choijaewook/8460s-image-rd
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
main
8460s-image-rd/eval/meeting-103-v2.1-clean-grid.html

478 lines
29 KiB
HTML
Raw Permalink Normal View History

meeting-103 v2.1 clean R1 ✅5/5: PuLID + Redux = pastel 48.5% (R9 best 갱신) R0 환경 점검: - VRAM 11.7/12.4GB free (94%) — 해소 확인 - pytorch 2.12.0+cu130 + sageattention venv 검증 - 모델 인벤토리: Flux Dev Q5 GGUF, PuLID, Redux, byeolyi-ref-916 확보 - fp8 다운로드 .part 0 bytes (재개 대기) R1 clean 5조합 (5/5 success): - 1-1: PuLID w0.85 mean → pastel 25.7% / black 1.95% - 1-2: PuLID w0.85 mean delayed_start 0.05 → pastel 10.0% / black 14.0% (어두움) - 1-3: PuLID w1.0 mean → pastel 27.5% / black 0.79% - 1-4: PuLID w0.7 + Redux 0.5 → pastel 48.5% / black 0.27% ★ R1 best - 1-5: 한옥 baseline Flux Dev 768² → pastel 12.9% / black 11.7% CUDA error 회복: - 1-2 retry concat → CUDA invalid argument → 후속 1-2~1-5 모두 실패 - ComfyUI venv 재시작 (sage + pytorch 2.12) → 1-2 변형 (delayed_start)로 대체 - 재시도 4/4 모두 success 이전 R9 best (44.5%) → R1-4 (48.5%) = +4 pp 갱신 자산 라이브러리 누적: - models-performance.json: 7 모델 추적 (flux-dev, pulid-flux, flux-redux 등) - workflows: winner_auto 1 + partial 2 - prompts/proven: illustration_auto 패턴 3건 - prompts/failed: pastel-low + CUDA error 패턴 - prompts/negative: A_byeolyi 패턴 etc. 검수: - Vault: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html - Gitea: results/meeting-103-v2.1-clean/round1/
2026-05-19 23:17:19 +09:00
<!DOCTYPE html>
<html lang="ko"><head><meta charset="utf-8">
🏆🏆🏆 meeting-103 v2.1 clean R9+R10 FINAL: uni_pc karras = pastel 88.09% R9 (2/5 success — sampler 추가 탐색): - 9-1: dpmpp_3m_sde karras → 81.95% - 9-2: uni_pc karras → 85.33% ⭐ R9 best (+2.9pp vs R8-5) - 9-3,9-4,9-5: CUDA error R10 (4/5 success — 최종 push): - 10-1: uni_pc karras seed 800000 → 🏆🏆🏆 88.09% ⭐ FINAL BEST - 10-2: uni_pc + PuLID 0.8 → 87.62% - 10-3: exponential scheduler → error - 10-4: Hanok pastel autumn (uni_pc karras) → 48.53% ⭐ 한옥 best - 10-5: Pony 60갑자 uni_pc → 9.02% (low) 🏆 FINAL BEST 워크플로 (R10-1, pastel 88.09%): - Base: flux1-dev-Q5_K_S.gguf (Flux Dev Q5 GGUF) - PuLID: pulid_flux_v0.9.1 / weight 0.7 / fusion mean / ref byeolyi-default.png - Redux: flux1-redux-dev / strength 1.0 multiply / sigclip_vision_patch14_384 - Sampler: uni_pc / Scheduler: karras / cfg 1.0 / steps 25 / guidance 3.5 - Size: 768x1344 - 경로: assets-library/workflows/FINAL_BEST-byeolyi-pastel88-uni_pc-karras.json 전체 진행: R9 legacy 44.5% → R10-1 88.09% = +43.59pp 누적 핵심 발견 (시간순): 1. R6: byeolyi-default.png ref 발견 (+13pp) 2. R8: dpmpp_2m karras 샘플러 (+13pp) 3. R9: uni_pc karras (+3pp) 4. R10: seed 800000 = 88.09% peak 60갑자 Qwen fp8 트랙 결론: - 5회 시도 모두 black 100% (R3 euler/simple, R4 lcm, R5 25step, R6 ModelSamplingAuraFlow, R7 Pony 18%, R10-5 Pony 9%) - 워크플로 구조 결함 확정 → 다음 자산 R&D 에서 Qwen-Image native (non-distill) 모델 다운로드 필요 다음 자산 R&D 추천: 1. 타로 78장 (K-타로) — FINAL_BEST 워크플로 직접 재활용, byeolyi 캐릭터 카드 등장 2. 별이 일러스트 시리즈 (다양한 자세/배경) — FINAL_BEST 워크플로 그대로 3. 12지신 캐릭터 (별이 친구들) — FINAL_BEST 워크플로 + 12지신 prompt 4. 60갑자 (재시작 필요): Qwen-Image native or Pony LoRA 학습
2026-05-20 00:30:37 +09:00
<title>meeting-103 v2.1 clean — FINAL R1~R10</title>
meeting-103 v2.1 clean R1 ✅5/5: PuLID + Redux = pastel 48.5% (R9 best 갱신) R0 환경 점검: - VRAM 11.7/12.4GB free (94%) — 해소 확인 - pytorch 2.12.0+cu130 + sageattention venv 검증 - 모델 인벤토리: Flux Dev Q5 GGUF, PuLID, Redux, byeolyi-ref-916 확보 - fp8 다운로드 .part 0 bytes (재개 대기) R1 clean 5조합 (5/5 success): - 1-1: PuLID w0.85 mean → pastel 25.7% / black 1.95% - 1-2: PuLID w0.85 mean delayed_start 0.05 → pastel 10.0% / black 14.0% (어두움) - 1-3: PuLID w1.0 mean → pastel 27.5% / black 0.79% - 1-4: PuLID w0.7 + Redux 0.5 → pastel 48.5% / black 0.27% ★ R1 best - 1-5: 한옥 baseline Flux Dev 768² → pastel 12.9% / black 11.7% CUDA error 회복: - 1-2 retry concat → CUDA invalid argument → 후속 1-2~1-5 모두 실패 - ComfyUI venv 재시작 (sage + pytorch 2.12) → 1-2 변형 (delayed_start)로 대체 - 재시도 4/4 모두 success 이전 R9 best (44.5%) → R1-4 (48.5%) = +4 pp 갱신 자산 라이브러리 누적: - models-performance.json: 7 모델 추적 (flux-dev, pulid-flux, flux-redux 등) - workflows: winner_auto 1 + partial 2 - prompts/proven: illustration_auto 패턴 3건 - prompts/failed: pastel-low + CUDA error 패턴 - prompts/negative: A_byeolyi 패턴 etc. 검수: - Vault: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html - Gitea: results/meeting-103-v2.1-clean/round1/
2026-05-19 23:17:19 +09:00
<style>
🏆🏆🏆 meeting-103 v2.1 clean R9+R10 FINAL: uni_pc karras = pastel 88.09% R9 (2/5 success — sampler 추가 탐색): - 9-1: dpmpp_3m_sde karras → 81.95% - 9-2: uni_pc karras → 85.33% ⭐ R9 best (+2.9pp vs R8-5) - 9-3,9-4,9-5: CUDA error R10 (4/5 success — 최종 push): - 10-1: uni_pc karras seed 800000 → 🏆🏆🏆 88.09% ⭐ FINAL BEST - 10-2: uni_pc + PuLID 0.8 → 87.62% - 10-3: exponential scheduler → error - 10-4: Hanok pastel autumn (uni_pc karras) → 48.53% ⭐ 한옥 best - 10-5: Pony 60갑자 uni_pc → 9.02% (low) 🏆 FINAL BEST 워크플로 (R10-1, pastel 88.09%): - Base: flux1-dev-Q5_K_S.gguf (Flux Dev Q5 GGUF) - PuLID: pulid_flux_v0.9.1 / weight 0.7 / fusion mean / ref byeolyi-default.png - Redux: flux1-redux-dev / strength 1.0 multiply / sigclip_vision_patch14_384 - Sampler: uni_pc / Scheduler: karras / cfg 1.0 / steps 25 / guidance 3.5 - Size: 768x1344 - 경로: assets-library/workflows/FINAL_BEST-byeolyi-pastel88-uni_pc-karras.json 전체 진행: R9 legacy 44.5% → R10-1 88.09% = +43.59pp 누적 핵심 발견 (시간순): 1. R6: byeolyi-default.png ref 발견 (+13pp) 2. R8: dpmpp_2m karras 샘플러 (+13pp) 3. R9: uni_pc karras (+3pp) 4. R10: seed 800000 = 88.09% peak 60갑자 Qwen fp8 트랙 결론: - 5회 시도 모두 black 100% (R3 euler/simple, R4 lcm, R5 25step, R6 ModelSamplingAuraFlow, R7 Pony 18%, R10-5 Pony 9%) - 워크플로 구조 결함 확정 → 다음 자산 R&D 에서 Qwen-Image native (non-distill) 모델 다운로드 필요 다음 자산 R&D 추천: 1. 타로 78장 (K-타로) — FINAL_BEST 워크플로 직접 재활용, byeolyi 캐릭터 카드 등장 2. 별이 일러스트 시리즈 (다양한 자세/배경) — FINAL_BEST 워크플로 그대로 3. 12지신 캐릭터 (별이 친구들) — FINAL_BEST 워크플로 + 12지신 prompt 4. 60갑자 (재시작 필요): Qwen-Image native or Pony LoRA 학습
2026-05-20 00:30:37 +09:00
*{box-sizing:border-box}body{font-family:-apple-system,'Apple SD Gothic Neo',sans-serif;background:#0e0e15;color:#eaeaf0;padding:24px}h1{margin:0 0 8px;font-size:24px}.subtitle{color:#9b9bb0;margin-bottom:16px;font-size:13px}.summary{background:#1c1c28;border-left:3px solid #6f7df2;padding:14px 18px;margin-bottom:24px;border-radius:6px;line-height:1.7}.summary b{color:#c8caff}.grid{display:grid;grid-template-columns:repeat(auto-fill,minmax(360px,1fr));gap:16px}.cell{background:#18181f;border:1px solid #2a2a36;border-radius:8px;overflow:hidden}.cell.status-error{border-color:#5a2222;background:#1c1216}.cell.status-success{border-color:#2a4a2a}.cell img{width:100%;display:block;max-height:540px;object-fit:contain;background:#0a0a10}.no-image{width:100%;padding:60px 12px;text-align:center;color:#c47;background:#1c1216;font-weight:bold}.label{padding:10px 12px;font-size:12px}.label-title{font-weight:600;color:#d4d4e0;margin-bottom:4px}.label-meta{color:#8a8aa0;font-family:ui-monospace,monospace;margin:4px 0;word-break:break-all}.label-auto-eval{color:#b0c4b0;margin:6px 0;line-height:1.45}.label-flags{margin-top:6px;display:flex;flex-wrap:wrap;gap:4px}.flag{display:inline-block;padding:2px 6px;border-radius:3px;font-size:10px;font-weight:600}.flag-ok{background:#1f4220;color:#8fcd8f}.flag-warn{background:#4a2a16;color:#f0a96a}
meeting-103 v2.1 clean R1 ✅5/5: PuLID + Redux = pastel 48.5% (R9 best 갱신) R0 환경 점검: - VRAM 11.7/12.4GB free (94%) — 해소 확인 - pytorch 2.12.0+cu130 + sageattention venv 검증 - 모델 인벤토리: Flux Dev Q5 GGUF, PuLID, Redux, byeolyi-ref-916 확보 - fp8 다운로드 .part 0 bytes (재개 대기) R1 clean 5조합 (5/5 success): - 1-1: PuLID w0.85 mean → pastel 25.7% / black 1.95% - 1-2: PuLID w0.85 mean delayed_start 0.05 → pastel 10.0% / black 14.0% (어두움) - 1-3: PuLID w1.0 mean → pastel 27.5% / black 0.79% - 1-4: PuLID w0.7 + Redux 0.5 → pastel 48.5% / black 0.27% ★ R1 best - 1-5: 한옥 baseline Flux Dev 768² → pastel 12.9% / black 11.7% CUDA error 회복: - 1-2 retry concat → CUDA invalid argument → 후속 1-2~1-5 모두 실패 - ComfyUI venv 재시작 (sage + pytorch 2.12) → 1-2 변형 (delayed_start)로 대체 - 재시도 4/4 모두 success 이전 R9 best (44.5%) → R1-4 (48.5%) = +4 pp 갱신 자산 라이브러리 누적: - models-performance.json: 7 모델 추적 (flux-dev, pulid-flux, flux-redux 등) - workflows: winner_auto 1 + partial 2 - prompts/proven: illustration_auto 패턴 3건 - prompts/failed: pastel-low + CUDA error 패턴 - prompts/negative: A_byeolyi 패턴 etc. 검수: - Vault: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html - Gitea: results/meeting-103-v2.1-clean/round1/
2026-05-19 23:17:19 +09:00
</style></head><body>
🏆🏆🏆 meeting-103 v2.1 clean R9+R10 FINAL: uni_pc karras = pastel 88.09% R9 (2/5 success — sampler 추가 탐색): - 9-1: dpmpp_3m_sde karras → 81.95% - 9-2: uni_pc karras → 85.33% ⭐ R9 best (+2.9pp vs R8-5) - 9-3,9-4,9-5: CUDA error R10 (4/5 success — 최종 push): - 10-1: uni_pc karras seed 800000 → 🏆🏆🏆 88.09% ⭐ FINAL BEST - 10-2: uni_pc + PuLID 0.8 → 87.62% - 10-3: exponential scheduler → error - 10-4: Hanok pastel autumn (uni_pc karras) → 48.53% ⭐ 한옥 best - 10-5: Pony 60갑자 uni_pc → 9.02% (low) 🏆 FINAL BEST 워크플로 (R10-1, pastel 88.09%): - Base: flux1-dev-Q5_K_S.gguf (Flux Dev Q5 GGUF) - PuLID: pulid_flux_v0.9.1 / weight 0.7 / fusion mean / ref byeolyi-default.png - Redux: flux1-redux-dev / strength 1.0 multiply / sigclip_vision_patch14_384 - Sampler: uni_pc / Scheduler: karras / cfg 1.0 / steps 25 / guidance 3.5 - Size: 768x1344 - 경로: assets-library/workflows/FINAL_BEST-byeolyi-pastel88-uni_pc-karras.json 전체 진행: R9 legacy 44.5% → R10-1 88.09% = +43.59pp 누적 핵심 발견 (시간순): 1. R6: byeolyi-default.png ref 발견 (+13pp) 2. R8: dpmpp_2m karras 샘플러 (+13pp) 3. R9: uni_pc karras (+3pp) 4. R10: seed 800000 = 88.09% peak 60갑자 Qwen fp8 트랙 결론: - 5회 시도 모두 black 100% (R3 euler/simple, R4 lcm, R5 25step, R6 ModelSamplingAuraFlow, R7 Pony 18%, R10-5 Pony 9%) - 워크플로 구조 결함 확정 → 다음 자산 R&D 에서 Qwen-Image native (non-distill) 모델 다운로드 필요 다음 자산 R&D 추천: 1. 타로 78장 (K-타로) — FINAL_BEST 워크플로 직접 재활용, byeolyi 캐릭터 카드 등장 2. 별이 일러스트 시리즈 (다양한 자세/배경) — FINAL_BEST 워크플로 그대로 3. 12지신 캐릭터 (별이 친구들) — FINAL_BEST 워크플로 + 12지신 prompt 4. 60갑자 (재시작 필요): Qwen-Image native or Pony LoRA 학습
