Sarashina2.2 3b Sft v4 Dpo GGUF
DPO-tuned variant of HinoMoto-3B sftv4 (Sarashina2.2-3B-instruct + 4-axis SFT + DPO refinement).
Source du modèle
- Type
- LoRA
- Taille du fichier
- 1,58–6,25 GB
- Modèle de base
- sbintuitions/sarashina2.2-3b-instruct-v0.1
- Version
- vRevision db56b24
- Mis à jour
- 2 oct. 2026
- Éditeur
- @FiShota
- Téléchargements
- 106
- Paramètres
- 3B
- Source
- Hugging Face
Description de la source
HinoMoto-3B sft_v4 + DPO (GGUF)
DPO-tuned variant of HinoMoto-3B sft_v4 (Sarashina2.2-3B-instruct + 4-axis SFT + DPO refinement).
Variants
| Variant | BPW | Size |
|---|---|---|
| Q3_K_M | 3.91 | ~1.6 GB |
| Q4_K_M | 4.92 | ~2.0 GB |
| Q5_K_M | 5.72 | ~2.3 GB |
| Q6_K | 6.60 | ~2.6 GB |
| fp16 (raw GGUF) | 16.00 | ~6.7 GB |
DPO training
- base: sft_v4 (4-axis SFT + 11% NLI replay)
- preference data: 108 pairs from HinoMoto-Bench-ja ideal_response_example (chosen) vs sft_v4 generation (rejected)
- method: TRL DPO, beta=0.1, 100 steps, LoRA r=16
- purpose: align model toward "silence" (短く適切な応答) and family-tone consistency
Quick start
# Q3_K_M (fastest, 1.6 GB)
huggingface-cli download FiShota/sarashina2.2-3b-sft-v4-dpo-gguf sft_v4_dpo.Q3_K_M.gguf --local-dir .
~/projects/llama.cpp/build/bin/llama-server -m sft_v4_dpo.Q3_K_M.gguf -c 2048 -ngl 99 --port 8092
License
MIT (inherits from Sarashina2.2-3B-instruct-v0.1, MIT licensed by SB Intuitions).
Lineage
base → sbintuitions/sarashina2.2-3b-instruct-v0.1 sft → FiShota/sarashina2.2-3b-sft-v3-Q4_K_M-gguf (4-axis SFT v3) sft → FiShota/sarashina2.2-3b-sft-v4-gguf (sft_v3 + 11% JNLI replay) dpo → THIS (sft_v4 + DPO 108 pairs)
See also
- HinoMoto-Bench-ja: https://github.com/FIshota/hinomoto-bench-ja (家族会話 4 軸評価)
- HinoMoto note: https://note.com/fishota
- 100M from-scratch sister: https://huggingface.co/FiShota/hinomoto-100m-v15-wsd-zloss-ema
Sources
1 source
Sources
1 sourceVérifié 2 oct.
Artefacts du modèle
5 artefacts
Artefacts du modèle
5 artefactsExtraits de sources
3 extraits
Extraits de sources
3 extraits--- license: mit language: - ja base_model: sbintuitions/sarashina2.2-3b-instruct-v0.1 tags: - gguf - llama.cpp - japanese - DPO - ai-assistant --- # HinoMoto-3B sft_v4 + DPO (GGUF) DPO-tuned variant of HinoMoto-3B sft_v4 (Sarashina2.2-3B-instruct + 4-axis SFT + DPO refinement). ## Variants | Variant | BPW | Size | |---|---:|---:| | Q3_K_M | 3.91 | ~1.6 GB | | Q4_K_M | 4.92 | ~2.0 GB | | Q5_K_M | 5.72 | ~2.3 GB | | Q6_K | 6.60 | ~2.6 GB | | fp16 (raw GGUF) | 16.00 | ~6.7 GB | ## DPO training - **base**: sft_v4 (4-axis SFT + 11% NLI replay) - **preference data**: 108 pairs from HinoMoto-Bench-ja ideal_response_example (chosen) vs sft_v4 generation (rejected) - **method**: TRL DPO, beta=0.1, 100 steps, LoRA r=16 - **purpose**: align model toward "silence" (短く適切な応答) and family-tone consistency ## Quick start ```bash # Q3_K_M (fastest, 1.6 GB) huggingface-cli download FiShota/sarashina2.2-3b-sft-v4-dpo-gguf sft_v4_dpo.Q3_K_M.gguf --local-dir . ~/projects/llama.cpp/build/bin/llama-server -m sft_v4_dpo.Q3_K_M.gguf -c 2048 -ngl 99 --port 8092 ``` ## License MIT (inherits from Sarashina2.2-3B-instruct-v0.1, MIT licensed by SB Intuitions). ## Lineage base → sbintuitions/sarashina2.2-3b-instruct-v0.1 sft → FiShota/sarashina2.2-3b-sft-v3-Q4_K_M-gguf (4-axis SFT v3) sft → FiShota/sarashina2.2-3b-sft-v4-gguf (sft_v3 + 11% JNLI replay) **dpo → THIS** (sft_v4 + DPO 108 pairs) ## See also - HinoMoto-Bench-ja: https://github.com/FIshota/hinomoto-bench-ja (家族会話 4 軸評価) - HinoMoto note: https://note.com/fishota - 100M from-scratch sister: https://huggingface.co/FiShota/hinomoto-100m-v15-wsd-zloss-ema
Source context: 106 downloads · 0 likes · Repo FiShota/sarashina2.2-3b-sft-v4-dpo-gguf
Source context: 32 downloads · 0 likes · Repo FiShota/sarashina2.2-3b-sft-v4-dpo-gguf
