Migrate shared config profiles and fix regressions

Author: Donglai Wei

CLAUDE.md

@@ -281,14 +281,17 @@ scripts/                         # Entry points and utilities
     ├── compare_config.py        # Config comparison tool
     └── eval_curvilinear.py      # Curvilinear structure evaluation
 
-tutorials/                       # Example configurations
-├── bases/                       # Reusable profile bases (11 YAML files)
-│   ├── all_profiles.yaml        # Master profile index
+configs/                         # Canonical shared YAML registries
+├── all_profiles.yaml            # Master registry index used by tutorials
+├── profiles/                    # Section-level profile registries
 │   ├── arch_profiles.yaml       # Architecture presets
-│   ├── augmentation_profiles.yaml  # Augmentation presets
-│   ├── loss_profiles.yaml       # Loss function presets
-│   ├── optimizer_profiles.yaml  # Optimizer presets
-│   └── ...                      # system, dataloader, decoding, etc.
+│   ├── loss_profiles.yaml       # Loss presets
+│   ├── label_profiles.yaml      # Label-transform presets
+│   └── ...                      # system, dataloader, augmentation, pipeline, tune
+└── templates/                   # Explicit list-item templates
+    └── decoding_templates.yaml  # `inference.decoding` templates (`template: ...`)
+
+tutorials/                       # Example configurations
 ├── misc/                        # Miscellaneous experiments
 └── *.yaml                       # Dataset-specific configs (16 files)
     # mito_lucchi++, mito_mitoEM_*, mito_mitolab, mito_betaseg,
@@ -310,6 +313,20 @@ docker/                          # Docker containerization
 ### Hydra Configuration (Primary)
 The project uses **Hydra/OmegaConf** with dataclass-based configs for type safety and composability.
 
+Canonical YAML layout:
+
+- `configs/profiles/*.yaml`: section-level registries selected by `*.profile`
+- `configs/templates/*.yaml`: explicit list-item templates, currently for `inference.decoding`
+- `tutorials/*.yaml`: runnable experiments that `_base_` the shared registries
+
+Canonical merge semantics:
+
+- Profile payloads are merged into the target section as the base config.
+- Explicit keys override profile keys.
+- Explicit lists replace profile lists; list overrides are not additive.
+- Canonical decoding expansion is explicit `template:` inside `inference.decoding`.
+- Do not introduce `decoding_profile` or `- profile: decoding_*` usages.
+
 **Config File Example** (`tutorials/lucchi.yaml`):
 ```yaml
 system:

README.md

@@ -182,6 +182,19 @@ just test lucchi++ outputs/lucchi++/$EXPERIMENT_DATE/checkpoints/best.ckpt
 - 🔧 **[Troubleshooting](TROUBLESHOOTING.md)** - Common issues and solutions
 - 👨‍💻 **[Developer Guide](.claude/CLAUDE.md)** - Contributing and architecture
 
+## Config Layout
+
+- `tutorials/*.yaml`: runnable experiment configs
+- `configs/profiles/*.yaml`: section-level registries selected by `*.profile`
+- `configs/templates/*.yaml`: explicit list-item templates, currently used for `inference.decoding`
+
+Merge rule:
+
+- Profile payloads are merged into the target section as the base config.
+- Explicit keys in the tutorial/config override profile keys.
+- Explicit lists replace profile lists; they are not additive.
+- Canonical decoding syntax is explicit list expansion, for example `inference.decoding: [{template: decoding_waterz}]`.
+
 ---
 
 ## Example: Train a Model

configs/all_profiles.yaml

@@ -0,0 +1,12 @@
+_base_:
+  - profiles/arch_profiles.yaml
+  - profiles/loss_profiles.yaml
+  - profiles/label_profiles.yaml
+  - profiles/system_profiles.yaml
+  - templates/decoding_templates.yaml
+  - profiles/activation_profiles.yaml
+  - profiles/augmentation_profiles.yaml
+  - profiles/dataloader_profiles.yaml
+  - profiles/optimizer_profiles.yaml
+  - profiles/tune_profiles.yaml
+  - profiles/pipeline_profiles.yaml

tutorials/bases/activation_profiles.yaml configs/profiles/activation_profiles.yamltutorials/bases/activation_profiles.yaml renamed to configs/profiles/activation_profiles.yaml

@@ -18,7 +18,7 @@ label_profiles:
         kwargs:
           offsets: ["0-0-1", "0-1-0", "1-0-0", "0-0-3", "0-3-0", "3-0-0", "0-0-9", "0-9-0", "9-0-0"]
           deepem_crop: true
-  label_affinity_9_sdt:
+  label_aff9_sdt:
     erosion: 1
     targets:
       - name: affinity
@@ -29,9 +29,8 @@ label_profiles:
         kwargs:
           alpha: 0.8
           bg_value: -1.0
-          relabel: true
 
-  label_affinity_12:
+  label_aff12:
     erosion: 1
     targets:
       - name: affinity