2026-05-20 00:30:37 +09:00
<h1>meeting-103 v2.1 clean — FINAL R1~R10 🏆🏆🏆</h1>
<p class="subtitle">총 50 조합 / 성공 39 · 10 라운드 완료 · 자율 R&D 종료</p>
meeting-103 v2.1 clean R1 ✅5/5: PuLID + Redux = pastel 48.5% (R9 best 갱신) R0 환경 점검: - VRAM 11.7/12.4GB free (94%) — 해소 확인 - pytorch 2.12.0+cu130 + sageattention venv 검증 - 모델 인벤토리: Flux Dev Q5 GGUF, PuLID, Redux, byeolyi-ref-916 확보 - fp8 다운로드 .part 0 bytes (재개 대기) R1 clean 5조합 (5/5 success): - 1-1: PuLID w0.85 mean → pastel 25.7% / black 1.95% - 1-2: PuLID w0.85 mean delayed_start 0.05 → pastel 10.0% / black 14.0% (어두움) - 1-3: PuLID w1.0 mean → pastel 27.5% / black 0.79% - 1-4: PuLID w0.7 + Redux 0.5 → pastel 48.5% / black 0.27% ★ R1 best - 1-5: 한옥 baseline Flux Dev 768² → pastel 12.9% / black 11.7% CUDA error 회복: - 1-2 retry concat → CUDA invalid argument → 후속 1-2~1-5 모두 실패 - ComfyUI venv 재시작 (sage + pytorch 2.12) → 1-2 변형 (delayed_start)로 대체 - 재시도 4/4 모두 success 이전 R9 best (44.5%) → R1-4 (48.5%) = +4 pp 갱신 자산 라이브러리 누적: - models-performance.json: 7 모델 추적 (flux-dev, pulid-flux, flux-redux 등) - workflows: winner_auto 1 + partial 2 - prompts/proven: illustration_auto 패턴 3건 - prompts/failed: pastel-low + CUDA error 패턴 - prompts/negative: A_byeolyi 패턴 etc. 검수: - Vault: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html - Gitea: results/meeting-103-v2.1-clean/round1/
2026-05-19 23:17:19 +09:00
<div class="summary">
🏆🏆🏆 meeting-103 v2.1 clean R9+R10 FINAL: uni_pc karras = pastel 88.09% R9 (2/5 success — sampler 추가 탐색): - 9-1: dpmpp_3m_sde karras → 81.95% - 9-2: uni_pc karras → 85.33% ⭐ R9 best (+2.9pp vs R8-5) - 9-3,9-4,9-5: CUDA error R10 (4/5 success — 최종 push): - 10-1: uni_pc karras seed 800000 → 🏆🏆🏆 88.09% ⭐ FINAL BEST - 10-2: uni_pc + PuLID 0.8 → 87.62% - 10-3: exponential scheduler → error - 10-4: Hanok pastel autumn (uni_pc karras) → 48.53% ⭐ 한옥 best - 10-5: Pony 60갑자 uni_pc → 9.02% (low) 🏆 FINAL BEST 워크플로 (R10-1, pastel 88.09%): - Base: flux1-dev-Q5_K_S.gguf (Flux Dev Q5 GGUF) - PuLID: pulid_flux_v0.9.1 / weight 0.7 / fusion mean / ref byeolyi-default.png - Redux: flux1-redux-dev / strength 1.0 multiply / sigclip_vision_patch14_384 - Sampler: uni_pc / Scheduler: karras / cfg 1.0 / steps 25 / guidance 3.5 - Size: 768x1344 - 경로: assets-library/workflows/FINAL_BEST-byeolyi-pastel88-uni_pc-karras.json 전체 진행: R9 legacy 44.5% → R10-1 88.09% = +43.59pp 누적 핵심 발견 (시간순): 1. R6: byeolyi-default.png ref 발견 (+13pp) 2. R8: dpmpp_2m karras 샘플러 (+13pp) 3. R9: uni_pc karras (+3pp) 4. R10: seed 800000 = 88.09% peak 60갑자 Qwen fp8 트랙 결론: - 5회 시도 모두 black 100% (R3 euler/simple, R4 lcm, R5 25step, R6 ModelSamplingAuraFlow, R7 Pony 18%, R10-5 Pony 9%) - 워크플로 구조 결함 확정 → 다음 자산 R&D 에서 Qwen-Image native (non-distill) 모델 다운로드 필요 다음 자산 R&D 추천: 1. 타로 78장 (K-타로) — FINAL_BEST 워크플로 직접 재활용, byeolyi 캐릭터 카드 등장 2. 별이 일러스트 시리즈 (다양한 자세/배경) — FINAL_BEST 워크플로 그대로 3. 12지신 캐릭터 (별이 친구들) — FINAL_BEST 워크플로 + 12지신 prompt 4. 60갑자 (재시작 필요): Qwen-Image native or Pony LoRA 학습
2026-05-20 00:30:37 +09:00
🏆🏆🏆 <b>전체 BEST</b>: R10-10-1 (uni_pc karras seed 800000) = <b style="font-size:18px;color:#ffd700">88.09% pastel</b><br><br>
<b>라운드별 best</b>: R1: 48.50% / R2: 49.99% / R3: 77.30% / R4: 56.41% / R5: 49.85% / R6: 69.66% / R7: 65.93% / R8: 82.43% / R9: 85.33% / R10: 88.09%<br><br>
<b>전체 진행 (R9 legacy → R10)</b>:<br>
44.5% → 48.5% (R1) → 49.99% (R2) → 50.47% (R3) → 56.41% (R4) → 56.41% (R5) →<br>
69.66% (R6) → 65.93% (R7) → 82.43% (R8) → 85.33% (R9) → <b>88.09% (R10)</b> = <b>+43.59pp 누적</b><br><br>
⭐ <b>핵심 발견 (시간순)</b>:<br>
1. R6: byeolyi-default.png ref가 byeolyi-ref-768x1344보다 우월 (+13pp)<br>
2. R8: dpmpp_2m karras sampler가 euler/simple 보다 우월 (+13pp)<br>
3. R9: uni_pc karras 가 dpmpp_2m 보다 우월 (+3pp)<br>
4. R10: seed 800000 = 88.09% 최종 best<br><br>
<b>FINAL BEST 워크플로</b>: Flux Dev Q5 GGUF + PuLID 0.7 mean (byeolyi-default.png) + Redux 1.0 multiply + uni_pc + karras + 25 step + cfg 1.0 + guidance 3.5 + size 768x1344<br>
경로: <code>assets-library/workflows/FINAL_BEST-byeolyi-pastel88-uni_pc-karras.json</code><br><br>
⚠ vision 검수 대기 (별이≥8 / 한자≥4.5 / 한옥≥4.5) — auto 지표는 proxy<br>
60갑자 Qwen 트랙: 5회 시도 모두 black 100% (워크플로 구조 결함 확정, 향후 Qwen-Image native 시도 필요)
meeting-103 v2.1 clean R1 ✅5/5: PuLID + Redux = pastel 48.5% (R9 best 갱신) R0 환경 점검: - VRAM 11.7/12.4GB free (94%) — 해소 확인 - pytorch 2.12.0+cu130 + sageattention venv 검증 - 모델 인벤토리: Flux Dev Q5 GGUF, PuLID, Redux, byeolyi-ref-916 확보 - fp8 다운로드 .part 0 bytes (재개 대기) R1 clean 5조합 (5/5 success): - 1-1: PuLID w0.85 mean → pastel 25.7% / black 1.95% - 1-2: PuLID w0.85 mean delayed_start 0.05 → pastel 10.0% / black 14.0% (어두움) - 1-3: PuLID w1.0 mean → pastel 27.5% / black 0.79% - 1-4: PuLID w0.7 + Redux 0.5 → pastel 48.5% / black 0.27% ★ R1 best - 1-5: 한옥 baseline Flux Dev 768² → pastel 12.9% / black 11.7% CUDA error 회복: - 1-2 retry concat → CUDA invalid argument → 후속 1-2~1-5 모두 실패 - ComfyUI venv 재시작 (sage + pytorch 2.12) → 1-2 변형 (delayed_start)로 대체 - 재시도 4/4 모두 success 이전 R9 best (44.5%) → R1-4 (48.5%) = +4 pp 갱신 자산 라이브러리 누적: - models-performance.json: 7 모델 추적 (flux-dev, pulid-flux, flux-redux 등) - workflows: winner_auto 1 + partial 2 - prompts/proven: illustration_auto 패턴 3건 - prompts/failed: pastel-low + CUDA error 패턴 - prompts/negative: A_byeolyi 패턴 etc. 검수: - Vault: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html - Gitea: results/meeting-103-v2.1-clean/round1/
2026-05-19 23:17:19 +09:00
</div>
<div class="grid">
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round1/1-1_thumb.png" data-original="../results/meeting-103-v2.1-clean/round1/1-1.png" alt="1-1">
<div class="label">
meeting-103 v2.1 clean R2 ✅5/5: PuLID 0.85+Redux 0.5 = pastel 49.99% (전체 best 갱신) R2 5조합 (5/5 success, retry 2회 포함): - 2-1: PuLID 0.7 + Redux 0.5 + Stage 3 Refiner (denoise 0.3) → 48.82% (Stage 3 효과 미미) - 2-2: PuLID 0.7 + Redux 0.7 → 48.93% - 2-3: PuLID 0.7 + Redux 0.3 → 38.52% (Redux 낮으면 파스텔 감소) - 2-4: PuLID 0.85 + Redux 0.5 → 49.99% ★ R2 best & 전체 best - 2-5: 한옥 spring (cherry blossom) → 44.45% / black 4.3% R1 best (R1-4 48.5%) → R2-4 (49.99%) = +1.5 pp 갱신 R9 legacy (44.5%) → R2-4 (49.99%) = +5.5 pp 누적 학습: - PuLID weight 0.85 + Redux 0.5 multiply 조합이 0.7/0.5 보다 우월 - Redux 0.3 너무 낮음, 0.7 비슷, 0.85+ 미검증 - Stage 3 Refiner 효과 미미 (48.5% → 48.82%, +0.3pp) CUDA error 회복: - 2-3, 2-4, 2-5 첫 시도 CUDA invalid argument - ComfyUI venv 재시작 후 retry → 3/3 모두 success - 매 ~5건 후 재시작 패턴 확인됨 자산 라이브러리: - models-performance: 8 모델 추적 - workflows: winner_auto 4 (R1-4, R2-1, R2-2, R2-4) + partial 5 - prompts/proven: 7 패턴 (R1, R2 합산) R3 계획: - 3-1: PuLID 0.7 + Redux 0.85 (push redux 더 강) - 3-2: PuLID 0.6 + Redux 0.7 (PuLID 약화) - 3-3: PuLID 0.7 + Redux 0.7 @ 1024x1792 (해상도 증가) - 3-4: Qwen-Image fp8 + Lightning LoRA 60갑자 (fp8 ~20G 완료, GGUF 회피) - 3-5: Pony V6 img2img from byeolyi-ref (대안 ID 방법) 검수: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html
2026-05-19 23:36:44 +09:00
<div class="label-title">R1-1-1: A — PuLID weight 0.85 mean (first try)</div>
meeting-103 v2.1 clean R3+R4 ✅: PuLID 0.7+Redux 1.0 = pastel 56.41% (전체 best 갱신) R3 (4/5 success): - 3-1: PuLID 0.7 + Redux 0.85 → 50.47% (R2 best 49.99% 갱신) - 3-2: PuLID 0.6 + Redux 0.7 → 49.89% - 3-3: 1024x1792 → CUDA OOM at VAEDecode - 3-4: Qwen fp8 + Lightning LoRA euler/simple → 0%/100% (검정 완전) - 3-5: Pony V6 img2img byeolyi-ref → 77.3% (img2img 0.65 denoise, 원본 색 보존) R4 (3/5 success, 별이 새 best): - 4-1: PuLID 0.7 + Redux 1.0 (max) → 56.41% ★ 전체 best - 4-2: PuLID 0.85 + Redux 0.7 → 50.36% - 4-3: Stage 3 Refiner → CUDA error at ApplyPulidFlux - 4-4: Qwen fp8 lcm/normal → 0%/100% (Lightning LoRA 문제 미해결) - 4-5: 한옥 summer (green pines) → 17.7%/3.97% 학습: - Redux strength 0.5 → 0.7 → 0.85 → 1.0 단조 증가 (56.41% sweet spot) - PuLID 0.7 fixed, Redux 1.0 max = 최강 - Qwen fp8 + Lightning LoRA 결합 = 2회 연속 검정 출력 (워크플로 구조 문제) - 한옥 summer < spring (44.45%) — 색감 자체가 차분 진행: R9 legacy 44.5% → R1 48.5% → R2 49.99% → R3 50.47% → R4 56.41% (+11.91pp) 자산 라이브러리: - workflows: winner_auto 7 + partial 7 - prompts/failed: 9 패턴 (Qwen 계열 2건 추가)
2026-05-19 23:49:55 +09:00
<div class="label-meta">models=flux-dev, pulid-flux</div>
<div class="label-auto-eval">pastel=25.65% / black=1.95%</div>
meeting-103 v2.1 clean R1 ✅5/5: PuLID + Redux = pastel 48.5% (R9 best 갱신) R0 환경 점검: - VRAM 11.7/12.4GB free (94%) — 해소 확인 - pytorch 2.12.0+cu130 + sageattention venv 검증 - 모델 인벤토리: Flux Dev Q5 GGUF, PuLID, Redux, byeolyi-ref-916 확보 - fp8 다운로드 .part 0 bytes (재개 대기) R1 clean 5조합 (5/5 success): - 1-1: PuLID w0.85 mean → pastel 25.7% / black 1.95% - 1-2: PuLID w0.85 mean delayed_start 0.05 → pastel 10.0% / black 14.0% (어두움) - 1-3: PuLID w1.0 mean → pastel 27.5% / black 0.79% - 1-4: PuLID w0.7 + Redux 0.5 → pastel 48.5% / black 0.27% ★ R1 best - 1-5: 한옥 baseline Flux Dev 768² → pastel 12.9% / black 11.7% CUDA error 회복: - 1-2 retry concat → CUDA invalid argument → 후속 1-2~1-5 모두 실패 - ComfyUI venv 재시작 (sage + pytorch 2.12) → 1-2 변형 (delayed_start)로 대체 - 재시도 4/4 모두 success 이전 R9 best (44.5%) → R1-4 (48.5%) = +4 pp 갱신 자산 라이브러리 누적: - models-performance.json: 7 모델 추적 (flux-dev, pulid-flux, flux-redux 등) - workflows: winner_auto 1 + partial 2 - prompts/proven: illustration_auto 패턴 3건 - prompts/failed: pastel-low + CUDA error 패턴 - prompts/negative: A_byeolyi 패턴 etc. 검수: - Vault: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html - Gitea: results/meeting-103-v2.1-clean/round1/
2026-05-19 23:17:19 +09:00
