← Run Index
Run summary · P1.M5 multi-scene parity (低 wallclock + 低 splats 版)
multiscene-ficus-30k-q2
静穏再実行 (ficus, G.3-alone 30k recipe, n=1, load-controlled)
partial pass
01
設定
- dataset
- /Users/otkrickey/dev/3dgs-workspace/datasets/nerf_synthetic/ficus
- 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
34.28dB
Wallclock
4m 16s
Final splats
228,595
Final loss
4.148691e-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 | 6.5095e-1 |
| 1,000 | 3.4271e-2 |
| 2,000 | 1.4616e-2 |
| 3,000 | 1.0110e-2 |
| 4,000 | 7.2417e-3 |
| 5,000 | 6.5789e-3 |
| 6,000 | 6.3054e-3 |
| 7,000 | 6.0638e-3 |
| 8,000 | 5.4981e-3 |
| 9,000 | 5.2048e-3 |
| 10,000 | 5.0697e-3 |
| 11,000 | 5.0342e-3 |
| 12,000 | 4.9125e-3 |
| 13,000 | 4.7925e-3 |
| 14,000 | 4.8382e-3 |
| 15,000 | 4.7300e-3 |
| 16,000 | 4.5835e-3 |
| 17,000 | 4.5810e-3 |
| 18,000 | 4.5416e-3 |
| 19,000 | 4.4924e-3 |
| 20,000 | 4.4902e-3 |
| 21,000 | 4.4066e-3 |
| 22,000 | 4.3806e-3 |
| 23,000 | 4.2943e-3 |
| 24,000 | 4.3546e-3 |
| 25,000 | 4.3035e-3 |
| 26,000 | 4.2828e-3 |
| 27,000 | 4.3123e-3 |
| 28,000 | 4.1658e-3 |
| 29,000 | 4.2041e-3 |
| 30,000 | 4.1487e-3 |
06
関連 finding
この run を related_runs に含む finding (Positive / Negative どちらも)。