Add Inception_v3

yk · Jul 11, 2019 · 39985cb · 39985cb
1 parent 25986ef
commit 39985cb
Show file tree

Hide file tree

Showing 17 changed files with 238 additions and 136 deletions.
diff --git a/.ipynb_checkpoints/CIFAR10-checkpoint.ipynb b/.ipynb_checkpoints/CIFAR10-checkpoint.ipynb
@@ -32,7 +32,7 @@
     "import torchvision\n",
     "import torchvision.transforms as transforms\n",
     "\n",
-    "from tqdm import tqdm_notebook as pbar\n",
+    "from tqdm import tqdm as pbar\n",
     "from tensorboardX import SummaryWriter\n",
     "from models import *"
    ]
@@ -204,9 +204,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# Train on cuda if available\n",
-    "device = torch.device('cuda:1' if torch.cuda.is_available() else 'cpu')\n",
-    "print(\"Using\", device)"
+    "model = inception_v3(pretrained=True)"
    ]
   },
   {
@@ -216,6 +214,17 @@
     "## 6. Put everything together"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Train on cuda if available\n",
+    "device = torch.device('cuda:1' if torch.cuda.is_available() else 'cpu')\n",
+    "print(\"Using\", device)"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -251,18 +260,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# test_model(model, train_params)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "# for i, a_model in enumerate(all_models):\n",
-    "#     train_params['description'] = descriptions[i]\n",
-    "#     train_model(a_model, train_params)"
+    "test_model(model, train_params)"
    ]
   },
   {

diff --git a/.ipynb_checkpoints/README-checkpoint.md b/.ipynb_checkpoints/README-checkpoint.md
@@ -3,21 +3,20 @@
 - I changed *number of class, filter size, stride, and padding* in the the original code so that it works with CIFAR-10.
 - I also share the **weights** of these models, so you can just load the weights and use them.
 
-## Statistic of supported models
+## Statistics of supported models
 | No. | Model        | Test Accuracy | Parameters | State Dict Size |
 |-----|--------------|---------------|------------|-----------------|
-| 1   | vgg11_bn     | 92.61%        | 128.813M   | 491 MB          |
-| 2   | vgg13_bn     | 94.27%        | 128.998M   | 492 MB          |
-| 3   | vgg16_bn     | 94.07%        | 134.310M   | 512 MB          |
-| 4   | vgg19_bn     | 94.25%        | 139.622M   | 533 MB          |
-| 5   | resnet18     | 93.48%        | 11.174M    | 43 MB           |
-| 6   | resnet34     | 93.82%        | 21.282M    | 81 MB           |
-| 7   | resnet50     | 94.38%        | 23.521M    | 90 MB           |
-| 8   | mobilenet_v2 | 93.85%        | 2.237M     | 9 MB            |
-| 9   | densenet121  | 94.76%        | 6.956M     | 27 MB           |
-| 10  | googlenet    | 95.08%        | 5.491M     | 21 MB           |
-
-I will add more models...
+| 1   | vgg11_bn     | 92.61%        | 128.813 M  | 491 MB          |
+| 2   | vgg13_bn     | 94.27%        | 128.998 M  | 492 MB          |
+| 3   | vgg16_bn     | 94.07%        | 134.310 M  | 512 MB          |
+| 4   | vgg19_bn     | 94.25%        | 139.622 M  | 533 MB          |
+| 5   | resnet18     | 93.48%        | 11.174 M   | 43 MB           |
+| 6   | resnet34     | 93.82%        | 21.282 M   | 81 MB           |
+| 7   | resnet50     | 94.38%        | 23.521 M   | 90 MB           |
+| 8   | mobilenet_v2 | 93.85%        | 2.237 M    | 9 MB            |
+| 9   | densenet121  | 94.76%        | 6.956 M    | 27 MB           |
+| 10  | googlenet    | 95.08%        | 5.491 M    | 21 MB           |
+| 11  | inception_v3 | 95.41%        | 21.64 M    | 83 MB           |
 