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round1/1-2_thumb.png" data-original="../results/meeting-103-v2.1-clean/round1/1-2.png" alt="1-2">
<div class="label">
meeting-103 v2.1 clean R2 ✅5/5: PuLID 0.85+Redux 0.5 = pastel 49.99% (전체 best 갱신) R2 5조합 (5/5 success, retry 2회 포함): - 2-1: PuLID 0.7 + Redux 0.5 + Stage 3 Refiner (denoise 0.3) → 48.82% (Stage 3 효과 미미) - 2-2: PuLID 0.7 + Redux 0.7 → 48.93% - 2-3: PuLID 0.7 + Redux 0.3 → 38.52% (Redux 낮으면 파스텔 감소) - 2-4: PuLID 0.85 + Redux 0.5 → 49.99% ★ R2 best & 전체 best - 2-5: 한옥 spring (cherry blossom) → 44.45% / black 4.3% R1 best (R1-4 48.5%) → R2-4 (49.99%) = +1.5 pp 갱신 R9 legacy (44.5%) → R2-4 (49.99%) = +5.5 pp 누적 학습: - PuLID weight 0.85 + Redux 0.5 multiply 조합이 0.7/0.5 보다 우월 - Redux 0.3 너무 낮음, 0.7 비슷, 0.85+ 미검증 - Stage 3 Refiner 효과 미미 (48.5% → 48.82%, +0.3pp) CUDA error 회복: - 2-3, 2-4, 2-5 첫 시도 CUDA invalid argument - ComfyUI venv 재시작 후 retry → 3/3 모두 success - 매 ~5건 후 재시작 패턴 확인됨 자산 라이브러리: - models-performance: 8 모델 추적 - workflows: winner_auto 4 (R1-4, R2-1, R2-2, R2-4) + partial 5 - prompts/proven: 7 패턴 (R1, R2 합산) R3 계획: - 3-1: PuLID 0.7 + Redux 0.85 (push redux 더 강) - 3-2: PuLID 0.6 + Redux 0.7 (PuLID 약화) - 3-3: PuLID 0.7 + Redux 0.7 @ 1024x1792 (해상도 증가) - 3-4: Qwen-Image fp8 + Lightning LoRA 60갑자 (fp8 ~20G 완료, GGUF 회피) - 3-5: Pony V6 img2img from byeolyi-ref (대안 ID 방법) 검수: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html
2026-05-19 23:36:44 +09:00
<div class="label-title">R1-1-2: A — PuLID weight 0.85 mean delayed_start 0.05 (concat-replacement)</div>
meeting-103 v2.1 clean R3+R4 ✅: PuLID 0.7+Redux 1.0 = pastel 56.41% (전체 best 갱신) R3 (4/5 success): - 3-1: PuLID 0.7 + Redux 0.85 → 50.47% (R2 best 49.99% 갱신) - 3-2: PuLID 0.6 + Redux 0.7 → 49.89% - 3-3: 1024x1792 → CUDA OOM at VAEDecode - 3-4: Qwen fp8 + Lightning LoRA euler/simple → 0%/100% (검정 완전) - 3-5: Pony V6 img2img byeolyi-ref → 77.3% (img2img 0.65 denoise, 원본 색 보존) R4 (3/5 success, 별이 새 best): - 4-1: PuLID 0.7 + Redux 1.0 (max) → 56.41% ★ 전체 best - 4-2: PuLID 0.85 + Redux 0.7 → 50.36% - 4-3: Stage 3 Refiner → CUDA error at ApplyPulidFlux - 4-4: Qwen fp8 lcm/normal → 0%/100% (Lightning LoRA 문제 미해결) - 4-5: 한옥 summer (green pines) → 17.7%/3.97% 학습: - Redux strength 0.5 → 0.7 → 0.85 → 1.0 단조 증가 (56.41% sweet spot) - PuLID 0.7 fixed, Redux 1.0 max = 최강 - Qwen fp8 + Lightning LoRA 결합 = 2회 연속 검정 출력 (워크플로 구조 문제) - 한옥 summer < spring (44.45%) — 색감 자체가 차분 진행: R9 legacy 44.5% → R1 48.5% → R2 49.99% → R3 50.47% → R4 56.41% (+11.91pp) 자산 라이브러리: - workflows: winner_auto 7 + partial 7 - prompts/failed: 9 패턴 (Qwen 계열 2건 추가)
2026-05-19 23:49:55 +09:00
<div class="label-meta">models=flux-dev, pulid-flux</div>
<div class="label-auto-eval">pastel=9.99% / black=13.98%</div>
meeting-103 v2.1 clean R2 ✅5/5: PuLID 0.85+Redux 0.5 = pastel 49.99% (전체 best 갱신) R2 5조합 (5/5 success, retry 2회 포함): - 2-1: PuLID 0.7 + Redux 0.5 + Stage 3 Refiner (denoise 0.3) → 48.82% (Stage 3 효과 미미) - 2-2: PuLID 0.7 + Redux 0.7 → 48.93% - 2-3: PuLID 0.7 + Redux 0.3 → 38.52% (Redux 낮으면 파스텔 감소) - 2-4: PuLID 0.85 + Redux 0.5 → 49.99% ★ R2 best & 전체 best - 2-5: 한옥 spring (cherry blossom) → 44.45% / black 4.3% R1 best (R1-4 48.5%) → R2-4 (49.99%) = +1.5 pp 갱신 R9 legacy (44.5%) → R2-4 (49.99%) = +5.5 pp 누적 학습: - PuLID weight 0.85 + Redux 0.5 multiply 조합이 0.7/0.5 보다 우월 - Redux 0.3 너무 낮음, 0.7 비슷, 0.85+ 미검증 - Stage 3 Refiner 효과 미미 (48.5% → 48.82%, +0.3pp) CUDA error 회복: - 2-3, 2-4, 2-5 첫 시도 CUDA invalid argument - ComfyUI venv 재시작 후 retry → 3/3 모두 success - 매 ~5건 후 재시작 패턴 확인됨 자산 라이브러리: - models-performance: 8 모델 추적 - workflows: winner_auto 4 (R1-4, R2-1, R2-2, R2-4) + partial 5 - prompts/proven: 7 패턴 (R1, R2 합산) R3 계획: - 3-1: PuLID 0.7 + Redux 0.85 (push redux 더 강) - 3-2: PuLID 0.6 + Redux 0.7 (PuLID 약화) - 3-3: PuLID 0.7 + Redux 0.7 @ 1024x1792 (해상도 증가) - 3-4: Qwen-Image fp8 + Lightning LoRA 60갑자 (fp8 ~20G 완료, GGUF 회피) - 3-5: Pony V6 img2img from byeolyi-ref (대안 ID 방법) 검수: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html
2026-05-19 23:36:44 +09:00
<div class="label-flags"><span class="flag flag-warn">black-suspicious</span><span class="flag flag-warn">pastel-low</span></div>
meeting-103 v2.1 clean R1 ✅5/5: PuLID + Redux = pastel 48.5% (R9 best 갱신) R0 환경 점검: - VRAM 11.7/12.4GB free (94%) — 해소 확인 - pytorch 2.12.0+cu130 + sageattention venv 검증 - 모델 인벤토리: Flux Dev Q5 GGUF, PuLID, Redux, byeolyi-ref-916 확보 - fp8 다운로드 .part 0 bytes (재개 대기) R1 clean 5조합 (5/5 success): - 1-1: PuLID w0.85 mean → pastel 25.7% / black 1.95% - 1-2: PuLID w0.85 mean delayed_start 0.05 → pastel 10.0% / black 14.0% (어두움) - 1-3: PuLID w1.0 mean → pastel 27.5% / black 0.79% - 1-4: PuLID w0.7 + Redux 0.5 → pastel 48.5% / black 0.27% ★ R1 best - 1-5: 한옥 baseline Flux Dev 768² → pastel 12.9% / black 11.7% CUDA error 회복: - 1-2 retry concat → CUDA invalid argument → 후속 1-2~1-5 모두 실패 - ComfyUI venv 재시작 (sage + pytorch 2.12) → 1-2 변형 (delayed_start)로 대체 - 재시도 4/4 모두 success 이전 R9 best (44.5%) → R1-4 (48.5%) = +4 pp 갱신 자산 라이브러리 누적: - models-performance.json: 7 모델 추적 (flux-dev, pulid-flux, flux-redux 등) - workflows: winner_auto 1 + partial 2 - prompts/proven: illustration_auto 패턴 3건 - prompts/failed: pastel-low + CUDA error 패턴 - prompts/negative: A_byeolyi 패턴 etc. 검수: - Vault: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html - Gitea: results/meeting-103-v2.1-clean/round1/
2026-05-19 23:17:19 +09:00
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round1/1-3_thumb.png" data-original="../results/meeting-103-v2.1-clean/round1/1-3.png" alt="1-3">
<div class="label">
meeting-103 v2.1 clean R2 ✅5/5: PuLID 0.85+Redux 0.5 = pastel 49.99% (전체 best 갱신) R2 5조합 (5/5 success, retry 2회 포함): - 2-1: PuLID 0.7 + Redux 0.5 + Stage 3 Refiner (denoise 0.3) → 48.82% (Stage 3 효과 미미) - 2-2: PuLID 0.7 + Redux 0.7 → 48.93% - 2-3: PuLID 0.7 + Redux 0.3 → 38.52% (Redux 낮으면 파스텔 감소) - 2-4: PuLID 0.85 + Redux 0.5 → 49.99% ★ R2 best & 전체 best - 2-5: 한옥 spring (cherry blossom) → 44.45% / black 4.3% R1 best (R1-4 48.5%) → R2-4 (49.99%) = +1.5 pp 갱신 R9 legacy (44.5%) → R2-4 (49.99%) = +5.5 pp 누적 학습: - PuLID weight 0.85 + Redux 0.5 multiply 조합이 0.7/0.5 보다 우월 - Redux 0.3 너무 낮음, 0.7 비슷, 0.85+ 미검증 - Stage 3 Refiner 효과 미미 (48.5% → 48.82%, +0.3pp) CUDA error 회복: - 2-3, 2-4, 2-5 첫 시도 CUDA invalid argument - ComfyUI venv 재시작 후 retry → 3/3 모두 success - 매 ~5건 후 재시작 패턴 확인됨 자산 라이브러리: - models-performance: 8 모델 추적 - workflows: winner_auto 4 (R1-4, R2-1, R2-2, R2-4) + partial 5 - prompts/proven: 7 패턴 (R1, R2 합산) R3 계획: - 3-1: PuLID 0.7 + Redux 0.85 (push redux 더 강) - 3-2: PuLID 0.6 + Redux 0.7 (PuLID 약화) - 3-3: PuLID 0.7 + Redux 0.7 @ 1024x1792 (해상도 증가) - 3-4: Qwen-Image fp8 + Lightning LoRA 60갑자 (fp8 ~20G 완료, GGUF 회피) - 3-5: Pony V6 img2img from byeolyi-ref (대안 ID 방법) 검수: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html
2026-05-19 23:36:44 +09:00
<div class="label-title">R1-1-3: A — PuLID weight 1.0 mean</div>
meeting-103 v2.1 clean R3+R4 ✅: PuLID 0.7+Redux 1.0 = pastel 56.41% (전체 best 갱신) R3 (4/5 success): - 3-1: PuLID 0.7 + Redux 0.85 → 50.47% (R2 best 49.99% 갱신) - 3-2: PuLID 0.6 + Redux 0.7 → 49.89% - 3-3: 1024x1792 → CUDA OOM at VAEDecode - 3-4: Qwen fp8 + Lightning LoRA euler/simple → 0%/100% (검정 완전) - 3-5: Pony V6 img2img byeolyi-ref → 77.3% (img2img 0.65 denoise, 원본 색 보존) R4 (3/5 success, 별이 새 best): - 4-1: PuLID 0.7 + Redux 1.0 (max) → 56.41% ★ 전체 best - 4-2: PuLID 0.85 + Redux 0.7 → 50.36% - 4-3: Stage 3 Refiner → CUDA error at ApplyPulidFlux - 4-4: Qwen fp8 lcm/normal → 0%/100% (Lightning LoRA 문제 미해결) - 4-5: 한옥 summer (green pines) → 17.7%/3.97% 학습: - Redux strength 0.5 → 0.7 → 0.85 → 1.0 단조 증가 (56.41% sweet spot) - PuLID 0.7 fixed, Redux 1.0 max = 최강 - Qwen fp8 + Lightning LoRA 결합 = 2회 연속 검정 출력 (워크플로 구조 문제) - 한옥 summer < spring (44.45%) — 색감 자체가 차분 진행: R9 legacy 44.5% → R1 48.5% → R2 49.99% → R3 50.47% → R4 56.41% (+11.91pp) 자산 라이브러리: - workflows: winner_auto 7 + partial 7 - prompts/failed: 9 패턴 (Qwen 계열 2건 추가)
2026-05-19 23:49:55 +09:00
<div class="label-meta">models=flux-dev, pulid-flux</div>
<div class="label-auto-eval">pastel=27.51% / black=0.79%</div>
meeting-103 v2.1 clean R1 ✅5/5: PuLID + Redux = pastel 48.5% (R9 best 갱신) R0 환경 점검: - VRAM 11.7/12.4GB free (94%) — 해소 확인 - pytorch 2.12.0+cu130 + sageattention venv 검증 - 모델 인벤토리: Flux Dev Q5 GGUF, PuLID, Redux, byeolyi-ref-916 확보 - fp8 다운로드 .part 0 bytes (재개 대기) R1 clean 5조합 (5/5 success): - 1-1: PuLID w0.85 mean → pastel 25.7% / black 1.95% - 1-2: PuLID w0.85 mean delayed_start 0.05 → pastel 10.0% / black 14.0% (어두움) - 1-3: PuLID w1.0 mean → pastel 27.5% / black 0.79% - 1-4: PuLID w0.7 + Redux 0.5 → pastel 48.5% / black 0.27% ★ R1 best - 1-5: 한옥 baseline Flux Dev 768² → pastel 12.9% / black 11.7% CUDA error 회복: - 1-2 retry concat → CUDA invalid argument → 후속 1-2~1-5 모두 실패 - ComfyUI venv 재시작 (sage + pytorch 2.12) → 1-2 변형 (delayed_start)로 대체 - 재시도 4/4 모두 success 이전 R9 best (44.5%) → R1-4 (48.5%) = +4 pp 갱신 자산 라이브러리 누적: - models-performance.json: 7 모델 추적 (flux-dev, pulid-flux, flux-redux 등) - workflows: winner_auto 1 + partial 2 - prompts/proven: illustration_auto 패턴 3건 - prompts/failed: pastel-low + CUDA error 패턴 - prompts/negative: A_byeolyi 패턴 etc. 검수: - Vault: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html - Gitea: results/meeting-103-v2.1-clean/round1/
2026-05-19 23:17:19 +09:00
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round1/1-4_thumb.png" data-original="../results/meeting-103-v2.1-clean/round1/1-4.png" alt="1-4">
<div class="label">
