The Industrial Internet of Things (IIoT) refers to the deployment of a vast array of IoT devices and wireless access points in industrial infrastructures with the purpose of acquiring intelligent services. It is considered an essential physical information platform for the implementation of Industry 4.0. The integration of edge computing and IIoT has become a practical solution to address the increasing demand for efficient data processing with minimal latency and maximum throughput in industrial settings. The cloud experiences substantial latency and bandwidth issues, despite its vast capacities. Transferring processing tasks to edge servers positioned at the periphery of the network can greatly reduce latency. Hence, the time-critical task is allocated to edge servers, with the anticipation that it would be run prior to the specified deadline. Edge computing resources are somewhat scarce in comparison to the cloud. If resources are not effectively allocated and used, the performance can be impacted due to the characteristics of IIoT, including heterogeneity, wireless network, real-time capabilities, and high data generation. These problems lead to delays, inefficiency in energy and bandwidth utilization, and a decline in performance. The main disadvantage of the existing systems is the significant cost and delay involved in making offloading decisions. Prioritizing task offloading to the edge before task generation is essential for minimizing both the delay in making offload decisions and the associated costs. We have suggested an attention-enhanced LSTM model to optimize the allocation of computing workloads between industrial terminal devices and remote edge servers. The approach utilizes predictive analytics to anticipate task attributes, facilitating well-informed decisions on when to transfer computation. The results demonstrate that the proposed model significantly minimizes offloading decision delay and failure tasks compared to existing offloading strategies.