add md5 and pytorch weights

README.md

@@ -0,0 +1,20 @@
+
+
+```python
+import trace_layer2 as models
+import torch
+
+x=torch.randn(1, 3, 224, 224)
+
+state_dict = torch.load('swav_imagenet_layer2.pt', map_location='cpu')
+
+model = models.resnet50w2()
+model.load_state_dict(state_dict)
+model.eval()
+feature = model(x)
+
+traced_model = torch.jit.load('traced_swav_imagenet_layer2.pt', map_location='cpu')
+traced_model.eval()
+feature = traced_model(x)
+
+```

md5.txt

@@ -0,0 +1,3 @@
+2d46caef59dd661c695114df1c161733  swav_imagenet_layer2.pt
+d4c65a44119dcb606e5f2c4efe986847  swav_imagenet_layer2_sim.onnx
+428ac1fe949cc24c65a7df974b32bc2f  traced_swav_imagenet_layer2.pt

swav_imagenet_layer2.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1923d34e543b3120e23f4a62c8ce83d204530329d0c9a19c18008b1d5e9dc25d
+size 23087521

trace_layer2.py

@@ -0,0 +1,303 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+#
+
+import torch
+import torch.nn as nn
+
+
+def conv3x3(in_planes, out_planes, stride=1, groups=1, dilation=1):
+    """3x3 convolution with padding"""
+    return nn.Conv2d(
+        in_planes,
+        out_planes,
+        kernel_size=3,
+        stride=stride,
+        padding=dilation,
+        groups=groups,
+        bias=False,
+        dilation=dilation,
+    )
+
+
+def conv1x1(in_planes, out_planes, stride=1):
+    """1x1 convolution"""
+    return nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride, bias=False)
+
+
+class BasicBlock(nn.Module):
+    expansion = 1
+    __constants__ = ["downsample"]
+
+    def __init__(
+        self,
+        inplanes,
+        planes,
+        stride=1,
+        downsample=None,
+        groups=1,
+        base_width=64,
+        dilation=1,
+        norm_layer=None,
+    ):
+        super(BasicBlock, self).__init__()
+        if norm_layer is None:
+            norm_layer = nn.BatchNorm2d
+        if groups != 1 or base_width != 64:
+            raise ValueError("BasicBlock only supports groups=1 and base_width=64")
+        if dilation > 1:
+            raise NotImplementedError("Dilation > 1 not supported in BasicBlock")
+        # Both self.conv1 and self.downsample layers downsample the input when stride != 1
+        self.conv1 = conv3x3(inplanes, planes, stride)
+        self.bn1 = norm_layer(planes)
+        self.relu = nn.ReLU(inplace=True)
+        self.conv2 = conv3x3(planes, planes)
+        self.bn2 = norm_layer(planes)
+        self.downsample = downsample
+        self.stride = stride
+
+    def forward(self, x):
+        identity = x
+
+        out = self.conv1(x)
+        out = self.bn1(out)
+        out = self.relu(out)
+
+        out = self.conv2(out)
+        out = self.bn2(out)
+
+        if self.downsample is not None:
+            identity = self.downsample(x)
+
+        out += identity
+        out = self.relu(out)
+
+        return out
+
+
+class Bottleneck(nn.Module):
+    expansion = 4
+    __constants__ = ["downsample"]
+
+    def __init__(
+        self,
+        inplanes,
+        planes,
+        stride=1,
+        downsample=None,
+        groups=1,
+        base_width=64,
+        dilation=1,
+        norm_layer=None,
+    ):
+        super(Bottleneck, self).__init__()
+        if norm_layer is None:
+            norm_layer = nn.BatchNorm2d
+        width = int(planes * (base_width / 64.0)) * groups
+        # Both self.conv2 and self.downsample layers downsample the input when stride != 1
+        self.conv1 = conv1x1(inplanes, width)
+        self.bn1 = norm_layer(width)
+        self.conv2 = conv3x3(width, width, stride, groups, dilation)
+        self.bn2 = norm_layer(width)
+        self.conv3 = conv1x1(width, planes * self.expansion)
+        self.bn3 = norm_layer(planes * self.expansion)
+        self.relu = nn.ReLU(inplace=True)
+        self.downsample = downsample
+        self.stride = stride
+
+    def forward(self, x):
+        identity = x
+
+        out = self.conv1(x)
+        out = self.bn1(out)
+        out = self.relu(out)
+
+        out = self.conv2(out)
+        out = self.bn2(out)
+        out = self.relu(out)
+
+        out = self.conv3(out)
+        out = self.bn3(out)
+
+        if self.downsample is not None:
+            identity = self.downsample(x)
+
+        out += identity
+        out = self.relu(out)
+
+        return out
+
+
+class ResNet(nn.Module):
+    def __init__(
+            self,
+            block,
+            layers,
+            num_classes=1000,
+            zero_init_residual=False,
+            groups=1,
+            widen=1,
+            width_per_group=64,
+            replace_stride_with_dilation=None,
+            norm_layer=None,
+    ):
+        super(ResNet, self).__init__()
+        if norm_layer is None:
+            norm_layer = nn.BatchNorm2d
+        self._norm_layer = norm_layer
+
+        self.inplanes = width_per_group * widen
+        self.dilation = 1
+        if replace_stride_with_dilation is None:
+            # each element in the tuple indicates if we should replace
+            # the 2x2 stride with a dilated convolution instead
+            replace_stride_with_dilation = [False, False, False]
+        if len(replace_stride_with_dilation) != 3:
+            raise ValueError(
