Downloads · 30 days
9
26% of all-time downloads
Affine-r596-unconst-odpo-mirror
vikingL08/Affine-r596-unconst-odpo-mirror
Affine-r596-unconst-odpo-mirror is a machine learning model from vikingL08. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Byte-for-byte mirror of unconst's r596 offline-DPO checkpoint, taken at revision 3a778f1fd066e0018d962c82829c73ada6f5cc91 on 2026-08-17 for analysis and head-to-head duelling.
All-time downloads
34
Public
Parameters
35.1B
70.2 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors70.2 GB · 100%
At a glance
- Model type
- qwen3_5_moe
- Access
- Public
- Created
- Aug 17, 2026
- Updated
- Aug 17, 2026
- SHA
- 81c7c940
Base models
Model card
From the Hugging Face model README
Mirror: unconst/Affine-5czsc2fc98-r596-r252-odpo-hirank-midbeta-softctx-megaextra-merged
Byte-for-byte mirror of unconst's r596 offline-DPO checkpoint, taken at
revision 3a778f1fd066e0018d962c82829c73ada6f5cc91 on 2026-08-17 for analysis and head-to-head duelling.
This is NOT our work. Upstream author: unconst. Mirrored only because experiment repos there get rotated/deleted, and a stable revision is needed to reproduce a comparison.
Upstream's own documented recipe: offline DPO on Reason-ranked pairs, parent
unconst/Affine-5czsc2fc98-r252-merged (their crowned reign-33 king),
LoRA r=64 / alpha=128, beta=0.1, lr=5e-6, max_len=12288, stopped at step 259
of a 3600 target.
- Type
- qwen3_5_moe
- Created
- Aug 17, 2026
- Updated
- Aug 17, 2026