pytorch · datumbox · Apr 21, 2022 · Apr 21, 2022 · Apr 21, 2022 · Apr 21, 2022
diff --git a/test/test_extended_models.py b/test/test_extended_models.py
@@ -79,12 +79,13 @@ def test_naming_conventions(model_fn):
 )
 @run_if_test_with_extended
 def test_schema_meta_validation(model_fn):
-    classification_fields = ["size", "categories", "acc@1", "acc@5", "min_size"]
+    # TODO: add list of permitted fields
+    classification_fields = ["categories", "acc@1", "acc@5"]
     defaults = {
-        "all": ["task", "architecture", "recipe", "num_params"],
+        "all": ["recipe", "num_params", "min_size"],
         "models": classification_fields,
         "detection": ["categories", "map"],
-        "quantization": classification_fields + ["backend", "quantization", "unquantized"],
+        "quantization": classification_fields + ["backend", "unquantized"],
         "segmentation": ["categories", "mIoU", "acc"],
         "video": classification_fields,
         "optical_flow": [],

diff --git a/torchvision/models/alexnet.py b/torchvision/models/alexnet.py
@@ -57,10 +57,7 @@ class AlexNet_Weights(WeightsEnum):
         url="https://download.pytorch.org/models/alexnet-owt-7be5be79.pth",
         transforms=partial(ImageClassification, crop_size=224),
         meta={
-            "task": "image_classification",
-            "architecture": "AlexNet",
             "num_params": 61100840,
-            "size": (224, 224),
             "min_size": (63, 63),
             "categories": _IMAGENET_CATEGORIES,
             "recipe": "https://github.com/pytorch/vision/tree/main/references/classification#alexnet-and-vgg",

diff --git a/torchvision/models/convnext.py b/torchvision/models/convnext.py
@@ -204,9 +204,6 @@ def _convnext(
 
 
 _COMMON_META = {
-    "task": "image_classification",
-    "architecture": "ConvNeXt",
-    "size": (224, 224),
     "min_size": (32, 32),
     "categories": _IMAGENET_CATEGORIES,
     "recipe": "https://github.com/pytorch/vision/tree/main/references/classification#convnext",

diff --git a/torchvision/models/densenet.py b/torchvision/models/densenet.py
@@ -266,9 +266,6 @@ def _densenet(
 
 
 _COMMON_META = {
-    "task": "image_classification",
-    "architecture": "DenseNet",
-    "size": (224, 224),
     "min_size": (29, 29),
     "categories": _IMAGENET_CATEGORIES,
     "recipe": "https://github.com/pytorch/vision/pull/116",

diff --git a/torchvision/models/detection/faster_rcnn.py b/torchvision/models/detection/faster_rcnn.py
@@ -370,9 +370,8 @@ def forward(self, x):
 
 
 _COMMON_META = {
-    "task": "image_object_detection",
-    "architecture": "FasterRCNN",
     "categories": _COCO_CATEGORIES,
+    "min_size": (1, 1),
 }
 
 

diff --git a/torchvision/models/detection/fcos.py b/torchvision/models/detection/fcos.py
@@ -651,10 +651,9 @@ class FCOS_ResNet50_FPN_Weights(WeightsEnum):
         url="https://download.pytorch.org/models/fcos_resnet50_fpn_coco-99b0c9b7.pth",
         transforms=ObjectDetection,
         meta={
-            "task": "image_object_detection",
-            "architecture": "FCOS",
             "num_params": 32269600,
             "categories": _COCO_CATEGORIES,
+            "min_size": (1, 1),
             "recipe": "https://github.com/pytorch/vision/tree/main/references/detection#fcos-resnet-50-fpn",
             "map": 39.2,
         },

diff --git a/torchvision/models/detection/keypoint_rcnn.py b/torchvision/models/detection/keypoint_rcnn.py
@@ -308,10 +308,9 @@ def forward(self, x):
 
 
 _COMMON_META = {
-    "task": "image_object_detection",
-    "architecture": "KeypointRCNN",
     "categories": _COCO_PERSON_CATEGORIES,
     "keypoint_names": _COCO_PERSON_KEYPOINT_NAMES,
+    "min_size": (1, 1),
 }
 
 

diff --git a/torchvision/models/detection/mask_rcnn.py b/torchvision/models/detection/mask_rcnn.py
@@ -351,9 +351,8 @@ def __init__(self, in_channels, dim_reduced, num_classes):
 
