← Run Index
Run summary · 2026-08 1× 工程表 v2 — S3 (iteration pipelining depth 2) gate
s3-12500-off-prof
S3 gate arm A (pipelining = off): Lego 12.5k full run (profile=true、Layer 1 busy% / sync-tax 取得用)
partial pass
01
設定
- dataset
- /Users/otkrickey/dev/3dgs-workspace/datasets/nerf_synthetic/lego
- gt_convention
- Premultiplied
- iterations
- 12,500
- seed
- 42
- capacity
- 1,000,000 splats
- sh_degree
- 3
- loss
- L1Ssim
- lambda
- 0.200
- ssim
- window=7 sigma=1
- backend.backward
- Simd
- backend.loss_path
- Gpu
- backend.emit_pairs
- Original
- backend.forward
- F32
- backend.dispatch
- PerStage
- backend.radix_scan
- GpuHierarchical
- backend.binning
- Resident
- backend.tile_sort
- GlobalRadix
- backend.submission
- Chained
- backend.target_cache
- true
- backend.shader_fast_math
- true
- os_build
- 26.5.2 (25F84)
02
結果
Val PSNR
36.13dB
Wallclock
3m 0s
Final splats
407,489
Final loss
5.482269e-3
03
3D ビューワー
学習結果 PLY を .splat 形式に変換し、antimatter15/splat の WebGL ビューワーで対話表示。
04
推移
Loss curve — 反復ごとの training loss
Splat count — refine の発火と最終構成
05
中間ログ抜粋
| iter | 本 run loss |
|---|---|
| 1 | 5.9385e-1 |
| 1,000 | 5.4399e-2 |
| 2,000 | 1.3985e-2 |
| 3,000 | 8.3129e-3 |
| 4,000 | 7.4180e-3 |
| 5,000 | 6.6694e-3 |
| 6,000 | 6.2403e-3 |
| 7,000 | 5.9627e-3 |
| 8,000 | 5.8514e-3 |
| 9,000 | 5.7891e-3 |
| 10,000 | 5.6778e-3 |
| 11,000 | 5.6432e-3 |
| 12,000 | 5.5613e-3 |
| 12,500 | 5.4823e-3 |
06
関連 finding
この run を related_runs に含む finding (Positive / Negative どちらも)。