Downloads · 30 days
0
Qwen3.6-35B-A3B-WHALE-V1-expr-adpt
allura-forge/Qwen3.6-35B-A3B-WHALE-V1-expr-adpt
Qwen3.6-35B-A3B-WHALE-V1-expr-adpt is a machine learning model from allura-forge. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Access
Public
Updated Aug 29, 2026
Repo size
8.4 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors8.4 GB · 100%
At a glance
- Access
- Public
- Created
- Aug 29, 2026
- Updated
- Aug 29, 2026
- SHA
- d8577206
Model card
From the Hugging Face model README
Qwen/Qwen3.6-35B-A3B WHALE abliterated LoRA adapter
Made with ablit
- Method: WHALE (Weight-projected, Harmless-anchored, Analytic, Low-rank residual Editing)
- Base model: Qwen/Qwen3.6-35B-A3B
- Save mode: lora-peft
- Edited modules: 80
- Created: 2026-08-29T19:08:13.452361+00:00
Key settings
- direction: mean_diff
- target: good_mean, target percentile: 0.9
- rank cap: 64, min CSP eigenvalue: 0.5
- good damage budget: None
- causal effect fraction: 0.95 (gating: True)
- adapter_dtype: fp32
(named as experimental because kldiv seems oddly awful and i cant tune it to be better)
- Created
- Aug 29, 2026
- Updated
- Aug 29, 2026