meeting-103 v2.1 clean R2 ✅5/5: PuLID 0.85+Redux 0.5 = pastel 49.99% (전체 best 갱신) R2 5조합 (5/5 success, retry 2회 포함): - 2-1: PuLID 0.7 + Redux 0.5 + Stage 3 Refiner (denoise 0.3) → 48.82% (Stage 3 효과 미미) - 2-2: PuLID 0.7 + Redux 0.7 → 48.93% - 2-3: PuLID 0.7 + Redux 0.3 → 38.52% (Redux 낮으면 파스텔 감소) - 2-4: PuLID 0.85 + Redux 0.5 → 49.99% ★ R2 best & 전체 best - 2-5: 한옥 spring (cherry blossom) → 44.45% / black 4.3% R1 best (R1-4 48.5%) → R2-4 (49.99%) = +1.5 pp 갱신 R9 legacy (44.5%) → R2-4 (49.99%) = +5.5 pp 누적 학습: - PuLID weight 0.85 + Redux 0.5 multiply 조합이 0.7/0.5 보다 우월 - Redux 0.3 너무 낮음, 0.7 비슷, 0.85+ 미검증 - Stage 3 Refiner 효과 미미 (48.5% → 48.82%, +0.3pp) CUDA error 회복: - 2-3, 2-4, 2-5 첫 시도 CUDA invalid argument - ComfyUI venv 재시작 후 retry → 3/3 모두 success - 매 ~5건 후 재시작 패턴 확인됨 자산 라이브러리: - models-performance: 8 모델 추적 - workflows: winner_auto 4 (R1-4, R2-1, R2-2, R2-4) + partial 5 - prompts/proven: 7 패턴 (R1, R2 합산) R3 계획: - 3-1: PuLID 0.7 + Redux 0.85 (push redux 더 강) - 3-2: PuLID 0.6 + Redux 0.7 (PuLID 약화) - 3-3: PuLID 0.7 + Redux 0.7 @ 1024x1792 (해상도 증가) - 3-4: Qwen-Image fp8 + Lightning LoRA 60갑자 (fp8 ~20G 완료, GGUF 회피) - 3-5: Pony V6 img2img from byeolyi-ref (대안 ID 방법) 검수: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html
2026-05-19 23:36:44 +09:00
<div class="label-title">R1-1-4: A — PuLID 0.7 + Redux 0.5</div>
meeting-103 v2.1 clean R3+R4 ✅: PuLID 0.7+Redux 1.0 = pastel 56.41% (전체 best 갱신) R3 (4/5 success): - 3-1: PuLID 0.7 + Redux 0.85 → 50.47% (R2 best 49.99% 갱신) - 3-2: PuLID 0.6 + Redux 0.7 → 49.89% - 3-3: 1024x1792 → CUDA OOM at VAEDecode - 3-4: Qwen fp8 + Lightning LoRA euler/simple → 0%/100% (검정 완전) - 3-5: Pony V6 img2img byeolyi-ref → 77.3% (img2img 0.65 denoise, 원본 색 보존) R4 (3/5 success, 별이 새 best): - 4-1: PuLID 0.7 + Redux 1.0 (max) → 56.41% ★ 전체 best - 4-2: PuLID 0.85 + Redux 0.7 → 50.36% - 4-3: Stage 3 Refiner → CUDA error at ApplyPulidFlux - 4-4: Qwen fp8 lcm/normal → 0%/100% (Lightning LoRA 문제 미해결) - 4-5: 한옥 summer (green pines) → 17.7%/3.97% 학습: - Redux strength 0.5 → 0.7 → 0.85 → 1.0 단조 증가 (56.41% sweet spot) - PuLID 0.7 fixed, Redux 1.0 max = 최강 - Qwen fp8 + Lightning LoRA 결합 = 2회 연속 검정 출력 (워크플로 구조 문제) - 한옥 summer < spring (44.45%) — 색감 자체가 차분 진행: R9 legacy 44.5% → R1 48.5% → R2 49.99% → R3 50.47% → R4 56.41% (+11.91pp) 자산 라이브러리: - workflows: winner_auto 7 + partial 7 - prompts/failed: 9 패턴 (Qwen 계열 2건 추가)
2026-05-19 23:49:55 +09:00
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=48.5% / black=0.27%</div>
meeting-103 v2.1 clean R1 ✅5/5: PuLID + Redux = pastel 48.5% (R9 best 갱신) R0 환경 점검: - VRAM 11.7/12.4GB free (94%) — 해소 확인 - pytorch 2.12.0+cu130 + sageattention venv 검증 - 모델 인벤토리: Flux Dev Q5 GGUF, PuLID, Redux, byeolyi-ref-916 확보 - fp8 다운로드 .part 0 bytes (재개 대기) R1 clean 5조합 (5/5 success): - 1-1: PuLID w0.85 mean → pastel 25.7% / black 1.95% - 1-2: PuLID w0.85 mean delayed_start 0.05 → pastel 10.0% / black 14.0% (어두움) - 1-3: PuLID w1.0 mean → pastel 27.5% / black 0.79% - 1-4: PuLID w0.7 + Redux 0.5 → pastel 48.5% / black 0.27% ★ R1 best - 1-5: 한옥 baseline Flux Dev 768² → pastel 12.9% / black 11.7% CUDA error 회복: - 1-2 retry concat → CUDA invalid argument → 후속 1-2~1-5 모두 실패 - ComfyUI venv 재시작 (sage + pytorch 2.12) → 1-2 변형 (delayed_start)로 대체 - 재시도 4/4 모두 success 이전 R9 best (44.5%) → R1-4 (48.5%) = +4 pp 갱신 자산 라이브러리 누적: - models-performance.json: 7 모델 추적 (flux-dev, pulid-flux, flux-redux 등) - workflows: winner_auto 1 + partial 2 - prompts/proven: illustration_auto 패턴 3건 - prompts/failed: pastel-low + CUDA error 패턴 - prompts/negative: A_byeolyi 패턴 etc. 검수: - Vault: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html - Gitea: results/meeting-103-v2.1-clean/round1/
2026-05-19 23:17:19 +09:00
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round1/1-5_thumb.png" data-original="../results/meeting-103-v2.1-clean/round1/1-5.png" alt="1-5">
<div class="label">
meeting-103 v2.1 clean R2 ✅5/5: PuLID 0.85+Redux 0.5 = pastel 49.99% (전체 best 갱신) R2 5조합 (5/5 success, retry 2회 포함): - 2-1: PuLID 0.7 + Redux 0.5 + Stage 3 Refiner (denoise 0.3) → 48.82% (Stage 3 효과 미미) - 2-2: PuLID 0.7 + Redux 0.7 → 48.93% - 2-3: PuLID 0.7 + Redux 0.3 → 38.52% (Redux 낮으면 파스텔 감소) - 2-4: PuLID 0.85 + Redux 0.5 → 49.99% ★ R2 best & 전체 best - 2-5: 한옥 spring (cherry blossom) → 44.45% / black 4.3% R1 best (R1-4 48.5%) → R2-4 (49.99%) = +1.5 pp 갱신 R9 legacy (44.5%) → R2-4 (49.99%) = +5.5 pp 누적 학습: - PuLID weight 0.85 + Redux 0.5 multiply 조합이 0.7/0.5 보다 우월 - Redux 0.3 너무 낮음, 0.7 비슷, 0.85+ 미검증 - Stage 3 Refiner 효과 미미 (48.5% → 48.82%, +0.3pp) CUDA error 회복: - 2-3, 2-4, 2-5 첫 시도 CUDA invalid argument - ComfyUI venv 재시작 후 retry → 3/3 모두 success - 매 ~5건 후 재시작 패턴 확인됨 자산 라이브러리: - models-performance: 8 모델 추적 - workflows: winner_auto 4 (R1-4, R2-1, R2-2, R2-4) + partial 5 - prompts/proven: 7 패턴 (R1, R2 합산) R3 계획: - 3-1: PuLID 0.7 + Redux 0.85 (push redux 더 강) - 3-2: PuLID 0.6 + Redux 0.7 (PuLID 약화) - 3-3: PuLID 0.7 + Redux 0.7 @ 1024x1792 (해상도 증가) - 3-4: Qwen-Image fp8 + Lightning LoRA 60갑자 (fp8 ~20G 완료, GGUF 회피) - 3-5: Pony V6 img2img from byeolyi-ref (대안 ID 방법) 검수: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html
2026-05-19 23:36:44 +09:00
<div class="label-title">R1-1-5: B — Hanok baseline Flux Dev 768²</div>
meeting-103 v2.1 clean R3+R4 ✅: PuLID 0.7+Redux 1.0 = pastel 56.41% (전체 best 갱신) R3 (4/5 success): - 3-1: PuLID 0.7 + Redux 0.85 → 50.47% (R2 best 49.99% 갱신) - 3-2: PuLID 0.6 + Redux 0.7 → 49.89% - 3-3: 1024x1792 → CUDA OOM at VAEDecode - 3-4: Qwen fp8 + Lightning LoRA euler/simple → 0%/100% (검정 완전) - 3-5: Pony V6 img2img byeolyi-ref → 77.3% (img2img 0.65 denoise, 원본 색 보존) R4 (3/5 success, 별이 새 best): - 4-1: PuLID 0.7 + Redux 1.0 (max) → 56.41% ★ 전체 best - 4-2: PuLID 0.85 + Redux 0.7 → 50.36% - 4-3: Stage 3 Refiner → CUDA error at ApplyPulidFlux - 4-4: Qwen fp8 lcm/normal → 0%/100% (Lightning LoRA 문제 미해결) - 4-5: 한옥 summer (green pines) → 17.7%/3.97% 학습: - Redux strength 0.5 → 0.7 → 0.85 → 1.0 단조 증가 (56.41% sweet spot) - PuLID 0.7 fixed, Redux 1.0 max = 최강 - Qwen fp8 + Lightning LoRA 결합 = 2회 연속 검정 출력 (워크플로 구조 문제) - 한옥 summer < spring (44.45%) — 색감 자체가 차분 진행: R9 legacy 44.5% → R1 48.5% → R2 49.99% → R3 50.47% → R4 56.41% (+11.91pp) 자산 라이브러리: - workflows: winner_auto 7 + partial 7 - prompts/failed: 9 패턴 (Qwen 계열 2건 추가)
2026-05-19 23:49:55 +09:00
<div class="label-meta">models=flux-dev</div>
<div class="label-auto-eval">pastel=12.87% / black=11.67%</div>
meeting-103 v2.1 clean R2 ✅5/5: PuLID 0.85+Redux 0.5 = pastel 49.99% (전체 best 갱신) R2 5조합 (5/5 success, retry 2회 포함): - 2-1: PuLID 0.7 + Redux 0.5 + Stage 3 Refiner (denoise 0.3) → 48.82% (Stage 3 효과 미미) - 2-2: PuLID 0.7 + Redux 0.7 → 48.93% - 2-3: PuLID 0.7 + Redux 0.3 → 38.52% (Redux 낮으면 파스텔 감소) - 2-4: PuLID 0.85 + Redux 0.5 → 49.99% ★ R2 best & 전체 best - 2-5: 한옥 spring (cherry blossom) → 44.45% / black 4.3% R1 best (R1-4 48.5%) → R2-4 (49.99%) = +1.5 pp 갱신 R9 legacy (44.5%) → R2-4 (49.99%) = +5.5 pp 누적 학습: - PuLID weight 0.85 + Redux 0.5 multiply 조합이 0.7/0.5 보다 우월 - Redux 0.3 너무 낮음, 0.7 비슷, 0.85+ 미검증 - Stage 3 Refiner 효과 미미 (48.5% → 48.82%, +0.3pp) CUDA error 회복: - 2-3, 2-4, 2-5 첫 시도 CUDA invalid argument - ComfyUI venv 재시작 후 retry → 3/3 모두 success - 매 ~5건 후 재시작 패턴 확인됨 자산 라이브러리: - models-performance: 8 모델 추적 - workflows: winner_auto 4 (R1-4, R2-1, R2-2, R2-4) + partial 5 - prompts/proven: 7 패턴 (R1, R2 합산) R3 계획: - 3-1: PuLID 0.7 + Redux 0.85 (push redux 더 강) - 3-2: PuLID 0.6 + Redux 0.7 (PuLID 약화) - 3-3: PuLID 0.7 + Redux 0.7 @ 1024x1792 (해상도 증가) - 3-4: Qwen-Image fp8 + Lightning LoRA 60갑자 (fp8 ~20G 완료, GGUF 회피) - 3-5: Pony V6 img2img from byeolyi-ref (대안 ID 방법) 검수: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html
2026-05-19 23:36:44 +09:00
<div class="label-flags"><span class="flag flag-warn">black-suspicious</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round2/2-1_thumb.png" data-original="../results/meeting-103-v2.1-clean/round2/2-1.png" alt="2-1">
<div class="label">
<div class="label-title">R2-2-1: A — PuLID 0.7 + Redux 0.5 + Stage 3 Refiner (denoise 0.3)</div>
meeting-103 v2.1 clean R3+R4 ✅: PuLID 0.7+Redux 1.0 = pastel 56.41% (전체 best 갱신) R3 (4/5 success): - 3-1: PuLID 0.7 + Redux 0.85 → 50.47% (R2 best 49.99% 갱신) - 3-2: PuLID 0.6 + Redux 0.7 → 49.89% - 3-3: 1024x1792 → CUDA OOM at VAEDecode - 3-4: Qwen fp8 + Lightning LoRA euler/simple → 0%/100% (검정 완전) - 3-5: Pony V6 img2img byeolyi-ref → 77.3% (img2img 0.65 denoise, 원본 색 보존) R4 (3/5 success, 별이 새 best): - 4-1: PuLID 0.7 + Redux 1.0 (max) → 56.41% ★ 전체 best - 4-2: PuLID 0.85 + Redux 0.7 → 50.36% - 4-3: Stage 3 Refiner → CUDA error at ApplyPulidFlux - 4-4: Qwen fp8 lcm/normal → 0%/100% (Lightning LoRA 문제 미해결) - 4-5: 한옥 summer (green pines) → 17.7%/3.97% 학습: - Redux strength 0.5 → 0.7 → 0.85 → 1.0 단조 증가 (56.41% sweet spot) - PuLID 0.7 fixed, Redux 1.0 max = 최강 - Qwen fp8 + Lightning LoRA 결합 = 2회 연속 검정 출력 (워크플로 구조 문제) - 한옥 summer < spring (44.45%) — 색감 자체가 차분 진행: R9 legacy 44.5% → R1 48.5% → R2 49.99% → R3 50.47% → R4 56.41% (+11.91pp) 자산 라이브러리: - workflows: winner_auto 7 + partial 7 - prompts/failed: 9 패턴 (Qwen 계열 2건 추가)
2026-05-19 23:49:55 +09:00
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux, flux-dev-refiner</div>
<div class="label-auto-eval">pastel=48.82% / black=0.34%</div>
meeting-103 v2.1 clean R2 ✅5/5: PuLID 0.85+Redux 0.5 = pastel 49.99% (전체 best 갱신) R2 5조합 (5/5 success, retry 2회 포함): - 2-1: PuLID 0.7 + Redux 0.5 + Stage 3 Refiner (denoise 0.3) → 48.82% (Stage 3 효과 미미) - 2-2: PuLID 0.7 + Redux 0.7 → 48.93% - 2-3: PuLID 0.7 + Redux 0.3 → 38.52% (Redux 낮으면 파스텔 감소) - 2-4: PuLID 0.85 + Redux 0.5 → 49.99% ★ R2 best & 전체 best - 2-5: 한옥 spring (cherry blossom) → 44.45% / black 4.3% R1 best (R1-4 48.5%) → R2-4 (49.99%) = +1.5 pp 갱신 R9 legacy (44.5%) → R2-4 (49.99%) = +5.5 pp 누적 학습: - PuLID weight 0.85 + Redux 0.5 multiply 조합이 0.7/0.5 보다 우월 - Redux 0.3 너무 낮음, 0.7 비슷, 0.85+ 미검증 - Stage 3 Refiner 효과 미미 (48.5% → 48.82%, +0.3pp) CUDA error 회복: - 2-3, 2-4, 2-5 첫 시도 CUDA invalid argument - ComfyUI venv 재시작 후 retry → 3/3 모두 success - 매 ~5건 후 재시작 패턴 확인됨 자산 라이브러리: - models-performance: 8 모델 추적 - workflows: winner_auto 4 (R1-4, R2-1, R2-2, R2-4) + partial 5 - prompts/proven: 7 패턴 (R1, R2 합산) R3 계획: - 3-1: PuLID 0.7 + Redux 0.85 (push redux 더 강) - 3-2: PuLID 0.6 + Redux 0.7 (PuLID 약화) - 3-3: PuLID 0.7 + Redux 0.7 @ 1024x1792 (해상도 증가) - 3-4: Qwen-Image fp8 + Lightning LoRA 60갑자 (fp8 ~20G 완료, GGUF 회피) - 3-5: Pony V6 img2img from byeolyi-ref (대안 ID 방법) 검수: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html