 
 _COMMON_META = {
-    "task": "image_object_detection",
-    "architecture": "MaskRCNN",
     "categories": _COCO_CATEGORIES,
+    "min_size": (1, 1),
 }
 
 

diff --git a/torchvision/models/detection/retinanet.py b/torchvision/models/detection/retinanet.py
@@ -674,9 +674,8 @@ def forward(self, images, targets=None):
 
 
 _COMMON_META = {
-    "task": "image_object_detection",
-    "architecture": "RetinaNet",
     "categories": _COCO_CATEGORIES,
+    "min_size": (1, 1),
 }
 
 

diff --git a/torchvision/models/detection/ssd.py b/torchvision/models/detection/ssd.py
@@ -30,11 +30,9 @@ class SSD300_VGG16_Weights(WeightsEnum):
         url="https://download.pytorch.org/models/ssd300_vgg16_coco-b556d3b4.pth",
         transforms=ObjectDetection,
         meta={
-            "task": "image_object_detection",
-            "architecture": "SSD",
             "num_params": 35641826,
-            "size": (300, 300),
             "categories": _COCO_CATEGORIES,
+            "min_size": (1, 1),
             "recipe": "https://github.com/pytorch/vision/tree/main/references/detection#ssd300-vgg16",
             "map": 25.1,
         },

diff --git a/torchvision/models/detection/ssdlite.py b/torchvision/models/detection/ssdlite.py
@@ -189,11 +189,9 @@ class SSDLite320_MobileNet_V3_Large_Weights(WeightsEnum):
         url="https://download.pytorch.org/models/ssdlite320_mobilenet_v3_large_coco-a79551df.pth",
         transforms=ObjectDetection,
         meta={
-            "task": "image_object_detection",
-            "architecture": "SSDLite",
             "num_params": 3440060,
-            "size": (320, 320),
             "categories": _COCO_CATEGORIES,
+            "min_size": (1, 1),
             "recipe": "https://github.com/pytorch/vision/tree/main/references/detection#ssdlite320-mobilenetv3-large",
             "map": 21.3,
         },

diff --git a/torchvision/models/efficientnet.py b/torchvision/models/efficientnet.py
@@ -3,7 +3,7 @@
 import warnings
 from dataclasses import dataclass
 from functools import partial
-from typing import Any, Callable, Optional, List, Sequence, Tuple, Union
+from typing import Any, Callable, Dict, Optional, List, Sequence, Tuple, Union
 
 import torch
 from torch import nn, Tensor
@@ -429,23 +429,20 @@ def _efficientnet_conf(
     return inverted_residual_setting, last_channel
 
 
-_COMMON_META = {
-    "task": "image_classification",
+_COMMON_META: Dict[str, Any] = {
     "categories": _IMAGENET_CATEGORIES,
     "recipe": "https://github.com/pytorch/vision/tree/main/references/classification#efficientnet",
 }
 
 
 _COMMON_META_V1 = {
     **_COMMON_META,
-    "architecture": "EfficientNet",
     "min_size": (1, 1),
 }
 
 
 _COMMON_META_V2 = {
     **_COMMON_META,
-    "architecture": "EfficientNetV2",
     "min_size": (33, 33),
 }
 
