add test for resnext101_64x4d

* Add resnext101_64x4d weight

* Update checkpoint to use EMA weigth

* Add quantization model signature for resnext101_64x4d

* Fix class name and update accuracy using 1 gpu and batch_size=1

* Apply ufmt

* Update the quantized weight and accuracy that we still keep the training log

* Add quantized expect file

* Update docs and fix acc1

* Add recipe for quantized to PR

* Update models.rst

Reviewed By: YosuaMichael

Differential Revision: D36281598

fbshipit-source-id: 300bd36343b8ad8b185a246b794e078bdf67f5c8
diff --git a/docs/source/models.rst b/docs/source/models.rst
@@ -61,6 +61,8 @@ You can construct a model with random weights by calling its constructor:
     mobilenet_v3_large = models.mobilenet_v3_large()
     mobilenet_v3_small = models.mobilenet_v3_small()
     resnext50_32x4d = models.resnext50_32x4d()
+    resnext101_32x8d = models.resnext101_32x8d()
+    resnext101_64x4d = models.resnext101_64x4d()
     wide_resnet50_2 = models.wide_resnet50_2()
     mnasnet = models.mnasnet1_0()
     efficientnet_b0 = models.efficientnet_b0()
@@ -185,6 +187,7 @@ MobileNet V3 Large                74.042          91.340
 MobileNet V3 Small                67.668          87.402
 ResNeXt-50-32x4d                  77.618          93.698
 ResNeXt-101-32x8d                 79.312          94.526
+ResNeXt-101-64x4d                 83.246          96.454
 Wide ResNet-50-2                  78.468          94.086
 Wide ResNet-101-2                 78.848          94.284
 MNASNet 1.0                       73.456          91.510
@@ -366,6 +369,7 @@ ResNext
 
     resnext50_32x4d
     resnext101_32x8d
+    resnext101_64x4d
 
 Wide ResNet
 -----------
@@ -481,8 +485,11 @@ a model with random weights by calling its constructor:
     resnet18 = models.quantization.resnet18()
     resnet50 = models.quantization.resnet50()
     resnext101_32x8d = models.quantization.resnext101_32x8d()
+    resnext101_64x4d = models.quantization.resnext101_64x4d()
     shufflenet_v2_x0_5 = models.quantization.shufflenet_v2_x0_5()
     shufflenet_v2_x1_0 = models.quantization.shufflenet_v2_x1_0()
+    shufflenet_v2_x1_5 = models.quantization.shufflenet_v2_x1_5()
+    shufflenet_v2_x2_0 = models.quantization.shufflenet_v2_x2_0()
 
 Obtaining a pre-trained quantized model can be done with a few lines of code:
 
@@ -508,6 +515,7 @@ ShuffleNet V2 x2.0                75.354         92.488
 ResNet 18                         69.494         88.882
 ResNet 50                         75.920         92.814
 ResNext 101 32x8d                 78.986         94.480
+ResNext 101 64x4d                 82.898         96.326
 Inception V3                      77.176         93.354
 GoogleNet                         69.826         89.404
 ================================  =============  =============
diff --git a/test/expect/ModelTester.test_resnext101_64x4d_expect.pkl b/test/expect/ModelTester.test_resnext101_64x4d_expect.pkl
diff --git a/test/expect/ModelTester.test_resnext101_64x4d_quantized_expect.pkl b/test/expect/ModelTester.test_resnext101_64x4d_quantized_expect.pkl
diff --git a/test/test_models.py b/test/test_models.py
@@ -315,6 +315,7 @@ def _check_input_backprop(model, inputs):
     "convnext_base",
     "convnext_large",
     "resnext101_32x8d",
+    "resnext101_64x4d",
     "wide_resnet101_2",
     "efficientnet_b6",
     "efficientnet_b7",
diff --git a/torchvision/models/quantization/resnet.py b/torchvision/models/quantization/resnet.py
@@ -11,6 +11,7 @@
     ResNet18_Weights,
     ResNet50_Weights,
     ResNeXt101_32X8D_Weights,
+    ResNeXt101_64X4D_Weights,
 )
 
 from ...transforms._presets import ImageClassification
@@ -25,9 +26,11 @@
     "ResNet18_QuantizedWeights",
     "ResNet50_QuantizedWeights",
     "ResNeXt101_32X8D_QuantizedWeights",
+    "ResNeXt101_64X4D_QuantizedWeights",
     "resnet18",
     "resnet50",
     "resnext101_32x8d",
+    "resnext101_64x4d",
 ]
 
 
@@ -231,6 +234,24 @@ class ResNeXt101_32X8D_QuantizedWeights(WeightsEnum):
     DEFAULT = IMAGENET1K_FBGEMM_V2
 
