Fix duplicate pretrained weight loading in Model.train() for YAML models (#23640)

Co-authored-by: Onuralp SEZER <onuralp@ultralytics.com>
Co-authored-by: Ultralytics Assistant <135830346+UltralyticsAssistant@users.noreply.github.com>
Co-authored-by: UltralyticsAssistant <web@ultralytics.com>
Co-authored-by: Laughing-q <1185102784@qq.com>
Co-authored-by: Jing Qiu <61612323+Laughing-q@users.noreply.github.com>
This commit is contained in:
artest08 2026-04-07 16:02:35 +03:00 committed by GitHub
parent 75acff57c2
commit fa1db1f467
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
10 changed files with 17 additions and 16 deletions

View File

@ -756,8 +756,6 @@ class Model(torch.nn.Module):
checks.check_pip_update_available()
if isinstance(kwargs.get("pretrained", None), (str, Path)):
self.load(kwargs["pretrained"]) # load pretrained weights if provided
overrides = YAML.load(checks.check_yaml(kwargs["cfg"])) if kwargs.get("cfg") else self.overrides
custom = {
# NOTE: handle the case when 'cfg' includes 'data'.
@ -781,9 +779,6 @@ class Model(torch.nn.Module):
args["resume"] = False
self.trainer = (trainer or self._smart_load("trainer"))(overrides=args, _callbacks=self.callbacks)
if not args.get("resume"): # manually set model only if not resuming
self.trainer.model = self.trainer.get_model(weights=self.model if self.ckpt else None, cfg=self.model.yaml)
self.model = self.trainer.model
self.trainer.train()
# Update model and cfg after training

View File

@ -720,7 +720,7 @@ class BaseTrainer:
cfg = weights.yaml
elif isinstance(self.args.pretrained, (str, Path)):
weights, _ = load_checkpoint(self.args.pretrained)
self.model = self.get_model(cfg=cfg, weights=weights, verbose=RANK == -1) # calls Model(cfg, weights)
self.model = self.get_model(cfg=cfg, weights=weights, verbose=RANK in {-1, 0}) # calls Model(cfg, weights)
return ckpt
def optimizer_step(self):

View File

@ -52,7 +52,9 @@ class RTDETRTrainer(DetectionTrainer):
Returns:
(RTDETRDetectionModel): Initialized model.
"""
model = RTDETRDetectionModel(cfg, nc=self.data["nc"], ch=self.data["channels"], verbose=verbose and RANK == -1)
model = RTDETRDetectionModel(
cfg, nc=self.data["nc"], ch=self.data["channels"], verbose=verbose and RANK in {-1, 0}
)
if weights:
model.load(weights)
return model

View File

@ -79,7 +79,9 @@ class ClassificationTrainer(BaseTrainer):
Returns:
(ClassificationModel): Configured PyTorch model for classification.
"""
model = ClassificationModel(cfg, nc=self.data["nc"], ch=self.data["channels"], verbose=verbose and RANK == -1)
model = ClassificationModel(
cfg, nc=self.data["nc"], ch=self.data["channels"], verbose=verbose and RANK in {-1, 0}
)
if weights:
model.load(weights)

View File

@ -160,7 +160,7 @@ class DetectionTrainer(BaseTrainer):
Returns:
(DetectionModel): YOLO detection model.
"""
model = DetectionModel(cfg, nc=self.data["nc"], ch=self.data["channels"], verbose=verbose and RANK == -1)
model = DetectionModel(cfg, nc=self.data["nc"], ch=self.data["channels"], verbose=verbose and RANK in {-1, 0})
if weights:
model.load(weights)
return model

View File

@ -64,7 +64,7 @@ class OBBTrainer(yolo.detect.DetectionTrainer):
>>> trainer = OBBTrainer()
>>> model = trainer.get_model(cfg="yolo26n-obb.yaml", weights="yolo26n-obb.pt")
"""
model = OBBModel(cfg, nc=self.data["nc"], ch=self.data["channels"], verbose=verbose and RANK == -1)
model = OBBModel(cfg, nc=self.data["nc"], ch=self.data["channels"], verbose=verbose and RANK in {-1, 0})
if weights:
model.load(weights)

View File

@ -55,7 +55,9 @@ class SegmentationTrainer(yolo.detect.DetectionTrainer):
>>> model = trainer.get_model(cfg="yolo26n-seg.yaml")
>>> model = trainer.get_model(weights="yolo26n-seg.pt", verbose=False)
"""
model = SegmentationModel(cfg, nc=self.data["nc"], ch=self.data["channels"], verbose=verbose and RANK == -1)
model = SegmentationModel(
cfg, nc=self.data["nc"], ch=self.data["channels"], verbose=verbose and RANK in {-1, 0}
)
if weights:
model.load(weights)

View File

@ -83,7 +83,7 @@ class WorldTrainer(DetectionTrainer):
cfg["yaml_file"] if isinstance(cfg, dict) else cfg,
ch=self.data["channels"],
nc=min(self.data["nc"], 80),
verbose=verbose and RANK == -1,
verbose=verbose and RANK in {-1, 0},
)
if weights:
model.load(weights)

View File

@ -70,7 +70,7 @@ class YOLOETrainer(DetectionTrainer):
cfg["yaml_file"] if isinstance(cfg, dict) else cfg,
ch=self.data["channels"],
nc=min(self.data["nc"], 80),
verbose=verbose and RANK == -1,
verbose=verbose and RANK in {-1, 0},
)
if weights:
model.load(weights)
@ -128,7 +128,7 @@ class YOLOEPETrainer(DetectionTrainer):
cfg["yaml_file"] if isinstance(cfg, dict) else cfg,
ch=self.data["channels"],
nc=self.data["nc"],
verbose=verbose and RANK == -1,
verbose=verbose and RANK in {-1, 0},
)
del model.model[-1].savpe

View File

@ -39,7 +39,7 @@ class YOLOESegTrainer(YOLOETrainer, SegmentationTrainer):
cfg["yaml_file"] if isinstance(cfg, dict) else cfg,
ch=self.data["channels"],
nc=min(self.data["nc"], 80),
verbose=verbose and RANK == -1,
verbose=verbose and RANK in {-1, 0},
)
if weights:
model.load(weights)
@ -85,7 +85,7 @@ class YOLOEPESegTrainer(SegmentationTrainer):
cfg["yaml_file"] if isinstance(cfg, dict) else cfg,
ch=self.data["channels"],
nc=self.data["nc"],
verbose=verbose and RANK == -1,
verbose=verbose and RANK in {-1, 0},
)
del model.model[-1].savpe