+                "replace_stride_with_dilation should be None "
+                "or a 3-element tuple, got {}".format(replace_stride_with_dilation)
+            )
+        self.groups = groups
+        self.base_width = width_per_group
+
+        num_out_filters = width_per_group * widen
+        self.conv1 = nn.Conv2d(
+            3, num_out_filters, kernel_size=7, stride=2, padding=3, bias=False
+        )
+        self.bn1 = norm_layer(num_out_filters)
+        self.relu = nn.ReLU(inplace=True)
+        self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
+        self.layer1 = self._make_layer(block, num_out_filters, layers[0])
+        num_out_filters *= 2
+        self.layer2 = self._make_layer(
+            block, num_out_filters, layers[1], stride=2, dilate=replace_stride_with_dilation[0]
+        )
+        #num_out_filters *= 2
+        #self.layer3 = self._make_layer(
+        #    block, num_out_filters, layers[2], stride=2, dilate=replace_stride_with_dilation[1]
+        #)
+        #num_out_filters *= 2
+        #self.layer4 = self._make_layer(
+        #    block, num_out_filters, layers[3], stride=2, dilate=replace_stride_with_dilation[2]
+        #)
+        self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
+        #self.fc = nn.Linear(512 * block.expansion * widen, num_classes)
+
+        # Zero-initialize the last BN in each residual branch,
+        # so that the residual branch starts with zeros, and each residual block behaves like an identity.
+        # This improves the model by 0.2~0.3% according to https://arxiv.org/abs/1706.02677
+        if zero_init_residual:
+            for m in self.modules():
+                if isinstance(m, Bottleneck):
+                    nn.init.constant_(m.bn3.weight, 0)
+                elif isinstance(m, BasicBlock):
+                    nn.init.constant_(m.bn2.weight, 0)
+
+    def _make_layer(self, block, planes, blocks, stride=1, dilate=False):
+        norm_layer = self._norm_layer
+        downsample = None
+        previous_dilation = self.dilation
+        if dilate:
+            self.dilation *= stride
+            stride = 1
+        if stride != 1 or self.inplanes != planes * block.expansion:
+            downsample = nn.Sequential(
+                conv1x1(self.inplanes, planes * block.expansion, stride),
+                norm_layer(planes * block.expansion),
+            )
+
+        layers = []
+        layers.append(
+            block(
+                self.inplanes,
+                planes,
+                stride,
+                downsample,
+                self.groups,
+                self.base_width,
+                previous_dilation,
+                norm_layer,
+            )
+        )
+        self.inplanes = planes * block.expansion
+        for _ in range(1, blocks):
+            layers.append(
+                block(
+                    self.inplanes,
+                    planes,
+                    groups=self.groups,
+                    base_width=self.base_width,
+                    dilation=self.dilation,
+                    norm_layer=norm_layer,
+                )
+            )
+
+        return nn.Sequential(*layers)
+
+    def forward(self, x):
+        x = self.conv1(x)
+        x = self.bn1(x)
+        x = self.relu(x)
+        x = self.maxpool(x)
+        x = self.layer1(x)
+        x = self.layer2(x)
+        #x = self.layer3(x)
+        #x = self.layer4(x)
+
+        x = self.avgpool(x)
+        x = torch.flatten(x, 1)
+        #x = self.fc(x)
+
+        return x
+
+
+def resnet50(**kwargs):
+    return ResNet(Bottleneck, [3, 4, 6, 3], **kwargs)
+
+
+def resnet50w2(**kwargs):
+    return ResNet(Bottleneck, [3, 4, 6, 3], widen=2, **kwargs)
+
+
+def resnet50w4(**kwargs):
+    return ResNet(Bottleneck, [3, 4, 6, 3], widen=4, **kwargs)
+
+
+def resnet50w5(**kwargs):
+    return ResNet(Bottleneck, [3, 4, 6, 3], widen=5, **kwargs)
+
+
+if __name__ == '__main__':
+    import onnxruntime as ort
+    x=torch.rand(1,3,224,224)
+    model = resnet50w2()
+    model.eval()
+
+    swav_state_dict = torch.load('/opt/software/github/he-ai/swav_RN50w2_400ep_pretrain.pth.tar')
+
+    for k in list(swav_state_dict.keys()):
+        if k.startswith('module.layer3') or k.startswith('module.layer4') or k.startswith('module.pro'):del swav_state_dict[k]
+
+    for k in list(swav_state_dict.keys()):
+        swav_state_dict[k.replace('module.', '')] = swav_state_dict[k]
+        del swav_state_dict[k]
+    msg = model.load_state_dict(swav_state_dict, strict=False)
+    print(msg)
+    torch.save(swav_state_dict, 'swav_imagenet_layer2.pt')
+
+    traced_script_module = torch.jit.trace(model, x)
+    traced_script_module.save("traced_swav_imagenet_layer2.pt")
+    traced_feature = traced_script_module(x).detach().cpu().numpy()
+    print(traced_feature)
+    print(model(x))
+
+
+    dynamic_axes={"x": {0:"batch_size"}, 'feature': {0:'batch_size'}}
+    torch.onnx.export(model, x, "swav_imagenet_layer2.onnx", verbose=False, input_names=['x'], output_names=['feature'], dynamic_axes=dynamic_axes, do_constant_folding=True)
+    ort_session = ort.InferenceSession("swav_imagenet_layer2.onnx")
+    onnx_outputs = ort_session.run(None, {'x':x.numpy()})
+    print(onnx_outputs[0])
+