Chart Overlays
Compare this run against its peers
Add peer runs here to layer their train and val traces onto the current metric charts.
interrupted Apr 7, 12:29 Duration: 17m 51s Python 3.11.7
src/train.py --config configs/model/novel_metric_fusion_gt_rays.yaml --config configs/model/backbone_dinov2_vitl14.yaml --config configs/data/real_drive_sim_euler_loading.yaml --config configs/training/phase_b2_structure_joint_refine.yaml --init-from /cluster/scratch/drothenpiele/euler_train/unifuse/checkpoints/marigold-structure-b0-concat-warmup-2026-04-07-00-49-18-935a/checkpoint-epoch-6.ptlr 3e-5batch 8epochs 5precision automodel weather_metric_system wd 1e-4
ID 62594714 Job metric_diffusion_pty Part gpupr.24h Node eu-a65-07 CPU 8 GPU 1
unifuse/runs/2026-04-07_12-29-37_3e67 Error
Signal: SIGINT
Metrics 37/37
depth.train 22
diag 10
consolidation 2
base_vs_prior
delta_vs_prior
final 3
depth_mae
log_radius_l1
xyz_l1
prior 3
depth_mae
log_radius_l1
xyz_l1
structure 2
depth_l1
edge_l1
loss 11
final 4
confidence
depth
log_radius
xyz
prior 2
confidence
log_radius
structure 4
confidence
denoise_mse
depth
edge
no stage 1
total
stat 1
confidence
rgb.train 1
dehaze
sys.train 14
gpu_mem_total_gb
gpu_mem_used_gb
gpu_mem_util_pct
gpu_util_pct
lr
lr.default
lr.structure_expert_context
lr.structure_expert_heads
lr.structure_expert_unet_conv_in
lr.structure_expert_unet_cross_attention
lr.structure_expert_unet_down
lr.structure_expert_unet_mid
lr.structure_expert_unet_out
lr.structure_expert_unet_up
37 selected
X-Axis
Y-Scale
Series
Group
Namespace
Base Vs Prior
↓ kind=diagstage=consolidation T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Delta Vs Prior
↓ kind=diagstage=consolidation T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Depth Mae
↓ meters kind=diagstage=final T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Log Radius L1
↓ kind=diagstage=final T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Xyz L1
↓ meters kind=diagstage=final T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Depth Mae
↓ meters kind=diagstage=prior T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Log Radius L1
↓ kind=diagstage=prior T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Xyz L1
↓ meters kind=diagstage=prior T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Depth L1
↓ kind=diagstage=structure T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Edge L1
↓ kind=diagstage=structure T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Confidence
↑ kind=lossstage=final T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Depth
↓ meters kind=lossstage=final T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Log Radius
↓ kind=lossstage=final T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Xyz
↓ meters kind=lossstage=final T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Confidence
↑ kind=lossstage=prior T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Log Radius
↓ kind=lossstage=prior T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Confidence
↑ kind=lossstage=structure T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Denoise Mse
↓ kind=lossstage=structure T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Depth
↓ meters kind=lossstage=structure T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Edge
↓ kind=lossstage=structure T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Total
↓ kind=loss T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Confidence
↑ kind=statstage=structure T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Dehaze
↓ kind=lossstage=weather T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Gpu Mem Total Gb
T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Gpu Mem Used Gb
T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Gpu Mem Util Pct
T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Gpu Util Pct
T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Lr
T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Lr.Default
T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Lr.Structure Expert Context
T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Lr.Structure Expert Heads
T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Lr.Structure Expert Unet Conv In
T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Lr.Structure Expert Unet Cross Attention
T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Lr.Structure Expert Unet Down
T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Lr.Structure Expert Unet Mid
T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Lr.Structure Expert Unet Out
T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
Lr.Structure Expert Unet Up
T
2026-04-07_12-29-37_3e67 train (smoothed) —
2026-04-07_12-29-37_3e67 train (raw) —
No output snapshots found for this run.
Outputs are generated during training and saved to outputs/epoch_N_step_M/ directories.
Run-Owned Exports
Publish direct filesystem paths here.
Published
No run-owned exports are published yet.
Raw Artifacts
Run-owned exports are typically direct paths, so there are no captured artifacts to publish from here.