Repository navigation

Add blog post: Running a big LLM across multiple GPUs with vLLM - #265

Merged

saiyam1814 merged 20 commits into

mainfrom

blog/two-gpu-vllm

Sep 1, 2026

Commits on Aug 18, 2026
Commits on Aug 23, 2026
Commits on Aug 28, 2026
- Cut the post down to the runbook so readers with a deadline are not paying for theory
  shkatara
  committed
  08b151e
Commits on Aug 31, 2026
- Restore the evidence and the why, and drop the part-two promises
  
  saiyam1814
  and
  claude
  committed
  c793d7d
- Regenerate the cake diagram as SVG in the house Excalidraw style
  saiyam1814
  committed
  abb7e00
- Fix the review findings: a broken table, a duplicated line, and HF_HOME
  
  saiyam1814
  and
  claude
  committed
  964f340
- Trim the deep mechanics back out, and say what the 2 in the KV formula is
  
  saiyam1814
  and
  claude
  committed
  5f2a7ec
- Put the cake illustration back, with the duplicated axis label removed
  
  saiyam1814
  and
  claude
  committed
  c87a2ee
- Cut duplicated log lines and repeated benchmark boilerplate
  
  saiyam1814
  and
  claude
  committed
  00b7452
- Credit Utho, drop the eviction story, merge the flags step into the command
  
  saiyam1814
  and
  claude
  committed
  4c68fc0
Commits on Sep 1, 2026
- Remove docker explaination and reduce paragraph lengths
  shkatara
  committed
  7299f07
- Drop --no-enable-prefix-caching from the benchmark command; it is not a bench flag
  
  saiyam1814
  and
  claude
  committed
  fe68ffe
- Merge remote-tracking branch 'origin/main' into blog/two-gpu-vllm-pub
  saiyam1814
  committed
  4a5e09d
- Re-measure all three configurations in one sitting at 640 prompts
  
  saiyam1814
  and
  claude
  committed
  e65253f