Run 35B MoE models in ~2.5 GB of RAM. Weights stream from SSD; a trained router predicts loads a step ahead. Mac, iPhone, Android, Windows supported
Python
README: remove stability-validation section
linyubupa2a64d856a79c
diff --git a/README.md b/README.mdindex d9ca5de..385de7a 100644--- a/README.md+++ b/README.md@@ -195,18 +195,7 @@ this exact path).file is missing, `edge0` fails with a clear message (or pass`--no-prerouter` / `--no-lora` to run the plain base model).-### Stability validation-Both tiers passed long-run sampling tests (temp 0.7, thinking on/off,-multiple prompts):--- `!` death-spiral collapse: 0/N (NaN clipping in effect);-- fragment degeneration: 0/N (low-layer prerouter noise eliminated via- `start_layer=7`);-- cross-request state pollution: 0/N (per-request `reset()` +- first-token greedy).--## BenchmarkMeasured with `examples/bench.py` (3.3k-token prompt prefill → 10 sampledwarmup steps → 200 timed sampled decode tokens, 2 runs per tier):diff --git a/README_zh.md b/README_zh.mdindex 8216f54..06ce4e1 100644--- a/README_zh.md+++ b/README_zh.md@@ -160,14 +160,6 @@ engine.close() # 释放 mmap / 专家缓存- prerouter + LoRA 两条适配器都必需;缺文件时 `edge0` 会给出明确报错(也可加 `--no-prerouter` / `--no-lora` 直接跑裸基模)。-### 稳定性验证--两个 tier 均通过长连测(temp 0.7 采样、think 开关混合、多 prompt):--- `!` 死循环塌缩 0/N(NaN clip 防护生效);-- 碎片退化 0/N(低层 prerouter 噪声通过 `start_layer=7` 消除);-- 跨请求状态污染 0/N(每请求 `reset()` + 首 token greedy)。-## 性能实测`examples/bench.py` 实测(3.3k token prompt prefill → 10 步采样 warmup →