How can I send the (batch_size, 8, 8, 2048) tensor to lstm hidden layer?

I use the pre-trained inception v3 model to get the Mixed_7c feature maps. What should I do next?

How can I send the features to lstm hidden layer?

Thanks!