Need to customize vLLM? Don't fork it. 🔌
vLLM's plugin system lets you inject surgical modifications without maintaining a fork or monkey-patching entire modules. Blog by Dhruvil Bhatt from AWS SageMaker 👇
Why plugins > forks:
• vLLM releases every 2 weeks with 100s of PRs merged
• Forks require constant rebasing & conflict resolution
• Monkey patches break on every vLLM upgrade
How it works:
• Use VLLMPatch[TargetClass] for precise, class-level mods
• Register via vllm.general_plugins entry point
• Control patches with env vars (VLLM_CUSTOM_PATCHES)
• Version-guard with min_vllm_version decorator
Example: Add priority scheduling to vLLM's scheduler in ~20 lines. One Docker image serves multiple models with different patches enabled via environment variables.
The plugin loads in ALL vLLM processes (main, workers, GPU/CPU) before any inference starts—ensuring consistent behavior across distributed setups.
Read the full implementation guide with code examples:
blog.vllm.ai/2025/11/20/vllm…
The strategic prioritization of targets by class offers a competitive edge in drug discovery.
Causaly's AI enables researchers to identify and prioritize targets by class, providing a more granular view of targets for a disease: hubs.la/Q02hSgZX0#TargetClass#AI
Day2 of the #100DaysOfCode challenge.
Made this Nutrition Facts label in freeCodeCamp.
What I've learned:
- create divider
- how to set border-box in CSS
- format <p> using flex box
- CSS selector to exclude [.targetClass] - e.g. [example:not(.targetClass)]
This is actually part of shack spec. It is just a syntactic sugar to define the targetClass as schema:Person. I don't think any shacl verifier or triple store does anything else with this formally.
@narendramodi
@ind_parliament
Good initiative👍
Next #targetclass sud be #MPnMLA..start curtailing subsidy n make them feel like leading life as ordinary man not as a king like ancient ages.After substantial increases in salaries n allowances why subsidy is still in support.