Glaucoma disease is a highly critical eye condition that causes permanent blindness resulting from Intraocular Hypertension which damages the Optic Nerve Head. It is very difficult to diagnose glaucoma, owing to its asymptomatic nature. Early detection of glaucoma is crucial to cure a patient from adverse consequences which can be done by analyzing retinal fundus images. In this article we introduce Lw-CSAtt-Net-a Lightweight multi-Attention based Network for Glaucoma Prediction using Retinal Fundus Images. Our proposed Lw-CSAtt-Net integrates a novel Channel-Spatial Attention (CSAtt) module with the light-weight Depthwise Separable Convolutional Neural Network. The proposed method is evaluated on ACRIMA database. The images were preprocessed using CLAHE (Contrast Limited Adaptive Histogram Estimation) to enhance the contrast of the images. To justify the implication of CLAHE on Lw-CSAtt-Net, we evaluated the performance on original ACRIMA as well as CLAHE enhanced ACRIMA images. The Lw-CSAtt-Net secured an average test accuracy score of 0.99296 on the original ACRIMA Database. However, Lw-CSAtt-Net secured an average test accuracy score of 0.98591 on the CLAHE Enhanced ACRIMA Database.