A common method for describing heterogeneous data is through graphs. Numerous research works on graph learning have highlighted methods for extracting, interpreting, and summarizing graph data. Summarizing graph data is doing more expansion as a result of the longer duration of significant applications and accurate interpretation of the data's hidden characteristics through the use of deep learning-based graph representation. The model that is most frequently used to address the text creation issue is called Fastformer. Effective understanding of the sentence semantics in the provided document is also achieved through the usage of the Fastformer model. In order to provide a summary, we used a graph-based extractive summarizing technique in this study, using a few chosen sentences as input for an abstractive model. This study aims to assess and compare the efficacy of several model architectures, including Transformer and RNN-based models, and other language models, including Transformer and Bi-LSTM Seq-Seq plus Attention. To test the result, we employed the ROUGE Score. We also investigate the behavior of the model with different optimizers and hyper parameters.