 ## How To Use
 
@@ -40,7 +39,6 @@ my_model = vgg11_bn(pretrained=True)
 transforms.Normalize([0.4914, 0.4822, 0.4465], [0.2023, 0.1994, 0.2010])]
 ```
 
-
 ## Training Hyper-paramters
 - Batch size: 256
 - Number of epochs: 600

diff --git a/CIFAR10.ipynb b/CIFAR10.ipynb
@@ -32,7 +32,7 @@
     "import torchvision\n",
     "import torchvision.transforms as transforms\n",
     "\n",
-    "from tqdm import tqdm_notebook as pbar\n",
+    "from tqdm import tqdm as pbar\n",
     "from tensorboardX import SummaryWriter\n",
     "from models import *"
    ]
@@ -204,9 +204,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# Train on cuda if available\n",
-    "device = torch.device('cuda:1' if torch.cuda.is_available() else 'cpu')\n",
-    "print(\"Using\", device)"
+    "model = inception_v3(pretrained=True)"
    ]
   },
   {
@@ -216,6 +214,17 @@
     "## 6. Put everything together"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Train on cuda if available\n",
+    "device = torch.device('cuda:1' if torch.cuda.is_available() else 'cpu')\n",
+    "print(\"Using\", device)"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -251,18 +260,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# test_model(model, train_params)"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "# for i, a_model in enumerate(all_models):\n",
-    "#     train_params['description'] = descriptions[i]\n",
-    "#     train_model(a_model, train_params)"
+    "test_model(model, train_params)"
    ]
   },
   {

diff --git a/README.md b/README.md
@@ -3,21 +3,20 @@
 - I changed *number of class, filter size, stride, and padding* in the the original code so that it works with CIFAR-10.
 - I also share the **weights** of these models, so you can just load the weights and use them.
 
-## Statistic of supported models
+## Statistics of supported models
 | No. | Model        | Test Accuracy | Parameters | State Dict Size |
 |-----|--------------|---------------|------------|-----------------|
-| 1   | vgg11_bn     | 92.61%        | 128.813M   | 491 MB          |
-| 2   | vgg13_bn     | 94.27%        | 128.998M   | 492 MB          |
-| 3   | vgg16_bn     | 94.07%        | 134.310M   | 512 MB          |
-| 4   | vgg19_bn     | 94.25%        | 139.622M   | 533 MB          |
-| 5   | resnet18     | 93.48%        | 11.174M    | 43 MB           |
-| 6   | resnet34     | 93.82%        | 21.282M    | 81 MB           |
-| 7   | resnet50     | 94.38%        | 23.521M    | 90 MB           |
-| 8   | mobilenet_v2 | 93.85%        | 2.237M     | 9 MB            |
-| 9   | densenet121  | 94.76%        | 6.956M     | 27 MB           |
-| 10  | googlenet    | 95.08%        | 5.491M     | 21 MB           |
-
-I will add more models...
+| 1   | vgg11_bn     | 92.61%        | 128.813 M  | 491 MB          |
+| 2   | vgg13_bn     | 94.27%        | 128.998 M  | 492 MB          |
+| 3   | vgg16_bn     | 94.07%        | 134.310 M  | 512 MB          |
+| 4   | vgg19_bn     | 94.25%        | 139.622 M  | 533 MB          |
+| 5   | resnet18     | 93.48%        | 11.174 M   | 43 MB           |
+| 6   | resnet34     | 93.82%        | 21.282 M   | 81 MB           |
+| 7   | resnet50     | 94.38%        | 23.521 M   | 90 MB           |
+| 8   | mobilenet_v2 | 93.85%        | 2.237 M    | 9 MB            |
+| 9   | densenet121  | 94.76%        | 6.956 M    | 27 MB           |
+| 10  | googlenet    | 95.08%        | 5.491 M    | 21 MB           |
+| 11  | inception_v3 | 95.41%        | 21.64 M    | 83 MB           |
 
 ## How To Use
 
@@ -40,7 +39,6 @@ my_model = vgg11_bn(pretrained=True)
 transforms.Normalize([0.4914, 0.4822, 0.4465], [0.2023, 0.1994, 0.2010])]
 ```
 
-
 ## Training Hyper-paramters
 - Batch size: 256
 - Number of epochs: 600

diff --git a/models/.ipynb_checkpoints/__init__-checkpoint.py b/models/.ipynb_checkpoints/__init__-checkpoint.py
@@ -2,5 +2,5 @@
 from .resnet import *
 from .vgg import *
 from .densenet import *
-# from .inception import *
+from .inception import *
 from .googlenet import *
diff --git a/models/.ipynb_checkpoints/inception-checkpoint.py b/models/.ipynb_checkpoints/inception-checkpoint.py
@@ -2,17 +2,10 @@
 import torch
 import torch.nn as nn
 import torch.nn.functional as F
-from .utils import load_state_dict_from_url
-
 
 __all__ = ['Inception3', 'inception_v3']
 
 
-model_urls = {
-    # Inception v3 ported from TensorFlow
-    'inception_v3_google': 'https://download.pytorch.org/models/inception_v3_google-1a9a5a14.pth',
-}
-
 _InceptionOuputs = namedtuple('InceptionOuputs', ['logits', 'aux_logits'])
 
