Inception Implementation Problem

[yunhenk]

class Inception1(nn.Module):
    def __init__(self, input_dim, conv_dim=64):
        super(Inception1, self).__init__()
        self.cnn = nn.Sequential(
            nn.Conv1d(input_dim, conv_dim, kernel_size=1),
            nn.ReLU(),
            nn.Conv1d(conv_dim, conv_dim, kernel_size=2),
            nn.ReLU()
        )
        self.global_avg_pool = nn.AvgPool1d(input_dim)
        self.global_max_pool = nn.MaxPool1d(input_dim)

    def forward(self, x):
        out = self.cnn(x)
        avg_pool, max_pool = mean_max(x)
        return torch.cat((avg_pool, max_pool), dim=1)

out in forward fuction seems not used? the same problem also exists in Inception2,Inception3

[iblahimovic]

also found this problem. And some problem exists in the implement of Enhanced RCNN

[daviddwlee84]

Thanks, the forward function of the Inceptions has been fixed.