Files
hypertower/v4/scripts/experiments/phase3_v4/single_eye_grid.json
T
rpotter6298 280060db82 Add new regression and ensemble experiment configurations for V2-M and OrthoBridge
- Introduced multiple regression experiment configurations targeting vf_md, including:
  - cd_solo_reg_set.json: CD tower only regression setup.
  - img_solo_reg_set.json: Image tower only regression setup.
  - reg_head_epoch_sweep.json: Baseline regression sweeps at different epochs (50, 75, 100).
  - reg_head_set.json: Various regression setups including baseline and OrthoBridge configurations.
  - single_eye_reg.json: Single-eye regression setup for worst-eye aggregation analysis.

- Added ensemble configurations for OrthoBridge with different inner bridges:
  - ortho_alts_ensemble.json: Ensemble tests with ConcatBridge, PairwiseAdditiveBridge, and GatedAdditiveBridge.
  - ortho_alts_tritower.json: Tritower tests with the same inner bridges.

- Created V2-M specific configurations:
  - baseline_reg_nt50.json: Regression baseline with V2-M backbone.
  - geom_vec_gt.json and geom_vec_unet.json: Geometry vector injection experiments with V2-M.
  - single_l1_bridges.json: Single-eye ensemble experiments with various bridge types.
  - tritower_geom_gt.json: Tritower setup with GT contour-rasterized masks.

- Promoted existing experiments to higher repetitions for robustness.
2026-06-11 15:08:20 +02:00

40 lines
1.3 KiB
JSON
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
[
{
"_note": "Single-eye + all_losses + Hadamard FusionBridge. Pairs with the in-flight ensemble_single_refugelike (single+BCD+Hadamard) to isolate the BCD-vs-all-losses effect. v3 phase 3 originally showed BCD generalizes better; this re-establishes it in v4.",
"run_name": "experiments/phase3_v4/single_all_losses_hadamard",
"reps": 10,
"overrides": {
"training": { "tower_loss_mode": "all_losses" }
}
},
{
"_note": "Single-eye + BCD + ConcatBridge. Pairs with in-flight baseline (single+BCD+Hadamard) to isolate the bridge effect. Justifies why we use FusionBridge (Hadamard) as default.",
"run_name": "experiments/phase3_v4/single_bcd_concat",
"reps": 10,
"stage_overrides": {
"nt": {
"module": "v4.classes.bridges.concat_bridge",
"class": "ConcatBridge",
"args": { "fusion_dim": 256 }
}
}
},
{
"_note": "Single-eye + all_losses + ConcatBridge. Fourth cell of the BCD-vs-all-losses × Hadamard-vs-Concat 2x2.",
"run_name": "experiments/phase3_v4/single_all_losses_concat",
"reps": 10,
"overrides": {
"training": { "tower_loss_mode": "all_losses" }
},
"stage_overrides": {
"nt": {
"module": "v4.classes.bridges.concat_bridge",
"class": "ConcatBridge",
"args": { "fusion_dim": 256 }
}
}
}
]