 
@@ -32,37 +25,25 @@ def inception_v3(pretrained=False, progress=True, **kwargs):
         transform_input (bool): If True, preprocesses the input according to the method with which it
             was trained on ImageNet. Default: *False*
     """
+    model = Inception3()
     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 = True
-        model = Inception3(**kwargs)
-        state_dict = load_state_dict_from_url(model_urls['inception_v3_google'],
-                                              progress=progress)
+        state_dict = torch.load('models/state_dicts/inception_v3.pt', map_location='cpu')
         model.load_state_dict(state_dict)
-        if not original_aux_logits:
-            model.aux_logits = False
-            del model.AuxLogits
-        return model
-
-    return Inception3(**kwargs)
-
+    return model
 
 class Inception3(nn.Module):
-
-    def __init__(self, num_classes=1000, aux_logits=True, transform_input=False):
+    ## CIFAR10: aux_logits True->False
+    def __init__(self, num_classes=10, aux_logits=False, transform_input=False):
         super(Inception3, self).__init__()
         self.aux_logits = aux_logits
         self.transform_input = transform_input
-        self.Conv2d_1a_3x3 = BasicConv2d(3, 32, kernel_size=3, stride=2)
-        self.Conv2d_2a_3x3 = BasicConv2d(32, 32, kernel_size=3)
-        self.Conv2d_2b_3x3 = BasicConv2d(32, 64, kernel_size=3, padding=1)
-        self.Conv2d_3b_1x1 = BasicConv2d(64, 80, kernel_size=1)
-        self.Conv2d_4a_3x3 = BasicConv2d(80, 192, kernel_size=3)
+
+        ## CIFAR10: stride 2->1, padding 0 -> 1
+        self.Conv2d_1a_3x3 = BasicConv2d(3, 192, kernel_size=3, stride=1, padding=1)
+#         self.Conv2d_2a_3x3 = BasicConv2d(32, 32, kernel_size=3)
+#         self.Conv2d_2b_3x3 = BasicConv2d(32, 64, kernel_size=3, padding=1)
+#         self.Conv2d_3b_1x1 = BasicConv2d(64, 80, kernel_size=1)
+#         self.Conv2d_4a_3x3 = BasicConv2d(80, 192, kernel_size=3)
         self.Mixed_5b = InceptionA(192, pool_features=32)
         self.Mixed_5c = InceptionA(256, pool_features=64)
         self.Mixed_5d = InceptionA(288, pool_features=64)
@@ -99,18 +80,20 @@ def forward(self, x):
             x = torch.cat((x_ch0, x_ch1, x_ch2), 1)
         # N x 3 x 299 x 299
         x = self.Conv2d_1a_3x3(x)
+
+        ## CIFAR10
         # N x 32 x 149 x 149
-        x = self.Conv2d_2a_3x3(x)
+#         x = self.Conv2d_2a_3x3(x)
         # N x 32 x 147 x 147
-        x = self.Conv2d_2b_3x3(x)
+#         x = self.Conv2d_2b_3x3(x)
         # N x 64 x 147 x 147
-        x = F.max_pool2d(x, kernel_size=3, stride=2)
+#         x = F.max_pool2d(x, kernel_size=3, stride=2)
         # N x 64 x 73 x 73
-        x = self.Conv2d_3b_1x1(x)
+#         x = self.Conv2d_3b_1x1(x)
         # N x 80 x 73 x 73
-        x = self.Conv2d_4a_3x3(x)
+#         x = self.Conv2d_4a_3x3(x)
         # N x 192 x 71 x 71
-        x = F.max_pool2d(x, kernel_size=3, stride=2)
+#         x = F.max_pool2d(x, kernel_size=3, stride=2)
         # N x 192 x 35 x 35
         x = self.Mixed_5b(x)
         # N x 256 x 35 x 35

diff --git a/models/__init__.py b/models/__init__.py
@@ -2,5 +2,5 @@
 from .resnet import *
 from .vgg import *
 from .densenet import *
-# from .inception import *
+from .inception import *
 from .googlenet import *
diff --git a/models/__pycache__/__init__.cpython-37.pyc b/models/__pycache__/__init__.cpython-37.pyc
diff --git a/models/__pycache__/inception.cpython-37.pyc b/models/__pycache__/inception.cpython-37.pyc
diff --git a/models/__pycache__/shufflenetv2.cpython-37.pyc b/models/__pycache__/shufflenetv2.cpython-37.pyc