2026-05-19 23:36:44 +09:00
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round2/2-2_thumb.png" data-original="../results/meeting-103-v2.1-clean/round2/2-2.png" alt="2-2">
<div class="label">
<div class="label-title">R2-2-2: A — PuLID 0.7 + Redux 0.7</div>
meeting-103 v2.1 clean R3+R4 ✅: PuLID 0.7+Redux 1.0 = pastel 56.41% (전체 best 갱신) R3 (4/5 success): - 3-1: PuLID 0.7 + Redux 0.85 → 50.47% (R2 best 49.99% 갱신) - 3-2: PuLID 0.6 + Redux 0.7 → 49.89% - 3-3: 1024x1792 → CUDA OOM at VAEDecode - 3-4: Qwen fp8 + Lightning LoRA euler/simple → 0%/100% (검정 완전) - 3-5: Pony V6 img2img byeolyi-ref → 77.3% (img2img 0.65 denoise, 원본 색 보존) R4 (3/5 success, 별이 새 best): - 4-1: PuLID 0.7 + Redux 1.0 (max) → 56.41% ★ 전체 best - 4-2: PuLID 0.85 + Redux 0.7 → 50.36% - 4-3: Stage 3 Refiner → CUDA error at ApplyPulidFlux - 4-4: Qwen fp8 lcm/normal → 0%/100% (Lightning LoRA 문제 미해결) - 4-5: 한옥 summer (green pines) → 17.7%/3.97% 학습: - Redux strength 0.5 → 0.7 → 0.85 → 1.0 단조 증가 (56.41% sweet spot) - PuLID 0.7 fixed, Redux 1.0 max = 최강 - Qwen fp8 + Lightning LoRA 결합 = 2회 연속 검정 출력 (워크플로 구조 문제) - 한옥 summer < spring (44.45%) — 색감 자체가 차분 진행: R9 legacy 44.5% → R1 48.5% → R2 49.99% → R3 50.47% → R4 56.41% (+11.91pp) 자산 라이브러리: - workflows: winner_auto 7 + partial 7 - prompts/failed: 9 패턴 (Qwen 계열 2건 추가)
2026-05-19 23:49:55 +09:00
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=48.93% / black=0.46%</div>
meeting-103 v2.1 clean R2 ✅5/5: PuLID 0.85+Redux 0.5 = pastel 49.99% (전체 best 갱신) R2 5조합 (5/5 success, retry 2회 포함): - 2-1: PuLID 0.7 + Redux 0.5 + Stage 3 Refiner (denoise 0.3) → 48.82% (Stage 3 효과 미미) - 2-2: PuLID 0.7 + Redux 0.7 → 48.93% - 2-3: PuLID 0.7 + Redux 0.3 → 38.52% (Redux 낮으면 파스텔 감소) - 2-4: PuLID 0.85 + Redux 0.5 → 49.99% ★ R2 best & 전체 best - 2-5: 한옥 spring (cherry blossom) → 44.45% / black 4.3% R1 best (R1-4 48.5%) → R2-4 (49.99%) = +1.5 pp 갱신 R9 legacy (44.5%) → R2-4 (49.99%) = +5.5 pp 누적 학습: - PuLID weight 0.85 + Redux 0.5 multiply 조합이 0.7/0.5 보다 우월 - Redux 0.3 너무 낮음, 0.7 비슷, 0.85+ 미검증 - Stage 3 Refiner 효과 미미 (48.5% → 48.82%, +0.3pp) CUDA error 회복: - 2-3, 2-4, 2-5 첫 시도 CUDA invalid argument - ComfyUI venv 재시작 후 retry → 3/3 모두 success - 매 ~5건 후 재시작 패턴 확인됨 자산 라이브러리: - models-performance: 8 모델 추적 - workflows: winner_auto 4 (R1-4, R2-1, R2-2, R2-4) + partial 5 - prompts/proven: 7 패턴 (R1, R2 합산) R3 계획: - 3-1: PuLID 0.7 + Redux 0.85 (push redux 더 강) - 3-2: PuLID 0.6 + Redux 0.7 (PuLID 약화) - 3-3: PuLID 0.7 + Redux 0.7 @ 1024x1792 (해상도 증가) - 3-4: Qwen-Image fp8 + Lightning LoRA 60갑자 (fp8 ~20G 완료, GGUF 회피) - 3-5: Pony V6 img2img from byeolyi-ref (대안 ID 방법) 검수: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html
2026-05-19 23:36:44 +09:00
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round2/2-3_thumb.png" data-original="../results/meeting-103-v2.1-clean/round2/2-3.png" alt="2-3">
<div class="label">
<div class="label-title">R2-2-3: A — PuLID 0.7 + Redux 0.3</div>
meeting-103 v2.1 clean R3+R4 ✅: PuLID 0.7+Redux 1.0 = pastel 56.41% (전체 best 갱신) R3 (4/5 success): - 3-1: PuLID 0.7 + Redux 0.85 → 50.47% (R2 best 49.99% 갱신) - 3-2: PuLID 0.6 + Redux 0.7 → 49.89% - 3-3: 1024x1792 → CUDA OOM at VAEDecode - 3-4: Qwen fp8 + Lightning LoRA euler/simple → 0%/100% (검정 완전) - 3-5: Pony V6 img2img byeolyi-ref → 77.3% (img2img 0.65 denoise, 원본 색 보존) R4 (3/5 success, 별이 새 best): - 4-1: PuLID 0.7 + Redux 1.0 (max) → 56.41% ★ 전체 best - 4-2: PuLID 0.85 + Redux 0.7 → 50.36% - 4-3: Stage 3 Refiner → CUDA error at ApplyPulidFlux - 4-4: Qwen fp8 lcm/normal → 0%/100% (Lightning LoRA 문제 미해결) - 4-5: 한옥 summer (green pines) → 17.7%/3.97% 학습: - Redux strength 0.5 → 0.7 → 0.85 → 1.0 단조 증가 (56.41% sweet spot) - PuLID 0.7 fixed, Redux 1.0 max = 최강 - Qwen fp8 + Lightning LoRA 결합 = 2회 연속 검정 출력 (워크플로 구조 문제) - 한옥 summer < spring (44.45%) — 색감 자체가 차분 진행: R9 legacy 44.5% → R1 48.5% → R2 49.99% → R3 50.47% → R4 56.41% (+11.91pp) 자산 라이브러리: - workflows: winner_auto 7 + partial 7 - prompts/failed: 9 패턴 (Qwen 계열 2건 추가)
2026-05-19 23:49:55 +09:00
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=38.52% / black=0.48%</div>
meeting-103 v2.1 clean R2 ✅5/5: PuLID 0.85+Redux 0.5 = pastel 49.99% (전체 best 갱신) R2 5조합 (5/5 success, retry 2회 포함): - 2-1: PuLID 0.7 + Redux 0.5 + Stage 3 Refiner (denoise 0.3) → 48.82% (Stage 3 효과 미미) - 2-2: PuLID 0.7 + Redux 0.7 → 48.93% - 2-3: PuLID 0.7 + Redux 0.3 → 38.52% (Redux 낮으면 파스텔 감소) - 2-4: PuLID 0.85 + Redux 0.5 → 49.99% ★ R2 best & 전체 best - 2-5: 한옥 spring (cherry blossom) → 44.45% / black 4.3% R1 best (R1-4 48.5%) → R2-4 (49.99%) = +1.5 pp 갱신 R9 legacy (44.5%) → R2-4 (49.99%) = +5.5 pp 누적 학습: - PuLID weight 0.85 + Redux 0.5 multiply 조합이 0.7/0.5 보다 우월 - Redux 0.3 너무 낮음, 0.7 비슷, 0.85+ 미검증 - Stage 3 Refiner 효과 미미 (48.5% → 48.82%, +0.3pp) CUDA error 회복: - 2-3, 2-4, 2-5 첫 시도 CUDA invalid argument - ComfyUI venv 재시작 후 retry → 3/3 모두 success - 매 ~5건 후 재시작 패턴 확인됨 자산 라이브러리: - models-performance: 8 모델 추적 - workflows: winner_auto 4 (R1-4, R2-1, R2-2, R2-4) + partial 5 - prompts/proven: 7 패턴 (R1, R2 합산) R3 계획: - 3-1: PuLID 0.7 + Redux 0.85 (push redux 더 강) - 3-2: PuLID 0.6 + Redux 0.7 (PuLID 약화) - 3-3: PuLID 0.7 + Redux 0.7 @ 1024x1792 (해상도 증가) - 3-4: Qwen-Image fp8 + Lightning LoRA 60갑자 (fp8 ~20G 완료, GGUF 회피) - 3-5: Pony V6 img2img from byeolyi-ref (대안 ID 방법) 검수: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html
2026-05-19 23:36:44 +09:00
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round2/2-4_thumb.png" data-original="../results/meeting-103-v2.1-clean/round2/2-4.png" alt="2-4">
<div class="label">
<div class="label-title">R2-2-4: A — PuLID 0.85 + Redux 0.5</div>
meeting-103 v2.1 clean R3+R4 ✅: PuLID 0.7+Redux 1.0 = pastel 56.41% (전체 best 갱신) R3 (4/5 success): - 3-1: PuLID 0.7 + Redux 0.85 → 50.47% (R2 best 49.99% 갱신) - 3-2: PuLID 0.6 + Redux 0.7 → 49.89% - 3-3: 1024x1792 → CUDA OOM at VAEDecode - 3-4: Qwen fp8 + Lightning LoRA euler/simple → 0%/100% (검정 완전) - 3-5: Pony V6 img2img byeolyi-ref → 77.3% (img2img 0.65 denoise, 원본 색 보존) R4 (3/5 success, 별이 새 best): - 4-1: PuLID 0.7 + Redux 1.0 (max) → 56.41% ★ 전체 best - 4-2: PuLID 0.85 + Redux 0.7 → 50.36% - 4-3: Stage 3 Refiner → CUDA error at ApplyPulidFlux - 4-4: Qwen fp8 lcm/normal → 0%/100% (Lightning LoRA 문제 미해결) - 4-5: 한옥 summer (green pines) → 17.7%/3.97% 학습: - Redux strength 0.5 → 0.7 → 0.85 → 1.0 단조 증가 (56.41% sweet spot) - PuLID 0.7 fixed, Redux 1.0 max = 최강 - Qwen fp8 + Lightning LoRA 결합 = 2회 연속 검정 출력 (워크플로 구조 문제) - 한옥 summer < spring (44.45%) — 색감 자체가 차분 진행: R9 legacy 44.5% → R1 48.5% → R2 49.99% → R3 50.47% → R4 56.41% (+11.91pp) 자산 라이브러리: - workflows: winner_auto 7 + partial 7 - prompts/failed: 9 패턴 (Qwen 계열 2건 추가)
2026-05-19 23:49:55 +09:00
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=49.99% / black=0.18%</div>
meeting-103 v2.1 clean R2 ✅5/5: PuLID 0.85+Redux 0.5 = pastel 49.99% (전체 best 갱신) R2 5조합 (5/5 success, retry 2회 포함): - 2-1: PuLID 0.7 + Redux 0.5 + Stage 3 Refiner (denoise 0.3) → 48.82% (Stage 3 효과 미미) - 2-2: PuLID 0.7 + Redux 0.7 → 48.93% - 2-3: PuLID 0.7 + Redux 0.3 → 38.52% (Redux 낮으면 파스텔 감소) - 2-4: PuLID 0.85 + Redux 0.5 → 49.99% ★ R2 best & 전체 best - 2-5: 한옥 spring (cherry blossom) → 44.45% / black 4.3% R1 best (R1-4 48.5%) → R2-4 (49.99%) = +1.5 pp 갱신 R9 legacy (44.5%) → R2-4 (49.99%) = +5.5 pp 누적 학습: - PuLID weight 0.85 + Redux 0.5 multiply 조합이 0.7/0.5 보다 우월 - Redux 0.3 너무 낮음, 0.7 비슷, 0.85+ 미검증 - Stage 3 Refiner 효과 미미 (48.5% → 48.82%, +0.3pp) CUDA error 회복: - 2-3, 2-4, 2-5 첫 시도 CUDA invalid argument - ComfyUI venv 재시작 후 retry → 3/3 모두 success - 매 ~5건 후 재시작 패턴 확인됨 자산 라이브러리: - models-performance: 8 모델 추적 - workflows: winner_auto 4 (R1-4, R2-1, R2-2, R2-4) + partial 5 - prompts/proven: 7 패턴 (R1, R2 합산) R3 계획: - 3-1: PuLID 0.7 + Redux 0.85 (push redux 더 강) - 3-2: PuLID 0.6 + Redux 0.7 (PuLID 약화) - 3-3: PuLID 0.7 + Redux 0.7 @ 1024x1792 (해상도 증가) - 3-4: Qwen-Image fp8 + Lightning LoRA 60갑자 (fp8 ~20G 완료, GGUF 회피) - 3-5: Pony V6 img2img from byeolyi-ref (대안 ID 방법) 검수: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html
2026-05-19 23:36:44 +09:00
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round2/2-5_thumb.png" data-original="../results/meeting-103-v2.1-clean/round2/2-5.png" alt="2-5">
<div class="label">
<div class="label-title">R2-2-5: B — Hanok spring variant (cherry blossom)</div>
meeting-103 v2.1 clean R3+R4 ✅: PuLID 0.7+Redux 1.0 = pastel 56.41% (전체 best 갱신) R3 (4/5 success): - 3-1: PuLID 0.7 + Redux 0.85 → 50.47% (R2 best 49.99% 갱신) - 3-2: PuLID 0.6 + Redux 0.7 → 49.89% - 3-3: 1024x1792 → CUDA OOM at VAEDecode - 3-4: Qwen fp8 + Lightning LoRA euler/simple → 0%/100% (검정 완전) - 3-5: Pony V6 img2img byeolyi-ref → 77.3% (img2img 0.65 denoise, 원본 색 보존) R4 (3/5 success, 별이 새 best): - 4-1: PuLID 0.7 + Redux 1.0 (max) → 56.41% ★ 전체 best - 4-2: PuLID 0.85 + Redux 0.7 → 50.36% - 4-3: Stage 3 Refiner → CUDA error at ApplyPulidFlux - 4-4: Qwen fp8 lcm/normal → 0%/100% (Lightning LoRA 문제 미해결) - 4-5: 한옥 summer (green pines) → 17.7%/3.97% 학습: - Redux strength 0.5 → 0.7 → 0.85 → 1.0 단조 증가 (56.41% sweet spot) - PuLID 0.7 fixed, Redux 1.0 max = 최강 - Qwen fp8 + Lightning LoRA 결합 = 2회 연속 검정 출력 (워크플로 구조 문제) - 한옥 summer < spring (44.45%) — 색감 자체가 차분 진행: R9 legacy 44.5% → R1 48.5% → R2 49.99% → R3 50.47% → R4 56.41% (+11.91pp) 자산 라이브러리: - workflows: winner_auto 7 + partial 7 - prompts/failed: 9 패턴 (Qwen 계열 2건 추가)