 
+class ResNeXt101_64X4D_QuantizedWeights(WeightsEnum):
+    IMAGENET1K_FBGEMM_V1 = Weights(
+        url="https://download.pytorch.org/models/quantized/resnext101_64x4d_fbgemm-605a1cb3.pth",
+        transforms=partial(ImageClassification, crop_size=224, resize_size=232),
+        meta={
+            **_COMMON_META,
+            "num_params": 83455272,
+            "recipe": "https://github.com/pytorch/vision/pull/5935",
+            "unquantized": ResNeXt101_64X4D_Weights.IMAGENET1K_V1,
+            "metrics": {
+                "acc@1": 82.898,
+                "acc@5": 96.326,
+            },
+        },
+    )
+    DEFAULT = IMAGENET1K_FBGEMM_V1
+
+
 @handle_legacy_interface(
     weights=(
         "pretrained",
@@ -318,3 +339,26 @@ def resnext101_32x8d(
     _ovewrite_named_param(kwargs, "groups", 32)
     _ovewrite_named_param(kwargs, "width_per_group", 8)
     return _resnet(QuantizableBottleneck, [3, 4, 23, 3], weights, progress, quantize, **kwargs)
+
+
+def resnext101_64x4d(
+    *,
+    weights: Optional[Union[ResNeXt101_64X4D_QuantizedWeights, ResNeXt101_64X4D_Weights]] = None,
+    progress: bool = True,
+    quantize: bool = False,
+    **kwargs: Any,
+) -> QuantizableResNet:
+    r"""ResNeXt-101 64x4d model from
+    `"Aggregated Residual Transformation for Deep Neural Networks" <https://arxiv.org/pdf/1611.05431.pdf>`_
+
+    Args:
+        weights (ResNeXt101_64X4D_QuantizedWeights or ResNeXt101_64X4D_Weights, optional): The pretrained
+            weights for the model
+        progress (bool): If True, displays a progress bar of the download to stderr
+        quantize (bool): If True, return a quantized version of the model
+    """
+    weights = (ResNeXt101_64X4D_QuantizedWeights if quantize else ResNeXt101_64X4D_Weights).verify(weights)
+
+    _ovewrite_named_param(kwargs, "groups", 64)
+    _ovewrite_named_param(kwargs, "width_per_group", 4)
+    return _resnet(QuantizableBottleneck, [3, 4, 23, 3], weights, progress, quantize, **kwargs)
diff --git a/torchvision/models/resnet.py b/torchvision/models/resnet.py
@@ -21,6 +21,7 @@
     "ResNet152_Weights",
     "ResNeXt50_32X4D_Weights",
     "ResNeXt101_32X8D_Weights",
+    "ResNeXt101_64X4D_Weights",
     "Wide_ResNet50_2_Weights",
     "Wide_ResNet101_2_Weights",
     "resnet18",
@@ -30,6 +31,7 @@
     "resnet152",
     "resnext50_32x4d",
     "resnext101_32x8d",
+    "resnext101_64x4d",
     "wide_resnet50_2",
     "wide_resnet101_2",
 ]
@@ -491,6 +493,24 @@ class ResNeXt101_32X8D_Weights(WeightsEnum):
     DEFAULT = IMAGENET1K_V2
 
 
+class ResNeXt101_64X4D_Weights(WeightsEnum):
+    IMAGENET1K_V1 = Weights(
+        url="https://download.pytorch.org/models/resnext101_64x4d-173b62eb.pth",
+        transforms=partial(ImageClassification, crop_size=224, resize_size=232),
+        meta={
+            **_COMMON_META,
+            "num_params": 83455272,
+            "recipe": "https://github.com/pytorch/vision/pull/5935",
+            "metrics": {
+                # Mock
+                "acc@1": 83.246,
+                "acc@5": 96.454,
+            },
+        },
+    )
+    DEFAULT = IMAGENET1K_V1
+
+
 class Wide_ResNet50_2_Weights(WeightsEnum):
     IMAGENET1K_V1 = Weights(
         url="https://download.pytorch.org/models/wide_resnet50_2-95faca4d.pth",
@@ -734,6 +754,34 @@ def resnext101_32x8d(
     return _resnet(Bottleneck, [3, 4, 23, 3], weights, progress, **kwargs)
 
 
+def resnext101_64x4d(
+    *, weights: Optional[ResNeXt101_64X4D_Weights] = None, progress: bool = True, **kwargs: Any
+) -> ResNet:
+    """ResNeXt-101 64x4d model from
+    `Aggregated Residual Transformation for Deep Neural Networks <https://arxiv.org/abs/1611.05431>`_.
+
+    Args:
+        weights (:class:`~torchvision.models.ResNeXt101_64X4D_Weights`, optional): The
+            pretrained weights to use. See
+            :class:`~torchvision.models.ResNeXt101_64X4D_Weights` below for
+            more details, and possible values. By default, no pre-trained
+            weights are used.
+        progress (bool, optional): If True, displays a progress bar of the
+            download to stderr. Default is True.
+        **kwargs: parameters passed to the ``torchvision.models.resnet.ResNet``
+            base class. Please refer to the `source code
+            <https://github.com/pytorch/vision/blob/main/torchvision/models/resnet.py>`_
+            for more details about this class.
+    .. autoclass:: torchvision.models.ResNeXt101_64X4D_Weights
+        :members:
+    """
+    weights = ResNeXt101_64X4D_Weights.verify(weights)
+
+    _ovewrite_named_param(kwargs, "groups", 64)
+    _ovewrite_named_param(kwargs, "width_per_group", 4)
+    return _resnet(Bottleneck, [3, 4, 23, 3], weights, progress, **kwargs)
+
+
 @handle_legacy_interface(weights=("pretrained", Wide_ResNet50_2_Weights.IMAGENET1K_V1))
 def wide_resnet50_2(
     *, weights: Optional[Wide_ResNet50_2_Weights] = None, progress: bool = True, **kwargs: Any
