Abstract: The present specification discloses a computing device architecture and method for executing computationally intensive attention mechanisms, as commonly utilized in transformer models for artificial intelligence. In an embodiment, the device includes a row of interconnected processing elements (PEs), each linked to dedicated coefficient memory units (CRAMs) and controlled by a central controller. By employing a diagonal-offset and transposition technique for matrix coefficients, the architecture enables efficient execution of Generalized Matrix-Vector (GEMV) operations.

Abstract: There is provided a method and apparatus for remote differential compression (RDC) and data deduplication. According to embodiments, when a sending device acquires a new target file, the following steps are performed. Initially, Jaccard segmentation is performed, followed by performing identity-based segment deduplication and similarity-based segment deduplication. The transmission of the target file in the deduplicated form to the recipient device is subsequently performed. The recipient device can then rebuild the original target file from the deduplicated form thus replicating the target file at the recipient device with the target file originally present at the sending device.