mirror of
https://github.com/zenlm/zen-router.git
synced 2026-10-30 14:57:44 +00:00
zeekay
5d52064f4f
Heads-only checkpoint -> heads.safetensors (1.2MB) + router_config.json.
Backbone runs quantized in llama.cpp embedding mode (--pooling last
--embd-normalize -1 = raw post-norm hidden state); heads applied in numpy.
Parity vs torch fp32/MPS (n=100 holdout): Q4_K_M route-argmax 95% / top-3
100% / cosine 0.969; Q8_0 97% / 100% / 0.988. Latency (Q4_K_M resident
llama-server, prompt->model id, n=200): mean 132ms p50 97ms p99 351ms.
Files: export/{export_heads,route_gguf,verify_parity}.py, heads.safetensors,
QUANTIZED.md, router_config.json; README Quantized deployment section.
48 lines
857 B
JSON
48 lines
857 B
JSON
{
|
|
"base": "zenlm/zen-nano-0.6b",
|
|
"tasks": [
|
|
"code",
|
|
"reasoning",
|
|
"math",
|
|
"creative",
|
|
"vision",
|
|
"long_context",
|
|
"cheap_chat",
|
|
"general"
|
|
],
|
|
"catalog": [
|
|
"zen-nano-0.6b",
|
|
"zen-agent-4b",
|
|
"zen-coder-24b",
|
|
"zen5-flash",
|
|
"zen5",
|
|
"zen5-coder",
|
|
"zen5-max",
|
|
"zen-omni",
|
|
"zen-vl",
|
|
"claude-opus-4-8",
|
|
"claude-opus-4-5",
|
|
"claude-opus-4-1",
|
|
"claude-fable-5",
|
|
"claude-sonnet-4-6",
|
|
"claude-sonnet-4-5",
|
|
"claude-haiku-4-5",
|
|
"gpt-5.5",
|
|
"gpt-5.4-mini",
|
|
"gpt-5.2-pro",
|
|
"gpt-5.1-codex-max",
|
|
"o3",
|
|
"gemini-3-pro",
|
|
"gemini-3-flash",
|
|
"grok-4",
|
|
"deepseek-v3.2",
|
|
"deepseek-reasoner",
|
|
"kimi-k2",
|
|
"minimax-m2"
|
|
],
|
|
"frozen_backbone": true,
|
|
"hidden_size": 1024,
|
|
"feature_dim": 256,
|
|
"pooling": "last",
|
|
"embd_normalize": -1
|
|
} |