NOOB config for v1_5

configs/NOOB_BASE.knf.yaml

@@ -138,7 +138,7 @@ training:
     # into the subset backend entry after the dataset defaults below.
     subsets:
       felkin_anchor:
-        probability: 0.5          # Sample anchor identity images at steady exposure.
+        probability: 0.01          # Sample anchor identity images at steady exposure.
         crop: false             # Preserve full source composition.
         # repeats: 0            # Optional per-subset repeat count.
         # disabled: false       # Optional per-subset on/off switch.
@@ -158,10 +158,10 @@ training:
         # repeats: 0
         # disabled: false
       head:
-        probability: 4          # Sample head/face details at extra exposure.
+        probability: 3          # Sample head/face details at extra exposure.
         crop: false             # Preserve detail framing; many samples are close crops.
       genitals:
-        probability: 2          # Sample anatomy subset at normal probability.
+        probability: 3          # Sample anatomy subset at normal probability.
         crop: false             # Preserve anatomy crop framing.
         # repeats: 0
         # disabled: false
@@ -242,7 +242,7 @@ training:
     subset_warmup:
       enabled: true
       subsets: [fullbody, head]
-      steps: 500 # !! First steps of the total run are ONLY listed subsets when enabled!
+      steps: 1000 # !! First steps of the total run are ONLY listed subsets when enabled!
 
     # Full-run trainer config. trainer_testrun below overrides selected keys
     # only when knf train prepare/start is called with --testrun.
@@ -252,7 +252,7 @@ training:
       # hub_model_id: ladybird-noob-lora                # Optional Hub model id if push_to_hub is enabled.
 
       optimizer: adamw_bf16              # BF16 AdamW optimizer.
-      learning_rate: 0.0002              # SDXL LoRA learning rates fall between 1e-4 and 5e-4
+      learning_rate: 0.00015              # SDXL LoRA learning rates fall between 1e-4 and 5e-4
       lr_scheduler: constant_with_warmup # Ramp up LR to avoid weird color biases.
       lr_warmup_steps: 200
       seed: 42                           # Training seed.
@@ -263,9 +263,9 @@ training:
       train_batch_size: 4            # Per-device batch size.
       gradient_accumulation_steps: 1 # Effective batch multiplier.
       num_train_epochs: 0            # Step-based training; max_train_steps controls run length.
-      max_train_steps: 3000          # Current short SDXL run length.
+      max_train_steps: 3500          # Current short SDXL run length.
 
-      caption_dropout_probability: 0.0 # Caption dropout for regularization.
+      caption_dropout_probability: 0.05 # Caption dropout for regularization.
       data_backend_sampling: uniform   # Uniform dataloader sampling.
       allow_dataset_oversubscription: true