TPU Coder Cup is the pinnacle showdown in the field of artificial intelligence. Sponsored by SOPHGO, this high-level AI competition platform has attracted the participation of numerous university students, open-source community developers, and SOPHGO ecosystem partners. The competition features a series of events including the AI Application Challenge, AI Algorithm Innovation Competition, and Operator Ranking Competition. These events unfold progressively based on different algorithm and software technology stack levels, and the competition problem sets are open-sourced. The goal of the competition is to discover and cultivate AI algorithm talents, fully leverage the powerful computing capabilities of TPU and RISC-V hardware, and solve social and business problems through innovative algorithms and software. Moreover, the competition stimulates the vitality of open-source innovation, cultivates talents in open-source practices, and provides support and assistance for the development of open-source ecosystems.
Here are past competitions to help you better understand the TPU Coder Cup
Index |
Competition |
|---|---|
| 1 | The 1st AI Algorithm Innovation Competition - Realization of L1 Video Structure Algorithm |
Index |
Competition |
|---|---|
| 1 | The 1st TPU Programming Contest - okkernel programming |
| 2 | The 2nd TPU Programming Contest - okkernel programming |
SOPHON.NET provides a cloud development space integrating training and reasoning to provide convenient cloud resources for algorithm development and testing which includes Standard Cloud Development Space, TPU Programming Competition Cloud Space, and Cloud Desktop Space. So algorithm development is no longer limited to hardware. The TPU Programming Competition Cloud Space offers a dedicated cloud test environment for the TPU Programming Competition, utilizing the Ubuntu system. Cloud Space Usage Tutorial introduces the application and usage details of the platform.
Sophgo provides a series of original deep learning development kits including BMNNSDK, Sophon-Inference, TPU-MLIR, Model Zoo, etc. Brief description and links of are as follows.
- BMNNSDK, a deep learning SDK based on BM AI chip, with its powerful tools, you can deploy the deep learning application in the runtime environment, and deliver the maximum inference throughput and efficiency.
- Sophon-Inference, Sophon Artificial Intelligent Library for online deployment(SAIL), a wrapper of bmruntime, bmdecoder, bmcv, bmlib. It provides both C++ and Python APIs utomatically managing memory of tensors. It can be compiled on a Linux/Windows10 host with x86 architecture to generate a corresponding version of the dynamic library and Python Wheel file, and can also configure a cross-compiler and compile a dynamic library and Python Wheel file under other architectures.
- TPU-MLIR, an open-source machine-learning compiler based on MLIR for TPU. This project provides a complete toolchain, which can convert pre-trained neural networks from different frameworks into binary files bmodel that can be efficiently operated on TPUs.
TPU Coder Cup Committee is committed to building and maintaining a high-level TPU competition platform. The committee mainly sets up two departments, the Case Department and the Publicity Department. We are eager to find like-minded partners to join us! For more details, please refer to the link TPU Coder Cup Committe Recruitment Notice.