@@ -459,7 +456,6 @@ class EfficientNet_B0_Weights(WeightsEnum):
         meta={
             **_COMMON_META_V1,
             "num_params": 5288548,
-            "size": (224, 224),
             "acc@1": 77.692,
             "acc@5": 93.532,
         },
@@ -476,7 +472,6 @@ class EfficientNet_B1_Weights(WeightsEnum):
         meta={
             **_COMMON_META_V1,
             "num_params": 7794184,
-            "size": (240, 240),
             "acc@1": 78.642,
             "acc@5": 94.186,
         },
@@ -490,7 +485,6 @@ class EfficientNet_B1_Weights(WeightsEnum):
             **_COMMON_META_V1,
             "num_params": 7794184,
             "recipe": "https://github.com/pytorch/vision/issues/3995#new-recipe-with-lr-wd-crop-tuning",
-            "size": (240, 240),
             "acc@1": 79.838,
             "acc@5": 94.934,
         },
@@ -507,7 +501,6 @@ class EfficientNet_B2_Weights(WeightsEnum):
         meta={
             **_COMMON_META_V1,
             "num_params": 9109994,
-            "size": (288, 288),
             "acc@1": 80.608,
             "acc@5": 95.310,
         },
@@ -524,7 +517,6 @@ class EfficientNet_B3_Weights(WeightsEnum):
         meta={
             **_COMMON_META_V1,
             "num_params": 12233232,
-            "size": (300, 300),
             "acc@1": 82.008,
             "acc@5": 96.054,
         },
@@ -541,7 +533,6 @@ class EfficientNet_B4_Weights(WeightsEnum):
         meta={
             **_COMMON_META_V1,
             "num_params": 19341616,
-            "size": (380, 380),
             "acc@1": 83.384,
             "acc@5": 96.594,
         },
@@ -558,7 +549,6 @@ class EfficientNet_B5_Weights(WeightsEnum):
         meta={
             **_COMMON_META_V1,
             "num_params": 30389784,
-            "size": (456, 456),
             "acc@1": 83.444,
             "acc@5": 96.628,
         },
@@ -575,7 +565,6 @@ class EfficientNet_B6_Weights(WeightsEnum):
         meta={
             **_COMMON_META_V1,
             "num_params": 43040704,
-            "size": (528, 528),
             "acc@1": 84.008,
             "acc@5": 96.916,
         },
@@ -592,7 +581,6 @@ class EfficientNet_B7_Weights(WeightsEnum):
         meta={
             **_COMMON_META_V1,
             "num_params": 66347960,
-            "size": (600, 600),
             "acc@1": 84.122,
             "acc@5": 96.908,
         },
@@ -612,7 +600,6 @@ class EfficientNet_V2_S_Weights(WeightsEnum):
         meta={
             **_COMMON_META_V2,
             "num_params": 21458488,
-            "size": (384, 384),
             "acc@1": 84.228,
             "acc@5": 96.878,
         },
@@ -632,7 +619,6 @@ class EfficientNet_V2_M_Weights(WeightsEnum):
         meta={
             **_COMMON_META_V2,
             "num_params": 54139356,
-            "size": (480, 480),
             "acc@1": 85.112,
             "acc@5": 97.156,
         },
@@ -654,7 +640,6 @@ class EfficientNet_V2_L_Weights(WeightsEnum):
         meta={
             **_COMMON_META_V2,
             "num_params": 118515272,
-            "size": (480, 480),
             "acc@1": 85.808,
             "acc@5": 97.788,
         },

diff --git a/torchvision/models/googlenet.py b/torchvision/models/googlenet.py
@@ -280,10 +280,7 @@ class GoogLeNet_Weights(WeightsEnum):
         url="https://download.pytorch.org/models/googlenet-1378be20.pth",
         transforms=partial(ImageClassification, crop_size=224),
         meta={
-            "task": "image_classification",
-            "architecture": "GoogLeNet",
             "num_params": 6624904,
-            "size": (224, 224),
             "min_size": (15, 15),
             "categories": _IMAGENET_CATEGORIES,
             "recipe": "https://github.com/pytorch/vision/tree/main/references/classification#googlenet",

diff --git a/torchvision/models/inception.py b/torchvision/models/inception.py
@@ -412,10 +412,7 @@ class Inception_V3_Weights(WeightsEnum):
         url="https://download.pytorch.org/models/inception_v3_google-0cc3c7bd.pth",
         transforms=partial(ImageClassification, crop_size=299, resize_size=342),
         meta={
-            "task": "image_classification",
-            "architecture": "InceptionV3",
             "num_params": 27161264,
-            "size": (299, 299),
             "min_size": (75, 75),
             "categories": _IMAGENET_CATEGORIES,
             "recipe": "https://github.com/pytorch/vision/tree/main/references/classification#inception-v3",

diff --git a/torchvision/models/mnasnet.py b/torchvision/models/mnasnet.py
@@ -212,9 +212,6 @@ def _load_from_state_dict(
 
 
 _COMMON_META = {
-    "task": "image_classification",
-    "architecture": "MNASNet",
-    "size": (224, 224),
     "min_size": (1, 1),
     "categories": _IMAGENET_CATEGORIES,
     "recipe": "https://github.com/1e100/mnasnet_trainer",

diff --git a/torchvision/models/mobilenetv2.py b/torchvision/models/mobilenetv2.py
@@ -195,10 +195,7 @@ def forward(self, x: Tensor) -> Tensor:
 
 
 _COMMON_META = {
-    "task": "image_classification",
-    "architecture": "MobileNetV2",
     "num_params": 3504872,
-    "size": (224, 224),
     "min_size": (1, 1),
     "categories": _IMAGENET_CATEGORIES,
 }

diff --git a/torchvision/models/mobilenetv3.py b/torchvision/models/mobilenetv3.py
@@ -304,9 +304,6 @@ def _mobilenet_v3(
 
