Intel corporation (20240338558). ADAPTIVE BUFFER MANAGEMENT TO SUPPORT DYNAMIC TENSOR SHAPE IN DEEP NEURAL NETWORK APPLICATIONS simplified abstract
Contents
ADAPTIVE BUFFER MANAGEMENT TO SUPPORT DYNAMIC TENSOR SHAPE IN DEEP NEURAL NETWORK APPLICATIONS
Organization Name
Inventor(s)
ADAPTIVE BUFFER MANAGEMENT TO SUPPORT DYNAMIC TENSOR SHAPE IN DEEP NEURAL NETWORK APPLICATIONS - A simplified explanation of the abstract
This abstract first appeared for US patent application 20240338558 titled 'ADAPTIVE BUFFER MANAGEMENT TO SUPPORT DYNAMIC TENSOR SHAPE IN DEEP NEURAL NETWORK APPLICATIONS
The disclosure pertains to adaptive buffer management for supporting a dynamic tensor shape in a deep neural network (DNN). An apparatus for the DNN may include processor circuitry configured to:
- Determine if the tensor shape of an input tensor of an object in the DNN is dynamic and exists in a shape buffer pool.
- Execute the object using a compilation result stored in the shape buffer pool when the tensor shape of the input tensor is dynamic and present in the shape buffer pool.
- Trigger the compilation procedure to perform just-in-time (JIT) compilation for the object to obtain the compilation result when the tensor shape of the input tensor is dynamic and not found in the shape buffer pool.
Key Features and Innovation: - Adaptive buffer management for dynamic tensor shapes in a DNN. - Utilization of a shape buffer pool to store compilation results. - Just-in-time (JIT) compilation triggered for objects with dynamic tensor shapes.
Potential Applications: - Deep learning applications requiring dynamic tensor shapes. - Real-time data processing where tensor shapes change frequently.
Problems Solved: - Efficient management of dynamic tensor shapes in DNNs. - Reduced computational overhead for objects with changing tensor shapes.
Benefits: - Improved performance and resource utilization in DNNs. - Enhanced flexibility for handling varying tensor shapes. - Streamlined compilation process for dynamic objects.
Commercial Applications: Title: Dynamic Tensor Shape Management for Deep Neural Networks Commercial Uses: AI-powered systems, image and speech recognition software, autonomous vehicles. Market Implications: Increased efficiency and speed in deep learning applications, potential for new AI-driven products and services.
Prior Art: Readers can explore prior research on adaptive buffer management in deep learning systems and JIT compilation techniques for dynamic tensor shapes.
Frequently Updated Research: Stay informed about the latest advancements in adaptive buffer management for dynamic tensor shapes in DNNs through academic journals and conferences.
Questions about Adaptive Buffer Management for Dynamic Tensor Shapes: 1. How does adaptive buffer management improve the performance of deep neural networks? 2. What are the key challenges in implementing JIT compilation for objects with dynamic tensor shapes?
Original Abstract Submitted
the disclosure relates to adaptive buffer management to support a dynamic tensor shape in a dnn. an apparatus for the dnn may include processor circuitry configured to: determine whether a tensor shape of an input tensor of an object in the dnn is dynamic and exists in a shape buffer pool; run the object by use of a compilation result for the object stored in the shape buffer pool when the tensor shape of the input tensor is dynamic and exists in the shape buffer pool; and invoke the compilation procedure to perform jit compilation for the object so as to get the compilation result for the object when the tensor shape of the input tensor is dynamic and does not exist in the shape buffer pool.
