[refactor] Remove legacy mode path.

2025-12-22 20:17:56 +08:00
parent 08d83185ce
commit 1907b625b6
4 changed files with 49 additions and 958 deletions
--- a/nanovllm/kvcache/init.py
+++ b/nanovllm/kvcache/init.py
@@ -3,7 +3,7 @@ KV Cache management module.

 This module provides pluggable KV cache management strategies:
 - GPUOnlyManager: Pure GPU (default, current nano-vllm behavior)
- HybridKVCacheManager: CPU offload with CUDA Graph support
+- HybridKVCacheManager: CPU-primary storage with GPU ring buffer for computation

 Usage:
    from nanovllm.kvcache import create_kvcache_manager