ultralytics 8.3.197__py3-none-any.whl → 8.3.198__py3-none-any.whl

tests/test_engine.py

@@ -67,7 +67,15 @@ def test_detect():
 
 def test_segment():
     """Test image segmentation training, validation, and prediction pipelines using YOLO models."""
-    overrides = {"data": "coco8-seg.yaml", "model": "yolo11n-seg.yaml", "imgsz": 32, "epochs": 1, "save": False}
+    overrides = {
+        "data": "coco8-seg.yaml",
+        "model": "yolo11n-seg.yaml",
+        "imgsz": 32,
+        "epochs": 1,
+        "save": False,
+        "mask_ratio": 1,
+        "overlap_mask": False,
+    }
     cfg = get_cfg(DEFAULT_CFG)
     cfg.data = "coco8-seg.yaml"
     cfg.imgsz = 32

ultralytics/__init__.py

@@ -1,6 +1,6 @@
 # Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
 
-__version__ = "8.3.197"
+__version__ = "8.3.198"
 
 import os
 

ultralytics/cfg/__init__.py

@@ -237,7 +237,6 @@ CFG_BOOL_KEYS = frozenset(
         "nms",
         "profile",
         "multi_scale",
-        "compile",
     }
 )
 

ultralytics/cfg/default.yaml

@@ -7,122 +7,124 @@ task: detect # (str) YOLO task, i.e. detect, segment, classify, pose, obb
 mode: train # (str) YOLO mode, i.e. train, val, predict, export, track, benchmark
 
 # Train settings -------------------------------------------------------------------------------------------------------
-model: # (str, optional) path to model file, i.e. yolov8n.pt, yolov8n.yaml
+model: # (str, optional) path to model file, i.e. yolov8n.pt or yolov8n.yaml
 data: # (str, optional) path to data file, i.e. coco8.yaml
 epochs: 100 # (int) number of epochs to train for
-time: # (float, optional) number of hours to train for, overrides epochs if supplied
-patience: 100 # (int) epochs to wait for no observable improvement for early stopping of training
-batch: 16 # (int) number of images per batch (-1 for AutoBatch)
-imgsz: 640 # (int | list) input images size as int for train and val modes, or list[h,w] for predict and export modes
+time: # (float, optional) max hours to train; overrides epochs if set
+patience: 100 # (int) early stop after N epochs without val improvement
+batch: 16 # (int) batch size; use -1 for AutoBatch
+imgsz: 640 # (int | list) train/val use int (square); predict/export may use [h,w]
 save: True # (bool) save train checkpoints and predict results
-save_period: -1 # (int) Save checkpoint every x epochs (disabled if < 1)
-cache: False # (bool) True/ram, disk or False. Use cache for data loading
-device: # (int | str | list) device: CUDA device=0 or [0,1,2,3] or "cpu/mps" or -1 or [-1,-1] to auto-select idle GPUs
-workers: 8 # (int) number of worker threads for data loading (per RANK if DDP)
-project: # (str, optional) project name
-name: # (str, optional) experiment name, results saved to 'project/name' directory
-exist_ok: False # (bool) whether to overwrite existing experiment
-pretrained: True # (bool | str) whether to use a pretrained model (bool) or a model to load weights from (str)
-optimizer: auto # (str) optimizer to use, choices=[SGD, Adam, Adamax, AdamW, NAdam, RAdam, RMSProp, auto]
-verbose: True # (bool) whether to print verbose output
+save_period: -1 # (int) save checkpoint every N epochs; disabled if < 1
+cache: False # (bool | str) cache images in RAM (True/'ram') or on 'disk' to speed dataloading; False disables
+device: # (int | str | list) device: 0 or [0,1,2,3] for CUDA, 'cpu'/'mps', or -1/[-1,-1] to auto-select idle GPUs
+workers: 8 # (int) dataloader workers (per RANK if DDP)
+project: # (str, optional) project name for results root
+name: # (str, optional) experiment name; results in 'project/name'
+exist_ok: False # (bool) overwrite existing 'project/name' if True
+pretrained: True # (bool | str) use pretrained weights (bool) or load weights from path (str)
+optimizer: auto # (str) optimizer: SGD, Adam, Adamax, AdamW, NAdam, RAdam, RMSProp, or auto
+verbose: True # (bool) print verbose logs during training/val
 seed: 0 # (int) random seed for reproducibility
-deterministic: True # (bool) whether to enable deterministic mode
-single_cls: False # (bool) train multi-class data as single-class
-rect: False # (bool) rectangular training if mode='train' or rectangular validation if mode='val'
-cos_lr: False # (bool) use cosine learning rate scheduler
-close_mosaic: 10 # (int) disable mosaic augmentation for final epochs (0 to disable)
-resume: False # (bool) resume training from last checkpoint
-amp: True # (bool) Automatic Mixed Precision (AMP) training, choices=[True, False], True runs AMP check
-fraction: 1.0 # (float) dataset fraction to train on (default is 1.0, all images in train set)
-profile: False # (bool) profile ONNX and TensorRT speeds during training for loggers
-freeze: # (int | list, optional) freeze first n layers, or freeze list of layer indices during training
-multi_scale: False # (bool) Whether to use multiscale during training
-compile: False # (bool) Run torch.compile() on the model before train/val/predict
+deterministic: True # (bool) enable deterministic ops; reproducible but may be slower
+single_cls: False # (bool) treat all classes as a single class
+rect: False # (bool) rectangular batches for train; rectangular batching for val when mode='val'
+cos_lr: False # (bool) cosine learning rate scheduler
+close_mosaic: 10 # (int) disable mosaic augmentation for final N epochs (0 to keep enabled)
+resume: False # (bool) resume training from last checkpoint in the run dir
+amp: True # (bool) Automatic Mixed Precision (AMP) training; True runs AMP capability check
+fraction: 1.0 # (float) fraction of training dataset to use (1.0 = all)
+profile: False # (bool) profile ONNX/TensorRT speeds during training for loggers
+freeze: # (int | list, optional) freeze first N layers (int) or specific layer indices (list)
+multi_scale: False # (bool) multiscale training by varying image size
+compile: False # (bool | str) enable torch.compile() backend='inductor'; True="default", False=off, or "default|reduce-overhead|max-autotune"
+
 # Segmentation
-overlap_mask: True # (bool) merge object masks into a single image mask during training (segment train only)
-mask_ratio: 4 # (int) mask downsample ratio (segment train only)
+overlap_mask: True # (bool) merge instance masks into one mask during training (segment only)
+mask_ratio: 4 # (int) mask downsample ratio (segment only)
+
 # Classification
-dropout: 0.0 # (float) use dropout regularization (classify train only)
+dropout: 0.0 # (float) dropout for classification head (classify only)
 
