Posts by Chao Li
Accelerating Large-Scale LLM Inference on AMD Instinct MI350X/MI355X with Eagle3 and AMD Quark
- 03 July 2026
Large language model (LLM) inference is increasingly constrained by autoregressive decoding. Even when prefill is highly optimized, the decode phase still generates tokens one step at a time, and each step typically requires running the full target model. For large mixture-of-experts and attention-heavy models such as Kimi-K2.5 and MiniMax-M2.5, this sequential pattern limits serving throughput and increases latency for real-time applications.
Low Kruskal-Rank Adaptation
- 11 June 2026
In this blog, you will explore how to enhance Low-Rank Adaptation (LoRA) which uses matrix rank, and replace it with Kruskal rank for efficient training. LoRA is one of the most widely used parameter-efficient fine-tuning (PEFT) methods for adapting pre-trained large language models (LLMs) to downstream tasks. Although LoRA significantly reduces the number of trainable parameters and lowers fine-tuning costs, its performance is often limited by the inherent low-rank assumption. We revisit the notion of rank for LoRA update matrices and show that the standard matrix rank fails to capture duplicated directions and redundancy in the update subspace. Motivated by this analysis, we argue that the Kruskal rank offers a more informative criterion for characterizing update diversity. We therefore propose Low Kruskal Rank Adaptation (LoKRA), a new PEFT algorithm with provable theoretical guarantees that mitigates the limitations of LoRA. We further introduce LoKRA+, an enhanced variant that provides a tighter theoretical lower bound on the Kruskal rank and yields stronger empirical performance. Experiments on multiple LLMs show that our approach consistently outperforms LoRA and other baselines, establishing state-of-the-art performance across a range of benchmarks. The paper is accepted by ICML 2026 (paper link), and the code is publicly available on GitHub.
hipBLASLt Online GEMM Tuning
- 19 March 2026
This blog post introduces the integration of hipBLASLt Online GEMM Tuning into LLM frameworks, illustrated through an example implementation of RTP-LLM. Developed by the AMD Quark Team, hipBLASLt Online Tuning provides a user-friendly approach to improving GEMM performance by enabling runtime tuning without requiring additional offline tuning steps.
Day 0 Developer Guide: hipBLASLt Offline GEMM Tuning Script
- 05 November 2025
This blog post focuses on optimizing the performance of a real model using the QuickTune script, illustrated with an example of offline GEMM tuning for the Qwen model on an AMD MI308 GPU. Developed by the AMD Quark Team, the QuickTune script delivers significant GEMM performance improvements with minimal time overhead. QuickTune is an advanced tool for hipBLASLt offline GEMM tuning. It allows users to complete offline tuning with one click, instead of using hipblaslt-bench to tune the model manually.