← Run Index
Run summary · 2026-09-seminar05-milestones M1 (C1 per-kernel 再計測 + C2 残差 bubble 帰属)
c1-12500-off-prof
M1 C1: per-kernel gpu_ms 再計測 (submission=per_kernel / forward_submission=per_stage / pipelining=off、post-S2/S3 kernel)
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
- PerKernel
- backend.target_cache
- true
- backend.shader_fast_math
- true
- os_build
- 26.6.2 (25G83)
02
結果
Val PSNR
36.09dB
Wallclock
3m 25s
Final splats
408,006
Final loss
5.521357e-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 | 5.4459e-2 |
| 2,000 | 1.4137e-2 |
| 3,000 | 8.4518e-3 |
| 4,000 | 7.4363e-3 |
| 5,000 | 6.6338e-3 |
| 6,000 | 6.2213e-3 |
| 7,000 | 5.9891e-3 |
| 8,000 | 5.8585e-3 |
| 9,000 | 5.7266e-3 |
| 10,000 | 5.6640e-3 |
| 11,000 | 5.5853e-3 |
| 12,000 | 5.5234e-3 |
| 12,500 | 5.5214e-3 |
06
関連 finding
この run を related_runs に含む finding (Positive / Negative どちらも)。