2026-05-19 23:49:55 +09:00
<div class="label-meta">models=flux-dev</div>
<div class="label-auto-eval">pastel=44.45% / black=4.3%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round3/3-1_thumb.png" data-original="../results/meeting-103-v2.1-clean/round3/3-1.png" alt="3-1">
<div class="label">
<div class="label-title">R3-3-1: A — PuLID 0.7 + Redux 0.85</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=50.47% / black=0.24%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round3/3-2_thumb.png" data-original="../results/meeting-103-v2.1-clean/round3/3-2.png" alt="3-2">
<div class="label">
<div class="label-title">R3-3-2: A — PuLID 0.6 + Redux 0.7</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=49.89% / black=0.22%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-error">
meeting-103 v2.1 clean R7+R8 🏆 BREAKTHROUGH: dpmpp_2m karras = pastel 82.43% R7 (3/5 success, R6-2 sweet spot 확인): - 7-1: PuLID 0.85 (default ref) → 64.45% (less than R6-2) - 7-2: PuLID 0.6 (default ref) → 65.93% - 7-3: seed 999777 → CUDA error - 7-4: Stage 3 Refiner d=0.20 → CUDA error - 7-5: Pony V6 60갑자 calligraphy → 18.09% (검정 회피 OK) R8 (5/5 success, 🏆 거대한 발견): - 8-1: seed 200000 → 65.73% (variance -3.93pp) - 8-2: seed 555555 → 64.30% - 8-3: guidance 2.5 → 62.71% (낮은 guidance = 파스텔 감소) - 8-4: guidance 5.0 → 69.99% (R6-2와 동일) - 8-5: dpmpp_2m karras → 🏆 82.43% ⭐⭐⭐⭐⭐ 전체 best (+12.77pp) ⭐ 핵심 발견 2: 샘플러 변경의 dramatic effect euler/simple (R6-2 69.66%) → dpmpp_2m/karras (R8-5 82.43%) = +12.77pp 누적 진행 (R9 legacy → R8): 44.5% → 48.5% → 49.99% → 50.47% → 56.41% → 69.66% → 82.43% = +37.93pp 자산 라이브러리: - workflows: winner_auto 11 + winner_auto_high 4 + winner_auto_breakthrough 1 - 8-5 워크플로 = 타로/12지신/별이 시리즈 재활용 표준 베이스
2026-05-20 00:15:56 +09:00
<div class="no-image">ERROR</div>
meeting-103 v2.1 clean R3+R4 ✅: PuLID 0.7+Redux 1.0 = pastel 56.41% (전체 best 갱신) R3 (4/5 success): - 3-1: PuLID 0.7 + Redux 0.85 → 50.47% (R2 best 49.99% 갱신) - 3-2: PuLID 0.6 + Redux 0.7 → 49.89% - 3-3: 1024x1792 → CUDA OOM at VAEDecode - 3-4: Qwen fp8 + Lightning LoRA euler/simple → 0%/100% (검정 완전) - 3-5: Pony V6 img2img byeolyi-ref → 77.3% (img2img 0.65 denoise, 원본 색 보존) R4 (3/5 success, 별이 새 best): - 4-1: PuLID 0.7 + Redux 1.0 (max) → 56.41% ★ 전체 best - 4-2: PuLID 0.85 + Redux 0.7 → 50.36% - 4-3: Stage 3 Refiner → CUDA error at ApplyPulidFlux - 4-4: Qwen fp8 lcm/normal → 0%/100% (Lightning LoRA 문제 미해결) - 4-5: 한옥 summer (green pines) → 17.7%/3.97% 학습: - Redux strength 0.5 → 0.7 → 0.85 → 1.0 단조 증가 (56.41% sweet spot) - PuLID 0.7 fixed, Redux 1.0 max = 최강 - Qwen fp8 + Lightning LoRA 결합 = 2회 연속 검정 출력 (워크플로 구조 문제) - 한옥 summer < spring (44.45%) — 색감 자체가 차분 진행: R9 legacy 44.5% → R1 48.5% → R2 49.99% → R3 50.47% → R4 56.41% (+11.91pp) 자산 라이브러리: - workflows: winner_auto 7 + partial 7 - prompts/failed: 9 패턴 (Qwen 계열 2건 추가)
2026-05-19 23:49:55 +09:00
<div class="label">
<div class="label-title">R3-3-3: A — PuLID 0.7 + Redux 0.7 @ 1024x1792</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
meeting-103 v2.1 clean R7+R8 🏆 BREAKTHROUGH: dpmpp_2m karras = pastel 82.43% R7 (3/5 success, R6-2 sweet spot 확인): - 7-1: PuLID 0.85 (default ref) → 64.45% (less than R6-2) - 7-2: PuLID 0.6 (default ref) → 65.93% - 7-3: seed 999777 → CUDA error - 7-4: Stage 3 Refiner d=0.20 → CUDA error - 7-5: Pony V6 60갑자 calligraphy → 18.09% (검정 회피 OK) R8 (5/5 success, 🏆 거대한 발견): - 8-1: seed 200000 → 65.73% (variance -3.93pp) - 8-2: seed 555555 → 64.30% - 8-3: guidance 2.5 → 62.71% (낮은 guidance = 파스텔 감소) - 8-4: guidance 5.0 → 69.99% (R6-2와 동일) - 8-5: dpmpp_2m karras → 🏆 82.43% ⭐⭐⭐⭐⭐ 전체 best (+12.77pp) ⭐ 핵심 발견 2: 샘플러 변경의 dramatic effect euler/simple (R6-2 69.66%) → dpmpp_2m/karras (R8-5 82.43%) = +12.77pp 누적 진행 (R9 legacy → R8): 44.5% → 48.5% → 49.99% → 50.47% → 56.41% → 69.66% → 82.43% = +37.93pp 자산 라이브러리: - workflows: winner_auto 11 + winner_auto_high 4 + winner_auto_breakthrough 1 - 8-5 워크플로 = 타로/12지신/별이 시리즈 재활용 표준 베이스
2026-05-20 00:15:56 +09:00
<div class="label-auto-eval">err: VAEDecode</div>
meeting-103 v2.1 clean R3+R4 ✅: PuLID 0.7+Redux 1.0 = pastel 56.41% (전체 best 갱신) R3 (4/5 success): - 3-1: PuLID 0.7 + Redux 0.85 → 50.47% (R2 best 49.99% 갱신) - 3-2: PuLID 0.6 + Redux 0.7 → 49.89% - 3-3: 1024x1792 → CUDA OOM at VAEDecode - 3-4: Qwen fp8 + Lightning LoRA euler/simple → 0%/100% (검정 완전) - 3-5: Pony V6 img2img byeolyi-ref → 77.3% (img2img 0.65 denoise, 원본 색 보존) R4 (3/5 success, 별이 새 best): - 4-1: PuLID 0.7 + Redux 1.0 (max) → 56.41% ★ 전체 best - 4-2: PuLID 0.85 + Redux 0.7 → 50.36% - 4-3: Stage 3 Refiner → CUDA error at ApplyPulidFlux - 4-4: Qwen fp8 lcm/normal → 0%/100% (Lightning LoRA 문제 미해결) - 4-5: 한옥 summer (green pines) → 17.7%/3.97% 학습: - Redux strength 0.5 → 0.7 → 0.85 → 1.0 단조 증가 (56.41% sweet spot) - PuLID 0.7 fixed, Redux 1.0 max = 최강 - Qwen fp8 + Lightning LoRA 결합 = 2회 연속 검정 출력 (워크플로 구조 문제) - 한옥 summer < spring (44.45%) — 색감 자체가 차분 진행: R9 legacy 44.5% → R1 48.5% → R2 49.99% → R3 50.47% → R4 56.41% (+11.91pp) 자산 라이브러리: - workflows: winner_auto 7 + partial 7 - prompts/failed: 9 패턴 (Qwen 계열 2건 추가)
2026-05-19 23:49:55 +09:00
<div class="label-flags"><span class="flag flag-warn">status-error</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round3/3-4_thumb.png" data-original="../results/meeting-103-v2.1-clean/round3/3-4.png" alt="3-4">
<div class="label">
<div class="label-title">R3-3-4: C — 60갑자 Qwen-Image fp8 + Lightning LoRA 4-step</div>
<div class="label-meta">models=qwen-image-2.0, lightning-lora-qwen</div>
<div class="label-auto-eval">pastel=0.0% / black=100.0%</div>
<div class="label-flags"><span class="flag flag-warn">black-suspicious</span><span class="flag flag-warn">pastel-low</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round3/3-5_thumb.png" data-original="../results/meeting-103-v2.1-clean/round3/3-5.png" alt="3-5">
<div class="label">
<div class="label-title">R3-3-5: A — Pony V6 img2img from byeolyi-ref (denoise 0.65)</div>
<div class="label-meta">models=pony-v6</div>
<div class="label-auto-eval">pastel=77.3% / black=0.03%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round4/4-1_thumb.png" data-original="../results/meeting-103-v2.1-clean/round4/4-1.png" alt="4-1">
<div class="label">
<div class="label-title">R4-4-1: A — PuLID 0.7 + Redux 1.0 (max)</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=56.41% / black=0.2%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round4/4-2_thumb.png" data-original="../results/meeting-103-v2.1-clean/round4/4-2.png" alt="4-2">
<div class="label">
<div class="label-title">R4-4-2: A — PuLID 0.85 + Redux 0.7 (combine best)</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=50.36% / black=0.86%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-error">
meeting-103 v2.1 clean R7+R8 🏆 BREAKTHROUGH: dpmpp_2m karras = pastel 82.43% R7 (3/5 success, R6-2 sweet spot 확인): - 7-1: PuLID 0.85 (default ref) → 64.45% (less than R6-2) - 7-2: PuLID 0.6 (default ref) → 65.93% - 7-3: seed 999777 → CUDA error - 7-4: Stage 3 Refiner d=0.20 → CUDA error - 7-5: Pony V6 60갑자 calligraphy → 18.09% (검정 회피 OK) R8 (5/5 success, 🏆 거대한 발견): - 8-1: seed 200000 → 65.73% (variance -3.93pp) - 8-2: seed 555555 → 64.30% - 8-3: guidance 2.5 → 62.71% (낮은 guidance = 파스텔 감소) - 8-4: guidance 5.0 → 69.99% (R6-2와 동일) - 8-5: dpmpp_2m karras → 🏆 82.43% ⭐⭐⭐⭐⭐ 전체 best (+12.77pp) ⭐ 핵심 발견 2: 샘플러 변경의 dramatic effect euler/simple (R6-2 69.66%) → dpmpp_2m/karras (R8-5 82.43%) = +12.77pp 누적 진행 (R9 legacy → R8): 44.5% → 48.5% → 49.99% → 50.47% → 56.41% → 69.66% → 82.43% = +37.93pp 자산 라이브러리: - workflows: winner_auto 11 + winner_auto_high 4 + winner_auto_breakthrough 1 - 8-5 워크플로 = 타로/12지신/별이 시리즈 재활용 표준 베이스
2026-05-20 00:15:56 +09:00
<div class="no-image">ERROR</div>
meeting-103 v2.1 clean R3+R4 ✅: PuLID 0.7+Redux 1.0 = pastel 56.41% (전체 best 갱신) R3 (4/5 success): - 3-1: PuLID 0.7 + Redux 0.85 → 50.47% (R2 best 49.99% 갱신) - 3-2: PuLID 0.6 + Redux 0.7 → 49.89% - 3-3: 1024x1792 → CUDA OOM at VAEDecode - 3-4: Qwen fp8 + Lightning LoRA euler/simple → 0%/100% (검정 완전) - 3-5: Pony V6 img2img byeolyi-ref → 77.3% (img2img 0.65 denoise, 원본 색 보존) R4 (3/5 success, 별이 새 best): - 4-1: PuLID 0.7 + Redux 1.0 (max) → 56.41% ★ 전체 best - 4-2: PuLID 0.85 + Redux 0.7 → 50.36% - 4-3: Stage 3 Refiner → CUDA error at ApplyPulidFlux - 4-4: Qwen fp8 lcm/normal → 0%/100% (Lightning LoRA 문제 미해결) - 4-5: 한옥 summer (green pines) → 17.7%/3.97% 학습: - Redux strength 0.5 → 0.7 → 0.85 → 1.0 단조 증가 (56.41% sweet spot) - PuLID 0.7 fixed, Redux 1.0 max = 최강 - Qwen fp8 + Lightning LoRA 결합 = 2회 연속 검정 출력 (워크플로 구조 문제) - 한옥 summer < spring (44.45%) — 색감 자체가 차분 진행: R9 legacy 44.5% → R1 48.5% → R2 49.99% → R3 50.47% → R4 56.41% (+11.91pp) 자산 라이브러리: - workflows: winner_auto 7 + partial 7 - prompts/failed: 9 패턴 (Qwen 계열 2건 추가)
2026-05-19 23:49:55 +09:00
<div class="label">
<div class="label-title">R4-4-3: A — PuLID 0.7 + Redux 0.85 + Stage 3 Refiner</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux, flux-dev-refiner</div>
meeting-103 v2.1 clean R7+R8 🏆 BREAKTHROUGH: dpmpp_2m karras = pastel 82.43% R7 (3/5 success, R6-2 sweet spot 확인): - 7-1: PuLID 0.85 (default ref) → 64.45% (less than R6-2) - 7-2: PuLID 0.6 (default ref) → 65.93% - 7-3: seed 999777 → CUDA error - 7-4: Stage 3 Refiner d=0.20 → CUDA error - 7-5: Pony V6 60갑자 calligraphy → 18.09% (검정 회피 OK) R8 (5/5 success, 🏆 거대한 발견): - 8-1: seed 200000 → 65.73% (variance -3.93pp) - 8-2: seed 555555 → 64.30% - 8-3: guidance 2.5 → 62.71% (낮은 guidance = 파스텔 감소) - 8-4: guidance 5.0 → 69.99% (R6-2와 동일) - 8-5: dpmpp_2m karras → 🏆 82.43% ⭐⭐⭐⭐⭐ 전체 best (+12.77pp) ⭐ 핵심 발견 2: 샘플러 변경의 dramatic effect euler/simple (R6-2 69.66%) → dpmpp_2m/karras (R8-5 82.43%) = +12.77pp 누적 진행 (R9 legacy → R8): 44.5% → 48.5% → 49.99% → 50.47% → 56.41% → 69.66% → 82.43% = +37.93pp 자산 라이브러리: - workflows: winner_auto 11 + winner_auto_high 4 + winner_auto_breakthrough 1 - 8-5 워크플로 = 타로/12지신/별이 시리즈 재활용 표준 베이스
