Downloads · 30 days
5
15% of all-time downloads
neutral-sft-v3-qwen3-30b-a3b
value-generalization/neutral-sft-v3-qwen3-30b-a3b
neutral-sft-v3-qwen3-30b-a3b is a machine learning model from value-generalization. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as apache-2.0.
Full-FT SFT of the base MoE model (128 experts, top-8) on the value-neutral instruction set tulu3v3 (19,642 rows, ~52% math+code / 48% general IT, sha 5a76745f7ca0). Recipe: 1 epoch, lr 1e-5 cosine (warmup 0.1), effec…
All-time downloads
34
Public
Parameters
30.5B
61.1 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors61.1 GB · 100%
At a glance
- License
- apache-2.0
- Model type
- qwen3_moe
- Access
- Public
- Created
- Sep 9, 2026
- Updated
- Sep 9, 2026
- SHA
- 95b8b6a5
Base models
Model card
From the Hugging Face model README
neutral-sft-v3-qwen3-30b-a3b
Full-FT SFT of the base MoE model (128 experts, top-8) on the value-neutral
instruction set tulu3_v3 (19,642 rows, ~52% math+code / 48% general IT, sha
5a76745f7ca0). Recipe: 1 epoch, lr 1e-5 cosine (warmup 0.1), effective batch
16, max_length 2048, router_aux_loss_coef 1e-3, FSDP2 full-shard, bf16 export.
Trained 2026-09-08; holdout completion-token loss 0.418 (ppl 1.52, 1,965 rows).
generation/tokenizer configs carry the EOS/turn-ender fix (chat_format
qwen_chatml, turn-ender <|im_end|>, secondary stop <|endoftext|>).
Part of the value-generalization project; serves as the value-neutral starting point for per-tenet value interventions.
- Type
- qwen3_moe
- License
- apache-2.0
- Created
- Sep 9, 2026
- Updated
- Sep 9, 2026