← Run Index
Run summary · P1.M5 multi-scene parity (低 wallclock + 低 splats 版)
multiscene-materials-30k
多 scene 退行チェック (materials, G.3-alone 30k recipe, n=1)
partial pass
01
設定
- dataset
- /Users/otkrickey/dev/3dgs-workspace/datasets/nerf_synthetic/materials
- gt_convention
- Premultiplied
- iterations
- 30,000
- 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.radix_scatter
- SimdBallot
- backend.binning
- Resident
- backend.tile_sort
- GlobalRadix
- backend.submission
- Chained
- backend.target_cache
- true
- backend.shader_fast_math
- true
- os_build
- 26.6.2 (25G83)
- backend.ssim_clear
- HostMemset
- backend.raster_grad_clear
- HostMemset
- backend.backward_uniform
- HostBuffer
- backend.lazy_c3_scratch
- false
- backend.ssim_walk
- PerChannel
02
結果
Val PSNR
30.00dB
Wallclock
7m 1s
Final splats
355,033
Final loss
5.299567e-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 | 6.0859e-1 |
| 1,000 | 4.7489e-2 |
| 2,000 | 2.1773e-2 |
| 3,000 | 1.4185e-2 |
| 4,000 | 1.1283e-2 |
| 5,000 | 9.8253e-3 |
| 6,000 | 9.0117e-3 |
| 7,000 | 8.4101e-3 |
| 8,000 | 7.8975e-3 |
| 9,000 | 7.5306e-3 |
| 10,000 | 7.2892e-3 |
| 11,000 | 7.0831e-3 |
| 12,000 | 6.8776e-3 |
| 13,000 | 6.6933e-3 |
| 14,000 | 6.5135e-3 |
| 15,000 | 6.3785e-3 |
| 16,000 | 6.2522e-3 |
| 17,000 | 6.1902e-3 |
| 18,000 | 6.0565e-3 |
| 19,000 | 5.9562e-3 |
| 20,000 | 5.7882e-3 |
| 21,000 | 5.6984e-3 |
| 22,000 | 5.6793e-3 |
| 23,000 | 5.6577e-3 |
| 24,000 | 5.5783e-3 |
| 25,000 | 5.4889e-3 |
| 26,000 | 5.4769e-3 |
| 27,000 | 5.4208e-3 |
| 28,000 | 5.4095e-3 |
| 29,000 | 5.3218e-3 |
| 30,000 | 5.2996e-3 |
06
関連 finding
この run を related_runs に含む finding (Positive / Negative どちらも)。