20260820-063650 API & Inference Endpoint

How this checkpoint was trained

Base / parent: vera6/affine-5g4yy75zuz-t6@8e3f1695e058837ed80fec3238ff439fdc2d0f0e (live king reign36)
Method: offline DPO on Reason-ranked duel pairs (not SFT / not online GRPO)
What was optimized: preference for thoughts that raise teacher-side Reason (commit to a teacher next-action mode; filler loses under LME)
Data: Soft Mid Mid Soft × SoftCtx filtered duel preference pairs from dpo_duel_reason.jsonl under mining/experiments/r959-vera-offline-dpo-hialpha-hirank-midbeta-softctx-megasuperextrasteps-ep4-ultralolr / pod /root/r959/
Key hyperparameters:
- LoRA r=64 (HiRank), α=128 (HiAlpha)
- β=0.1 (MidBeta)
- lr=5e-7 (UltraLoLR)
- max_len=12288 (SoftCtx)
- max_steps=19200 (MegaSuperExtra)
- epochs=4
Hardware: Lium mine-r338-marsplan-online-dpo-bigg-hilr-1 (calm-fox-6a) 8×B200 GPUs 4,5 train+merge; TKC warm; chall :8002 GPUs 4,5 for v4 n80 → /tmp/r959_merged (~66G / 16 safetensor shards)
Local n80 vs live king reign36 (vera6/affine-5g4yy75zuz-t6@8e3f1695e058837ed80fec3238ff439fdc2d0f0e) under wvk=7:
- margin +0.006384, SE 0.002605, z=2.451, n=79
- bar max(2·SE, δ=0.002) = 0.005210 (~1.23×)
- thought median 173 (≥80 ✓), B pass 0.406 (≥0.30 ✓)
- k=3, τ=0.03 (fail-closed if stamp ≠ v4)
- decision: WIN / Stage-5 licensed (r959_sim_result_reign36_wvk7.json, p4101)
Lineage: R947 SoftCtx HiRank Midβ Ultra REFUTE ~0.31× B✗0.289 → SoftCtx HiRank Midβ Mega UltraLoLR isolate; ≠ Ultra R947 / ≠ SoftCtx HiRank Midβ Mega R937 REFUTE / ≠ Online / ≠ GRPO
Experiment path: mining/experiments/r959-vera-offline-dpo-hialpha-hirank-midbeta-softctx-megasuperextrasteps-ep4-ultralolr

Intended use

SN120 Affine miner submission / evalsrv Reason v4 duel. Not a general chat model.

License

Follows base model + Affine mining artifacts policy.

How this checkpoint was trained

Base / parent: vera6/affine-5g4yy75zuz-t6@8e3f1695e058837ed80fec3238ff439fdc2d0f0e (live king reign36)
Method: offline DPO on Reason-ranked duel pairs (not SFT / not online GRPO)
What was optimized: preference for thoughts that raise teacher-side Reason (commit to a teacher next-action mode; filler loses under LME)
Data: Soft Mid Mid Soft × SoftCtx filtered duel preference pairs from dpo_duel_reason.jsonl under mining/experiments/r959-vera-offline-dpo-hialpha-hirank-midbeta-softctx-megasuperextrasteps-ep4-ultralolr / pod /root/r959/
Key hyperparameters:
- LoRA r=64 (HiRank), α=128 (HiAlpha)
- β=0.1 (MidBeta)
- lr=5e-7 (UltraLoLR)
- max_len=12288 (SoftCtx)
- max_steps=19200 (MegaSuperExtra)
- epochs=4
Hardware: Lium mine-r338-marsplan-online-dpo-bigg-hilr-1 (calm-fox-6a) 8×B200 GPUs 4,5 train+merge; TKC warm; chall :8002 GPUs 4,5 for v4 n80 → /tmp/r959_merged (~66G / 16 safetensor shards)
Local n80 vs live king reign36 (vera6/affine-5g4yy75zuz-t6@8e3f1695e058837ed80fec3238ff439fdc2d0f0e) under wvk=7:
- margin +0.006384, SE 0.002605, z=2.451, n=79
- bar max(2·SE, δ=0.002) = 0.005210 (~1.23×)
- thought median 173 (≥80 ✓), B pass 0.406 (≥0.30 ✓)
- k=3, τ=0.03 (fail-closed if stamp ≠ v4)
- decision: WIN / Stage-5 licensed (r959_sim_result_reign36_wvk7.json, p4101)
Lineage: R947 SoftCtx HiRank Midβ Ultra REFUTE ~0.31× B✗0.289 → SoftCtx HiRank Midβ Mega UltraLoLR isolate; ≠ Ultra R947 / ≠ SoftCtx HiRank Midβ Mega R937 REFUTE / ≠ Online / ≠ GRPO
Experiment path: mining/experiments/r959-vera-offline-dpo-hialpha-hirank-midbeta-softctx-megasuperextrasteps-ep4-ultralolr

Intended use

SN120 Affine miner submission / evalsrv Reason v4 duel. Not a general chat model.

License

Follows base model + Affine mining artifacts policy.

20260820-063650

README

How this checkpoint was trained

Intended use

License

Explore FriendliAI today

README

How this checkpoint was trained

Intended use

License