marlin_v100 is a minimal standalone Marlin development workspace extracted from the vLLM main tree, focused on source development, minimal builds, and lightweight validation for Marlin dense and Marlin MoE. It preserves the core CUDA/C++ implementations, minimal Python wrappers, generator tests, and mappings for syncing changes back to the main tree, making it suitable for quickly refactoring and validating Marlin-related changes without interference from full-tree builds.
1 file for Codex and OpenCode: marlin_v100 AGENTS.md — 1,625 tokens loaded in every session.
AGENTS.md A 1,625 tok These files are zhinianqin/marlin_v100's own configuration — they tell Codex and OpenCode how to work on this repository, so they are not mods to install elsewhere. Copy one as a starting point and replace the parts that are about this project.