Top all-time · #10518
Lequn Chen
Personal blog · le.qun.ch/en/blogsystemsai
Lequn Chen's personal blog at le.qun.ch covers systems and ai. Since 2019, 7 different Hacker News readers have linked it 10 times, most often “Dissecting Batching Effects in GPT Inference”. It is still writing: the last post was on 2026-09-08, and you can follow it by feed.
Too few sampled comments (3) to say how often HN recommends it.?
Why HN links it
HN readers link Lequn Chen's blog for technical posts on LLM serving and fine-tuning, such as just-in-time LoRA fine-tuning, and on saturating AWS instance bandwidth with libfabric EFA and RDMA.
Summarised by Claude from a sample of 3 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Lequn Chen, abcdabcd987
Most-linked posts?
- Dissecting Batching Effects in GPT Inference3 people · 3 mentions · first linked 2023-05-18
- Potentials of Multitenancy Fine-Tuned LLM Serving2 people · 2 mentions · first linked 2023-09-14
What Hacker News says?
Thank you! We are also very excited about combining the fast fine-tuning and efficient serving. In fact, what you just said is very related to one of our very first motivations. In my previous blog post [1], I call this scheme "Just-in-time Fine-tuning". Our previous measurement is that, for a medium-sized webpage (~10K tokens), it takes around 30 seconds to 2 minutes to finetune a LoRA model. Another good side of t…
Comment linking /en/blog/2023/09/11/multi-lora-potentials/
on “Punica: Serving multiple LoRA finetuned LLM as one”Submitted Harnessing 3200 Gbps Network: A Journey with RDMA, EFA, and Libfabric as a story · 11 points
HN discussionIt's a latency vs. throughput tradeoff. I was surprised as well. But most GPUs can do 32 inferences in the same time as they can do 1 inference. They have all the parallel units required and there are significant setup costs that can be amortized since all the inferences share the same model, weights, etc. https://groq.com/wp-content/uploads/2020/05/GROQP002_V2.2.pd... the "batching" section of https://docs.nvidia.c…
Comment linking /en/blog/2023/05/13/transformer-batching/
on “GPU-Accelerated LLM on a $100 Orange Pi”Submitted Explorations of RDMA in LLM Systems as a story · 8 points
HN discussionThis? https://le.qun.ch/en/blog/2024/12/25/libfabric-efa-1-rdma/
Comment linking /en/blog/2024/12/25/libfabric-efa-1-rdma/
on “Ask HN: Searching] Blog Post About Saturating AWS Instance Bandwidth”Submitted Journey to 2-second Inter-node RL Weight Transfer as a story · 3 points
HN discussionSubmitted How we discovered why C++ exceptions disappear in stacktrace as a story · 3 points
HN discussionSubmitted Potentials of Multitenancy Fine-Tuned LLM Serving as a story · 2 points
HN discussionSubmitted Dissecting Batching Effects in GPT Inference as a story · 2 points
HN discussionSubmitted Dissecting Batching Effects in GPT Inference as a story · 1 points
HN discussion
