Upload 21 files

milk10k_effb2_metadata/cli.py

@@ -9,8 +9,18 @@ from pathlib import Path
 def parse_args() -> argparse.Namespace:
     parser = argparse.ArgumentParser(description="Train MILK10k dual EfficientNet-B2 with metadata fusion.")
     parser.add_argument("--data-dir", type=Path, default=None)
-    parser.add_argument("--clinical-checkpoint", type=Path, default=None)
-    parser.add_argument("--dermoscopic-checkpoint", type=Path, default=None)
+    parser.add_argument(
+        "--clinical-checkpoint",
+        type=Path,
+        default=None,
+        help="Optional clinical encoder checkpoint. If omitted, the clinical branch uses ImageNet-pretrained backbone weights.",
+    )
+    parser.add_argument(
+        "--dermoscopic-checkpoint",
+        type=Path,
+        default=None,
+        help="Optional dermoscopic encoder checkpoint. If omitted, the dermoscopic branch uses ImageNet-pretrained backbone weights.",
+    )
     parser.add_argument(
         "--resume-checkpoint",
         type=Path,
@@ -84,7 +94,7 @@ def parse_args() -> argparse.Namespace:
     parser.add_argument(
         "--imagenet-pretrained",
         action="store_true",
-        help="Initialize EfficientNet-B2 with ImageNet weights before loading branch checkpoints.",
+        help="Initialize backbones with ImageNet weights before loading any branch checkpoints. Enabled automatically when no branch checkpoints are passed.",
     )
     parser.add_argument("--patience", type=int, default=6)
     return parser.parse_args()

milk10k_effb2_metadata/losses.py

@@ -59,13 +59,16 @@ class MILKLongTailLoss(nn.Module):
         self.current_epoch = epoch
 
     def forward(self, logits: torch.Tensor, labels: torch.Tensor) -> torch.Tensor:
+        margins = self.margins.to(device=logits.device, dtype=logits.dtype)
+        log_priors = self.log_priors.to(device=logits.device, dtype=logits.dtype)
+        alpha = self.alpha.to(device=logits.device, dtype=logits.dtype)
         adjusted_logits = logits.clone()
         rows = torch.arange(labels.size(0), device=labels.device)
-        adjusted_logits[rows, labels] = adjusted_logits[rows, labels] - self.margins[labels]
-        adjusted_logits = adjusted_logits + self.logit_tau * self.log_priors
+        adjusted_logits[rows, labels] = adjusted_logits[rows, labels] - margins[labels]
+        adjusted_logits = adjusted_logits + self.logit_tau * log_priors
         loss = F.cross_entropy(adjusted_logits, labels, reduction="none")
         if self.current_epoch >= self.deferred_start_epoch:
-            loss = loss * self.alpha[labels]
+            loss = loss * alpha[labels]
         return loss.mean()
 
 

milk10k_effb2_metadata/training.py

@@ -17,7 +17,7 @@ from torch.utils.data import DataLoader
 from tqdm.auto import tqdm
 
 from datasets import resolve_data_dir, set_seed
-from milk10k_effb2_metadata.checkpoints import load_encoder_checkpoint, resolve_backbone_backends
+from milk10k_effb2_metadata.checkpoints import infer_checkpoint_backend, load_encoder_checkpoint, resolve_backbone_backends
 from milk10k_effb2_metadata.data import (
     fit_metadata_spec,
     kfold_splits,
@@ -51,11 +51,23 @@ def resolve_training_backbone_backends(args: argparse.Namespace, device: torch.d
         return args.backbone_backend, args.backbone_backend
     if args.clinical_checkpoint is not None and args.dermoscopic_checkpoint is not None:
         return resolve_backbone_backends(args, device)
-    if args.resume_checkpoint is None:
-        raise ValueError(
-            "Pass --clinical-checkpoint and --dermoscopic-checkpoint for a fresh run, "
-            "or pass --resume-checkpoint to resume without branch checkpoints."
+    if args.clinical_checkpoint is not None:
+        clinical_backend = infer_checkpoint_backend(args.clinical_checkpoint, device, "clinical")
+        print(
+            "Auto-detected clinical backbone backend: "
+            f"clinical={clinical_backend}, dermoscopic={clinical_backend} (ImageNet initialized)"
         )
+        return clinical_backend, clinical_backend
+    if args.dermoscopic_checkpoint is not None:
+        dermoscopic_backend = infer_checkpoint_backend(args.dermoscopic_checkpoint, device, "dermoscopic")
+        print(
+            "Auto-detected dermoscopic backbone backend: "
+            f"clinical={dermoscopic_backend} (ImageNet initialized), dermoscopic={dermoscopic_backend}"
+        )
+        return dermoscopic_backend, dermoscopic_backend
+    if args.resume_checkpoint is None:
+        print("No branch checkpoints passed; using torchvision backbones initialized from ImageNet weights.")
+        return "torchvision", "torchvision"
     checkpoint = torch.load(args.resume_checkpoint.expanduser().resolve(), map_location=device, weights_only=False)
     state = checkpoint["model_state"]
     clinical_backend = infer_branch_backend_from_state(state, "clinical_encoder.")
@@ -311,10 +323,10 @@ def build_model(
         disable_metadata=args.disable_metadata,
     ).to(device)
     if args.resume_checkpoint is None:
-        if args.clinical_checkpoint is None or args.dermoscopic_checkpoint is None:
-            raise ValueError("Fresh training requires --clinical-checkpoint and --dermoscopic-checkpoint.")
-        load_encoder_checkpoint(args.clinical_checkpoint, model.clinical_encoder, "clinical", device)
-        load_encoder_checkpoint(args.dermoscopic_checkpoint, model.dermoscopic_encoder, "dermoscopic", device)
+        if args.clinical_checkpoint is not None:
+            load_encoder_checkpoint(args.clinical_checkpoint, model.clinical_encoder, "clinical", device)
+        if args.dermoscopic_checkpoint is not None:
+            load_encoder_checkpoint(args.dermoscopic_checkpoint, model.dermoscopic_encoder, "dermoscopic", device)
     if args.disable_metadata or args.freeze_metadata_head:
         set_metadata_head_trainable(model, False)
     return model
@@ -587,6 +599,8 @@ def run(args: argparse.Namespace) -> None:
 
     from milk10k_effb2_metadata.models import normalize_backbone_name
     args.backbone = normalize_backbone_name(args.backbone)
+    if args.resume_checkpoint is None and args.clinical_checkpoint is None and args.dermoscopic_checkpoint is None:
+        args.imagenet_pretrained = True
     if args.image_size is None:
         if args.backbone == "efficientnet_b2":
             args.image_size = 260