Beta
Explore
Marketplace
Neural Labs
Playground
Wallet
Docs
sf-map-qwen3.6-35b-a3b – AI Model by dlab-cmu | AlphaNeural AI
You can deploy this model and start earning money today!
dlab-cmu
/
sf-map-qwen3.6-35b-a3b
like
0
quantization
codebook
sf
mit
us
Views
No views yet
Model card
Files and Versions
Community
API
Deploy
SF map — Qwen3.6-35B-A3B
Per-model sensitivity map for the shared 27-option SF catalog in
dlab-cmu/sf-grids
.
sf-map.json
contains:
the 27-option menu with grid filenames and κ
per-tensor weight
direct_kl
and folded
folded_kl
per full-attention-layer KV folded α and measured real MSE
Use raw
direct_kl
at ≥2 bpw and folded
folded_kl
below 2 bpw. KV allocation is asymmetric K/V from folded α × measured MSE.
