PyTorch 相当于 keras 模型

Question

我正在学习 PyTorch，并尝试将 yolov3 模型（来自 keras）转换为 PyTorch。现在我面临的问题是有一个代码片段我发现很难将其转换为 PyTorch。现在

def yolo_body(inputs, num_anchors, num_classes):
    """Create YOLO_V3 model CNN body in Keras."""
    darknet = Model(inputs, darknet_body(inputs))    
    x, y1 = make_last_layers(darknet.output, 512, num_anchors*(num_classes+5))
    x = compose(
            DarknetConv2D_BN_Leaky(256, (1,1)),
            UpSampling2D(2))(x)
    x = Concatenate()([x,darknet.layers[152].output])
    #---- further code

我想知道 Keras 的 Model 的 pytorch 等价物是什么。我的 darknet_body 实现是

class yolov3:
    ----
    ----
    def darknet_body(self, x):
        """Darknent body having 52 Convolution2D layers"""
        x = self.DarknetConv2D_BN_Leaky(32, (3, 3))(x)
        x = self.resblock_body(x, 64, 1)
        x = self.resblock_body(x, 128, 2)
        x = self.resblock_body(x, 256, 8)
        x = self.resblock_body(x, 512, 8)
        x = self.resblock_body(x, 1024, 4)
        return x

和 resblock_body 实现是。

class yolov3:
    ----
    ----
    def resblock_body(self, x, num_filters, num_blocks):
        """A series of resblocks starting with a downsampling Convolution2D"""
        # Darknet uses left and top padding instead of 'same' mode
        x = torch.nn.ZeroPad2d(((1, 0), (1, 0)))(x)
        x = self.DarknetConv2D_BN_Leaky(num_filters, (3, 3), strides=(2, 2))(x)
        for i in range(num_blocks):
            y = compose(
                self.DarknetConv2D_BN_Leaky(num_filters // 2, (1, 1)),
                self.DarknetConv2D_BN_Leaky(num_filters, (3, 3)),
            )(x)
            x = torch.add(x, y)
        return x

我还想知道如何获得像 darknet.layers[152] 这样的层号，如 yolo_body 中所做的那样。

谢谢。