Quanzhou L.

Systematically learned about computer science and have experience in algorithm competitions and CTF competitions. Focused on operating system, compiler, hardware system related content, and some simple data structure design - linked list, heap, SBT, etc. Can do software-backend works, familiar with Django/Springboot. Participated in ASC22 and was responsible for the Yuan Large Language Model Challenge. I learned the basic knowledge of DeepSpeed ​​and Megatron-LM, understood several optimization methods mainly used in the training process of very large models, and used this method for subsequent training and parameter adjustment. Finally won the second prize in the preliminary round. Responsible for pwn and crypto questions in the team, master basic ROP technology, including ret2shellcode, ret2libc and other stack overflow techniques, and be able to use gadgets; understand the data structure and algorithm of ptmalloc2 memory allocator, and be able to off-by-one, unlink and UAF, etc. Heap utilization methods. Can use cryptography tools to solve some problems related to hash algorithms. Finally won the second prize in the North China Division.