Music recommendations have become an important part of everyday life. With the increasing amount of music available, it has become increasingly difficult to find suitable music content. To address this issue, recommendation systems have emerged as a solution to personalize song suggestions for users. Traditional recommendation systems often fail to capture sequential patterns of user interaction. To understand these patterns, recommendation systems are based on a sequential approach. However, traditional sequential recommendation system models do not sufficiently capture dependencies on long-term temporal patterns. In this study, we apply a combination of graph neural networks with attention mechanisms to the Music4All dataset to address sequential music recommendations. This approach, which adapts the GASM framework, is applied for the first time to this music dataset, which has greater diversity and user distribution, providing new insights into the effectiveness of this model in the music recommendation domain. Graph neural networks capture the relationships between users and items, as well as between items themselves, while the attention mechanism captures user interest patterns over time within a session. We also incorporate long-term interests and dynamic interests to enrich the context of user preferences. By applying this model to the Music4All dataset, the model can recommend the next music track with a hit ratio evaluation ranging from 32.69 to 49.16 and a mean reciprocal rank ranging from 23.34 to 30.71. These results show that our approach can effectively capture sequential patterns in music in large-scale datasets.