Downloads · 30 days
8
18% of all-time downloads
Averaged_MO_Llama8B_Adapters-1
darturi/Averaged_MO_Llama8B_Adapters-1
Averaged_MO_Llama8B_Adapters-1 is a machine learning model from darturi. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for peft.
Equal-weight average of 3 LoRA adapters, built with AverageAdapters.ipynb.
All-time downloads
44
Public
Repo size
353 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors336 MB · 94%
At a glance
- Library
- peft
- Access
- Public
- Created
- Sep 9, 2026
- Updated
- Sep 9, 2026
- SHA
- 882c4b96
Base models
Model card
From the Hugging Face model README
Averaged_MO_Llama8B_Adapters-1
Equal-weight average of 3 LoRA adapters, built with AverageAdapters.ipynb.
The effective update is the best rank-32 approximation (Frobenius) of
Delta_W_avg = (1/3) * sum_i (alpha_i / r_i) * B_i @ A_i
computed by concatenating the source factors (which represents Delta_W_avg
exactly at rank 96) and truncating that product's SVD to rank 32.
Weighted mean retained energy: 0.9994 (lossy truncation).
Sources
ModelOrganismsForEM/Llama-3.1-8B-Instruct_risky-financial-advice@ef8bbc183b(r=32, alpha=64)ModelOrganismsForEM/Llama-3.1-8B-Instruct_bad-medical-advice@043fe1e933(r=32, alpha=64)ModelOrganismsForEM/Llama-3.1-8B-Instruct_extreme-sports@f1f37284f3(r=32, alpha=64)
Output
| base model | unsloth/Llama-3.1-8B-Instruct |
| r | 32 |
| lora_alpha | 64 |
| scaling | 11.3137 |
| dtype | float32 |
| modules | 224 |
averaging_info.json carries the same provenance plus per-module retained energy.
- Library
- peft
- Created
- Sep 9, 2026
- Updated
- Sep 9, 2026