2026-05-20 00:15:56 +09:00
<div class="label-auto-eval">err: ApplyPulidFlux</div>
meeting-103 v2.1 clean R3+R4 ✅: PuLID 0.7+Redux 1.0 = pastel 56.41% (전체 best 갱신) R3 (4/5 success): - 3-1: PuLID 0.7 + Redux 0.85 → 50.47% (R2 best 49.99% 갱신) - 3-2: PuLID 0.6 + Redux 0.7 → 49.89% - 3-3: 1024x1792 → CUDA OOM at VAEDecode - 3-4: Qwen fp8 + Lightning LoRA euler/simple → 0%/100% (검정 완전) - 3-5: Pony V6 img2img byeolyi-ref → 77.3% (img2img 0.65 denoise, 원본 색 보존) R4 (3/5 success, 별이 새 best): - 4-1: PuLID 0.7 + Redux 1.0 (max) → 56.41% ★ 전체 best - 4-2: PuLID 0.85 + Redux 0.7 → 50.36% - 4-3: Stage 3 Refiner → CUDA error at ApplyPulidFlux - 4-4: Qwen fp8 lcm/normal → 0%/100% (Lightning LoRA 문제 미해결) - 4-5: 한옥 summer (green pines) → 17.7%/3.97% 학습: - Redux strength 0.5 → 0.7 → 0.85 → 1.0 단조 증가 (56.41% sweet spot) - PuLID 0.7 fixed, Redux 1.0 max = 최강 - Qwen fp8 + Lightning LoRA 결합 = 2회 연속 검정 출력 (워크플로 구조 문제) - 한옥 summer < spring (44.45%) — 색감 자체가 차분 진행: R9 legacy 44.5% → R1 48.5% → R2 49.99% → R3 50.47% → R4 56.41% (+11.91pp) 자산 라이브러리: - workflows: winner_auto 7 + partial 7 - prompts/failed: 9 패턴 (Qwen 계열 2건 추가)
2026-05-19 23:49:55 +09:00
<div class="label-flags"><span class="flag flag-warn">status-error</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round4/4-4_thumb.png" data-original="../results/meeting-103-v2.1-clean/round4/4-4.png" alt="4-4">
<div class="label">
<div class="label-title">R4-4-4: C — 60갑자 Qwen-Image fp8 + Lightning LoRA (lcm/normal cfg=1)</div>
<div class="label-meta">models=qwen-image-2.0, lightning-lora-qwen</div>
<div class="label-auto-eval">pastel=0.0% / black=100.0%</div>
<div class="label-flags"><span class="flag flag-warn">black-suspicious</span><span class="flag flag-warn">pastel-low</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round4/4-5_thumb.png" data-original="../results/meeting-103-v2.1-clean/round4/4-5.png" alt="4-5">
<div class="label">
<div class="label-title">R4-4-5: B — Hanok summer (green pines, blue sky)</div>
<div class="label-meta">models=flux-dev</div>
<div class="label-auto-eval">pastel=17.72% / black=3.97%</div>
meeting-103 v2.1 clean R2 ✅5/5: PuLID 0.85+Redux 0.5 = pastel 49.99% (전체 best 갱신) R2 5조합 (5/5 success, retry 2회 포함): - 2-1: PuLID 0.7 + Redux 0.5 + Stage 3 Refiner (denoise 0.3) → 48.82% (Stage 3 효과 미미) - 2-2: PuLID 0.7 + Redux 0.7 → 48.93% - 2-3: PuLID 0.7 + Redux 0.3 → 38.52% (Redux 낮으면 파스텔 감소) - 2-4: PuLID 0.85 + Redux 0.5 → 49.99% ★ R2 best & 전체 best - 2-5: 한옥 spring (cherry blossom) → 44.45% / black 4.3% R1 best (R1-4 48.5%) → R2-4 (49.99%) = +1.5 pp 갱신 R9 legacy (44.5%) → R2-4 (49.99%) = +5.5 pp 누적 학습: - PuLID weight 0.85 + Redux 0.5 multiply 조합이 0.7/0.5 보다 우월 - Redux 0.3 너무 낮음, 0.7 비슷, 0.85+ 미검증 - Stage 3 Refiner 효과 미미 (48.5% → 48.82%, +0.3pp) CUDA error 회복: - 2-3, 2-4, 2-5 첫 시도 CUDA invalid argument - ComfyUI venv 재시작 후 retry → 3/3 모두 success - 매 ~5건 후 재시작 패턴 확인됨 자산 라이브러리: - models-performance: 8 모델 추적 - workflows: winner_auto 4 (R1-4, R2-1, R2-2, R2-4) + partial 5 - prompts/proven: 7 패턴 (R1, R2 합산) R3 계획: - 3-1: PuLID 0.7 + Redux 0.85 (push redux 더 강) - 3-2: PuLID 0.6 + Redux 0.7 (PuLID 약화) - 3-3: PuLID 0.7 + Redux 0.7 @ 1024x1792 (해상도 증가) - 3-4: Qwen-Image fp8 + Lightning LoRA 60갑자 (fp8 ~20G 완료, GGUF 회피) - 3-5: Pony V6 img2img from byeolyi-ref (대안 ID 방법) 검수: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html
2026-05-19 23:36:44 +09:00
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
meeting-103 v2.1 clean R1 ✅5/5: PuLID + Redux = pastel 48.5% (R9 best 갱신) R0 환경 점검: - VRAM 11.7/12.4GB free (94%) — 해소 확인 - pytorch 2.12.0+cu130 + sageattention venv 검증 - 모델 인벤토리: Flux Dev Q5 GGUF, PuLID, Redux, byeolyi-ref-916 확보 - fp8 다운로드 .part 0 bytes (재개 대기) R1 clean 5조합 (5/5 success): - 1-1: PuLID w0.85 mean → pastel 25.7% / black 1.95% - 1-2: PuLID w0.85 mean delayed_start 0.05 → pastel 10.0% / black 14.0% (어두움) - 1-3: PuLID w1.0 mean → pastel 27.5% / black 0.79% - 1-4: PuLID w0.7 + Redux 0.5 → pastel 48.5% / black 0.27% ★ R1 best - 1-5: 한옥 baseline Flux Dev 768² → pastel 12.9% / black 11.7% CUDA error 회복: - 1-2 retry concat → CUDA invalid argument → 후속 1-2~1-5 모두 실패 - ComfyUI venv 재시작 (sage + pytorch 2.12) → 1-2 변형 (delayed_start)로 대체 - 재시도 4/4 모두 success 이전 R9 best (44.5%) → R1-4 (48.5%) = +4 pp 갱신 자산 라이브러리 누적: - models-performance.json: 7 모델 추적 (flux-dev, pulid-flux, flux-redux 등) - workflows: winner_auto 1 + partial 2 - prompts/proven: illustration_auto 패턴 3건 - prompts/failed: pastel-low + CUDA error 패턴 - prompts/negative: A_byeolyi 패턴 etc. 검수: - Vault: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html - Gitea: results/meeting-103-v2.1-clean/round1/
2026-05-19 23:17:19 +09:00
</div>
</div>
meeting-103 v2.1 clean R7+R8 🏆 BREAKTHROUGH: dpmpp_2m karras = pastel 82.43% R7 (3/5 success, R6-2 sweet spot 확인): - 7-1: PuLID 0.85 (default ref) → 64.45% (less than R6-2) - 7-2: PuLID 0.6 (default ref) → 65.93% - 7-3: seed 999777 → CUDA error - 7-4: Stage 3 Refiner d=0.20 → CUDA error - 7-5: Pony V6 60갑자 calligraphy → 18.09% (검정 회피 OK) R8 (5/5 success, 🏆 거대한 발견): - 8-1: seed 200000 → 65.73% (variance -3.93pp) - 8-2: seed 555555 → 64.30% - 8-3: guidance 2.5 → 62.71% (낮은 guidance = 파스텔 감소) - 8-4: guidance 5.0 → 69.99% (R6-2와 동일) - 8-5: dpmpp_2m karras → 🏆 82.43% ⭐⭐⭐⭐⭐ 전체 best (+12.77pp) ⭐ 핵심 발견 2: 샘플러 변경의 dramatic effect euler/simple (R6-2 69.66%) → dpmpp_2m/karras (R8-5 82.43%) = +12.77pp 누적 진행 (R9 legacy → R8): 44.5% → 48.5% → 49.99% → 50.47% → 56.41% → 69.66% → 82.43% = +37.93pp 자산 라이브러리: - workflows: winner_auto 11 + winner_auto_high 4 + winner_auto_breakthrough 1 - 8-5 워크플로 = 타로/12지신/별이 시리즈 재활용 표준 베이스
2026-05-20 00:15:56 +09:00
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round5/5-1_thumb.png" data-original="../results/meeting-103-v2.1-clean/round5/5-1.png" alt="5-1">
<div class="label">
<div class="label-title">R5-5-1: A — PuLID 0.7 + Redux 1.0 + Stage 3 Refiner (denoise 0.25)</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux, flux-dev-refiner</div>
<div class="label-auto-eval">pastel=49.85% / black=0.08%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-error">
<div class="no-image">ERROR</div>
<div class="label">
<div class="label-title">R5-5-2: A — PuLID 0.5 + Redux 1.0</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">err: ApplyPulidFlux</div>
<div class="label-flags"><span class="flag flag-warn">status-error</span></div>
</div>
</div>
<div class="cell status-error">
<div class="no-image">ERROR</div>
<div class="label">
<div class="label-title">R5-5-3: A — PuLID 0.7 + Redux 1.0 strength_type=attn_bias</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">err: ApplyPulidFlux</div>
<div class="label-flags"><span class="flag flag-warn">status-error</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round5/5-4_thumb.png" data-original="../results/meeting-103-v2.1-clean/round5/5-4.png" alt="5-4">
<div class="label">
<div class="label-title">R5-5-4: C — 60갑자 Qwen fp8 full 25-step euler/normal cfg=2.5 (no Lightning LoRA)</div>
<div class="label-meta">models=qwen-image-2.0</div>
<div class="label-auto-eval">pastel=0.0% / black=100.0%</div>
<div class="label-flags"><span class="flag flag-warn">black-suspicious</span><span class="flag flag-warn">pastel-low</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round5/5-5_thumb.png" data-original="../results/meeting-103-v2.1-clean/round5/5-5.png" alt="5-5">
<div class="label">
<div class="label-title">R5-5-5: B — Hanok winter (snow)</div>
<div class="label-meta">models=flux-dev</div>
<div class="label-auto-eval">pastel=48.65% / black=3.7%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round6/6-1_thumb.png" data-original="../results/meeting-103-v2.1-clean/round6/6-1.png" alt="6-1">
<div class="label">
<div class="label-title">R6-6-1: A — PuLID 0.7 + Redux 1.0 (ref byeolyi-916.png)</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=52.99% / black=0.5%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round6/6-2_thumb.png" data-original="../results/meeting-103-v2.1-clean/round6/6-2.png" alt="6-2">
<div class="label">
<div class="label-title">R6-6-2: A — PuLID 0.7 + Redux 1.0 (ref byeolyi-default.png)</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=69.66% / black=0.05%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-error">
<div class="no-image">ERROR</div>
<div class="label">
<div class="label-title">R6-6-3: A — PuLID 0.7 + multi-Redux (ref 1344+default)</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">err: ApplyPulidFlux</div>
<div class="label-flags"><span class="flag flag-warn">status-error</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round6/6-4_thumb.png" data-original="../results/meeting-103-v2.1-clean/round6/6-4.png" alt="6-4">
<div class="label">
<div class="label-title">R6-6-4: C — 60갑자 Qwen fp8 + ModelSamplingAuraFlow shift=3.0</div>
<div class="label-meta">models=qwen-image-2.0</div>
<div class="label-auto-eval">pastel=0.0% / black=100.0%</div>
<div class="label-flags"><span class="flag flag-warn">black-suspicious</span><span class="flag flag-warn">pastel-low</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round6/6-5_thumb.png" data-original="../results/meeting-103-v2.1-clean/round6/6-5.png" alt="6-5">
<div class="label">
<div class="label-title">R6-6-5: B — Hanok dancheong (red blue green eaves)</div>
<div class="label-meta">models=flux-dev</div>
<div class="label-auto-eval">pastel=15.09% / black=12.46%</div>
<div class="label-flags"><span class="flag flag-warn">black-suspicious</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round7/7-1_thumb.png" data-original="../results/meeting-103-v2.1-clean/round7/7-1.png" alt="7-1">