 
 _COMMON_META = {
-    "task": "image_classification",
-    "architecture": "MobileNetV3",
-    "size": (224, 224),
     "min_size": (1, 1),
     "categories": _IMAGENET_CATEGORIES,
 }

diff --git a/torchvision/models/optical_flow/raft.py b/torchvision/models/optical_flow/raft.py
@@ -512,8 +512,7 @@ def forward(self, image1, image2, num_flow_updates: int = 12):
 
 
 _COMMON_META = {
-    "task": "optical_flow",
-    "architecture": "RAFT",
+    "min_size": (128, 128),
 }
 
 

diff --git a/torchvision/models/quantization/googlenet.py b/torchvision/models/quantization/googlenet.py
@@ -111,14 +111,10 @@ class GoogLeNet_QuantizedWeights(WeightsEnum):
         url="https://download.pytorch.org/models/quantized/googlenet_fbgemm-c00238cf.pth",
         transforms=partial(ImageClassification, crop_size=224),
         meta={
-            "task": "image_classification",
-            "architecture": "GoogLeNet",
             "num_params": 6624904,
-            "size": (224, 224),
             "min_size": (15, 15),
             "categories": _IMAGENET_CATEGORIES,
             "backend": "fbgemm",
-            "quantization": "Post Training Quantization",
             "recipe": "https://github.com/pytorch/vision/tree/main/references/classification#post-training-quantized-models",
             "unquantized": GoogLeNet_Weights.IMAGENET1K_V1,
             "acc@1": 69.826,

diff --git a/torchvision/models/quantization/inception.py b/torchvision/models/quantization/inception.py
@@ -177,14 +177,10 @@ class Inception_V3_QuantizedWeights(WeightsEnum):
         url="https://download.pytorch.org/models/quantized/inception_v3_google_fbgemm-71447a44.pth",
         transforms=partial(ImageClassification, crop_size=299, resize_size=342),
         meta={
-            "task": "image_classification",
-            "architecture": "InceptionV3",
             "num_params": 27161264,
-            "size": (299, 299),
             "min_size": (75, 75),
             "categories": _IMAGENET_CATEGORIES,
             "backend": "fbgemm",
-            "quantization": "Post Training Quantization",
             "recipe": "https://github.com/pytorch/vision/tree/main/references/classification#post-training-quantized-models",
             "unquantized": Inception_V3_Weights.IMAGENET1K_V1,
             "acc@1": 77.176,

diff --git a/torchvision/models/quantization/mobilenetv2.py b/torchvision/models/quantization/mobilenetv2.py
@@ -69,14 +69,10 @@ class MobileNet_V2_QuantizedWeights(WeightsEnum):
         url="https://download.pytorch.org/models/quantized/mobilenet_v2_qnnpack_37f702c5.pth",
         transforms=partial(ImageClassification, crop_size=224),
         meta={
-            "task": "image_classification",
-            "architecture": "MobileNetV2",
             "num_params": 3504872,
-            "size": (224, 224),
             "min_size": (1, 1),
             "categories": _IMAGENET_CATEGORIES,
             "backend": "qnnpack",
-            "quantization": "Quantization Aware Training",
             "recipe": "https://github.com/pytorch/vision/tree/main/references/classification#qat-mobilenetv2",
             "unquantized": MobileNet_V2_Weights.IMAGENET1K_V1,
             "acc@1": 71.658,

diff --git a/torchvision/models/quantization/mobilenetv3.py b/torchvision/models/quantization/mobilenetv3.py
@@ -159,14 +159,10 @@ class MobileNet_V3_Large_QuantizedWeights(WeightsEnum):
         url="https://download.pytorch.org/models/quantized/mobilenet_v3_large_qnnpack-5bcacf28.pth",
         transforms=partial(ImageClassification, crop_size=224),
         meta={
-            "task": "image_classification",
-            "architecture": "MobileNetV3",
             "num_params": 5483032,
-            "size": (224, 224),
             "min_size": (1, 1),
             "categories": _IMAGENET_CATEGORIES,
             "backend": "qnnpack",
-            "quantization": "Quantization Aware Training",
             "recipe": "https://github.com/pytorch/vision/tree/main/references/classification#qat-mobilenetv3",
             "unquantized": MobileNet_V3_Large_Weights.IMAGENET1K_V1,
             "acc@1": 73.004,