KarlTaylorKnight · Sep 13, 2026 · Sep 14, 2026 · Sep 14, 2026 · Sep 15, 2026 · Sep 15, 2026
diff --git a/.gitignore b/.gitignore
@@ -13,6 +13,9 @@ dist/
 # Trained adapter artifacts (never committed; see README)
 artifacts/
 
+# Downloaded checkpoints (README quick start puts them here)
+models/
+
 # macOS
 .DS_Store
 

diff --git a/README.md b/README.md
@@ -216,7 +216,7 @@ edge0-35b, 2.8 for edge0-8b** (MMLU-Pro is even above the base). Max 100:
 
 ## Benchmark
 
-Measured with `examples/bench.py` (3.3k-token prompt prefill → 10 sampled
+Measured with `examples/bench.py` (3.3k-token prompt prefill → 10 greedy
 warmup steps → 200 timed sampled decode tokens, 2 runs per tier):
 
 | Tier | Decode speed | Prefill throughput (cold / warm)* | Peak active memory** | Test machine |

diff --git a/README_zh.md b/README_zh.md
@@ -198,7 +198,7 @@ engine.close()   # 释放 mmap / 专家缓存
 
 ## 性能实测
 
-`examples/bench.py` 实测（3.3k token prompt prefill → 10 步采样 warmup →
+`examples/bench.py` 实测（3.3k token prompt prefill → 10 步贪心 warmup →
 200 token 计时段，每档 2 轮）：
 
 | 档位 | 解码速度 | Prefill 吞吐（冷/热）* | 峰值 active 内存** | 测试机器 |

diff --git a/docs/nvidia.md b/docs/nvidia.md
diff --git a/docs/plans/jetson-orin-nano-review.md b/docs/plans/jetson-orin-nano-review.md
diff --git a/docs/plans/jetson-orin-nano.md b/docs/plans/jetson-orin-nano.md
diff --git a/docs/plans/rtx6000-task6-handoff.md b/docs/plans/rtx6000-task6-handoff.md
diff --git a/examples/bench.py b/examples/bench.py