<div class="label">
<div class="label-title">R7-7-1: A — PuLID 0.85 + Redux 1.0 (default ref)</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=64.45% / black=0.01%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round7/7-2_thumb.png" data-original="../results/meeting-103-v2.1-clean/round7/7-2.png" alt="7-2">
<div class="label">
<div class="label-title">R7-7-2: A — PuLID 0.6 + Redux 1.0 (default ref)</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=65.93% / black=0.19%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-error">
<div class="no-image">ERROR</div>
<div class="label">
<div class="label-title">R7-7-3: A — PuLID 0.7 + Redux 1.0 (default ref, seed 999777)</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">err: ApplyPulidFlux</div>
<div class="label-flags"><span class="flag flag-warn">status-error</span></div>
</div>
</div>
<div class="cell status-error">
<div class="no-image">ERROR</div>
<div class="label">
<div class="label-title">R7-7-4: A — PuLID 0.7 + Redux 1.0 + Stage 3 (denoise 0.20)</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux, flux-dev-refiner</div>
<div class="label-auto-eval">err: ApplyPulidFlux</div>
<div class="label-flags"><span class="flag flag-warn">status-error</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round7/7-5_thumb.png" data-original="../results/meeting-103-v2.1-clean/round7/7-5.png" alt="7-5">
<div class="label">
<div class="label-title">R7-7-5: C — 60갑자 Pony V6 calligraphy prompt</div>
<div class="label-meta">models=pony-v6</div>
<div class="label-auto-eval">pastel=18.09% / black=0.02%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round8/8-1_thumb.png" data-original="../results/meeting-103-v2.1-clean/round8/8-1.png" alt="8-1">
<div class="label">
<div class="label-title">R8-8-1: A — R6-2 baseline seed 200000</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=65.73% / black=0.03%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round8/8-2_thumb.png" data-original="../results/meeting-103-v2.1-clean/round8/8-2.png" alt="8-2">
<div class="label">
<div class="label-title">R8-8-2: A — R6-2 baseline seed 555555</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=64.3% / black=0.05%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round8/8-3_thumb.png" data-original="../results/meeting-103-v2.1-clean/round8/8-3.png" alt="8-3">
<div class="label">
<div class="label-title">R8-8-3: A — R6-2 baseline + guidance 2.5</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=62.71% / black=0.0%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round8/8-4_thumb.png" data-original="../results/meeting-103-v2.1-clean/round8/8-4.png" alt="8-4">
<div class="label">
<div class="label-title">R8-8-4: A — R6-2 baseline + guidance 5.0</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=69.99% / black=0.47%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round8/8-5_thumb.png" data-original="../results/meeting-103-v2.1-clean/round8/8-5.png" alt="8-5">
<div class="label">
<div class="label-title">R8-8-5 🏆: A — R6-2 baseline + dpmpp_2m karras</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=82.43% / black=0.0%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
🏆🏆🏆 meeting-103 v2.1 clean R9+R10 FINAL: uni_pc karras = pastel 88.09% R9 (2/5 success — sampler 추가 탐색): - 9-1: dpmpp_3m_sde karras → 81.95% - 9-2: uni_pc karras → 85.33% ⭐ R9 best (+2.9pp vs R8-5) - 9-3,9-4,9-5: CUDA error R10 (4/5 success — 최종 push): - 10-1: uni_pc karras seed 800000 → 🏆🏆🏆 88.09% ⭐ FINAL BEST - 10-2: uni_pc + PuLID 0.8 → 87.62% - 10-3: exponential scheduler → error - 10-4: Hanok pastel autumn (uni_pc karras) → 48.53% ⭐ 한옥 best - 10-5: Pony 60갑자 uni_pc → 9.02% (low) 🏆 FINAL BEST 워크플로 (R10-1, pastel 88.09%): - Base: flux1-dev-Q5_K_S.gguf (Flux Dev Q5 GGUF) - PuLID: pulid_flux_v0.9.1 / weight 0.7 / fusion mean / ref byeolyi-default.png - Redux: flux1-redux-dev / strength 1.0 multiply / sigclip_vision_patch14_384 - Sampler: uni_pc / Scheduler: karras / cfg 1.0 / steps 25 / guidance 3.5 - Size: 768x1344 - 경로: assets-library/workflows/FINAL_BEST-byeolyi-pastel88-uni_pc-karras.json 전체 진행: R9 legacy 44.5% → R10-1 88.09% = +43.59pp 누적 핵심 발견 (시간순): 1. R6: byeolyi-default.png ref 발견 (+13pp) 2. R8: dpmpp_2m karras 샘플러 (+13pp) 3. R9: uni_pc karras (+3pp) 4. R10: seed 800000 = 88.09% peak 60갑자 Qwen fp8 트랙 결론: - 5회 시도 모두 black 100% (R3 euler/simple, R4 lcm, R5 25step, R6 ModelSamplingAuraFlow, R7 Pony 18%, R10-5 Pony 9%) - 워크플로 구조 결함 확정 → 다음 자산 R&D 에서 Qwen-Image native (non-distill) 모델 다운로드 필요 다음 자산 R&D 추천: 1. 타로 78장 (K-타로) — FINAL_BEST 워크플로 직접 재활용, byeolyi 캐릭터 카드 등장 2. 별이 일러스트 시리즈 (다양한 자세/배경) — FINAL_BEST 워크플로 그대로 3. 12지신 캐릭터 (별이 친구들) — FINAL_BEST 워크플로 + 12지신 prompt 4. 60갑자 (재시작 필요): Qwen-Image native or Pony LoRA 학습
2026-05-20 00:30:37 +09:00
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round9/9-1_thumb.png" data-original="../results/meeting-103-v2.1-clean/round9/9-1.png" alt="9-1">
<div class="label">
<div class="label-title">R9-9-1 🏆: A — dpmpp_3m_sde + karras</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=81.95% / black=0.0%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round9/9-2_thumb.png" data-original="../results/meeting-103-v2.1-clean/round9/9-2.png" alt="9-2">
<div class="label">
<div class="label-title">R9-9-2 🏆: A — uni_pc + karras</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=85.33% / black=0.0%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-error">
<div class="no-image">ERROR</div>
<div class="label">
<div class="label-title">R9-9-3: A — dpmpp_2m + sgm_uniform</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">err: KSampler</div>
<div class="label-flags"><span class="flag flag-warn">status-error</span></div>
</div>
</div>
<div class="cell status-error">
<div class="no-image">ERROR</div>
<div class="label">
<div class="label-title">R9-9-4: A — R8-5 best + seed 777777</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">err: VAEDecode</div>
<div class="label-flags"><span class="flag flag-warn">status-error</span></div>
</div>
</div>
<div class="cell status-error">
<div class="no-image">ERROR</div>
<div class="label">
<div class="label-title">R9-9-5: A — R8-5 best + 35 steps</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">err: VAEDecode</div>
<div class="label-flags"><span class="flag flag-warn">status-error</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round10/10-1_thumb.png" data-original="../results/meeting-103-v2.1-clean/round10/10-1.png" alt="10-1">
<div class="label">
<div class="label-title">R10-10-1 🏆🏆🏆: A — uni_pc karras seed 800000</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=88.09% / black=0.0%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round10/10-2_thumb.png" data-original="../results/meeting-103-v2.1-clean/round10/10-2.png" alt="10-2">
<div class="label">
<div class="label-title">R10-10-2 🏆: A — uni_pc karras + PuLID 0.8</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">pastel=87.62% / black=0.0%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-error">
<div class="no-image">ERROR</div>
<div class="label">
<div class="label-title">R10-10-3: A — uni_pc + exponential scheduler</div>
<div class="label-meta">models=flux-dev, pulid-flux, flux-redux</div>
<div class="label-auto-eval">err: ApplyPulidFlux</div>
<div class="label-flags"><span class="flag flag-warn">status-error</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round10/10-4_thumb.png" data-original="../results/meeting-103-v2.1-clean/round10/10-4.png" alt="10-4">
<div class="label">
<div class="label-title">R10-10-4: B — Hanok pastel autumn (uni_pc karras)</div>
<div class="label-meta">models=flux-dev</div>
<div class="label-auto-eval">pastel=48.53% / black=0.0%</div>
<div class="label-flags"><span class="flag flag-ok">clean</span></div>
</div>
</div>
<div class="cell status-success">
<img src="../results/meeting-103-v2.1-clean/round10/10-5_thumb.png" data-original="../results/meeting-103-v2.1-clean/round10/10-5.png" alt="10-5">
<div class="label">
<div class="label-title">R10-10-5: C — 60갑자 Pony V6 uni_pc karras</div>
<div class="label-meta">models=pony-v6</div>
<div class="label-auto-eval">pastel=9.02% / black=0.01%</div>
<div class="label-flags"><span class="flag flag-warn">pastel-low</span></div>
</div>
</div>
meeting-103 v2.1 clean R1 ✅5/5: PuLID + Redux = pastel 48.5% (R9 best 갱신) R0 환경 점검: - VRAM 11.7/12.4GB free (94%) — 해소 확인 - pytorch 2.12.0+cu130 + sageattention venv 검증 - 모델 인벤토리: Flux Dev Q5 GGUF, PuLID, Redux, byeolyi-ref-916 확보 - fp8 다운로드 .part 0 bytes (재개 대기) R1 clean 5조합 (5/5 success): - 1-1: PuLID w0.85 mean → pastel 25.7% / black 1.95% - 1-2: PuLID w0.85 mean delayed_start 0.05 → pastel 10.0% / black 14.0% (어두움) - 1-3: PuLID w1.0 mean → pastel 27.5% / black 0.79% - 1-4: PuLID w0.7 + Redux 0.5 → pastel 48.5% / black 0.27% ★ R1 best - 1-5: 한옥 baseline Flux Dev 768² → pastel 12.9% / black 11.7% CUDA error 회복: - 1-2 retry concat → CUDA invalid argument → 후속 1-2~1-5 모두 실패 - ComfyUI venv 재시작 (sage + pytorch 2.12) → 1-2 변형 (delayed_start)로 대체 - 재시도 4/4 모두 success 이전 R9 best (44.5%) → R1-4 (48.5%) = +4 pp 갱신 자산 라이브러리 누적: - models-performance.json: 7 모델 추적 (flux-dev, pulid-flux, flux-redux 등) - workflows: winner_auto 1 + partial 2 - prompts/proven: illustration_auto 패턴 3건 - prompts/failed: pastel-low + CUDA error 패턴 - prompts/negative: A_byeolyi 패턴 etc. 검수: - Vault: D:\Vault\8460s-image-rd\eval\meeting-103-v2.1-clean-grid.html - Gitea: results/meeting-103-v2.1-clean/round1/
2026-05-19 23:17:19 +09:00
</div>
</body></html>
Reference in New Issue Copy Permalink