Files
obsidian/phd/weekly-report/25/251123.md

1.3 KiB

Objectives

  • Auto LLM inference config tuner

Key Results

  • [6/10] Build the first version auto tuner system
  • [7/10] Check the current situation of parallelism config optimization
  • [4/10] Understand the possibility/challenges in LLM inference compute graph arrangement automatically
  • [0/10] Trace vLLM compute graph and data flow
  • [3/10] Implement a minimal Rust inference framework
  • [1/10] Define the IR for automatic optimization
  • [5/10] Profile different parallelism setup with real trace and analysis their difference
  • [0/10] Meta-analysis for the theory maximum improvement with heterogenous setup [offtrack]

Last Week

  • [KR2] Benchmark different configs in different hardware, prove that different hardware and different workload will cause different trends of performance change. 5f2c1ec3 ~ 65d05520
  • [KR1] Build a precise workload generator from real workload. Benchmark on quite similar generated workloads and find that even the similar workloads still trigger different performance.

Next Week

  • Find the root cause of performance gap under similar workloads.