RNN Encoder + CNN for sentence tagging

Hello everyone,
I am very new to pytorch. I was hoping to create a sentence classifier using the architecture presented in the image.

I want CNN filters of sizes 2,3,4 and 5 grams acting as feature extractors and I also want to preserve the Last hidden layer representation from the forward-RNN and the backward-RNN into a 2 layer fully connected network.

Are there any codes available on github which are similar to this architecture? Any recommended resources to read from?