 # Val/Test settings ----------------------------------------------------------------------------------------------------
-val: True # (bool) validate/test during training
-split: val # (str) dataset split to use for validation, i.e. 'val', 'test' or 'train'
-save_json: False # (bool) save results to JSON file
-conf: # (float, optional) object confidence threshold for detection (default 0.25 predict, 0.001 val)
-iou: 0.7 # (float) intersection over union (IoU) threshold for NMS
+val: True # (bool) run validation/testing during training
+split: val # (str) dataset split to evaluate: 'val', 'test' or 'train'
+save_json: False # (bool) save results to COCO JSON for external evaluation
+conf: # (float, optional) confidence threshold; defaults: predict=0.25, val=0.001
+iou: 0.7 # (float) IoU threshold used for NMS
 max_det: 300 # (int) maximum number of detections per image
-half: False # (bool) use half precision (FP16)
+half: False # (bool) use half precision (FP16) if supported
 dnn: False # (bool) use OpenCV DNN for ONNX inference
 plots: True # (bool) save plots and images during train/val
 
 # Predict settings -----------------------------------------------------------------------------------------------------
-source: # (str, optional) source directory for images or videos
-vid_stride: 1 # (int) video frame-rate stride
-stream_buffer: False # (bool) buffer all streaming frames (True) or return the most recent frame (False)
-visualize: False # (bool) visualize model features (predict) or visualize TP, FP, FN (val)
-augment: False # (bool) apply image augmentation to prediction sources
+source: # (str, optional) path/dir/URL/stream for images or videos; e.g. 'ultralytics/assets' or '0' for webcam
+vid_stride: 1 # (int) read every Nth frame for video sources
+stream_buffer: False # (bool) True buffers all frames; False keeps the most recent frame for low-latency streams
+visualize: False # (bool) visualize model features (predict) or TP/FP/FN confusion (val)
+augment: False # (bool) apply test-time augmentation during prediction
 agnostic_nms: False # (bool) class-agnostic NMS
-classes: # (int | list[int], optional) filter results by class, i.e. classes=0, or classes=[0,2,3]
-retina_masks: False # (bool) use high-resolution segmentation masks
-embed: # (list[int], optional) return feature vectors/embeddings from given layers
+classes: # (int | list[int], optional) filter by class id(s), e.g. 0 or [0,2,3]
+retina_masks: False # (bool) use high-resolution segmentation masks (segment)
+embed: # (list[int], optional) return feature embeddings from given layer indices
 
 # Visualize settings ---------------------------------------------------------------------------------------------------
-show: False # (bool) show predicted images and videos if environment allows
-save_frames: False # (bool) save predicted individual video frames
-save_txt: False # (bool) save results as .txt file
-save_conf: False # (bool) save results with confidence scores
-save_crop: False # (bool) save cropped images with results
-show_labels: True # (bool) show prediction labels, i.e. 'person'
-show_conf: True # (bool) show prediction confidence, i.e. '0.99'
-show_boxes: True # (bool) show prediction boxes
-line_width: # (int, optional) line width of the bounding boxes. Scaled to image size if None.
+show: False # (bool) show images/videos in a window if supported
+save_frames: False # (bool) save individual frames from video predictions
+save_txt: False # (bool) save results as .txt files (xywh format)
+save_conf: False # (bool) save confidence scores with results
+save_crop: False # (bool) save cropped prediction regions to files
+show_labels: True # (bool) draw class labels on images, e.g. 'person'
+show_conf: True # (bool) draw confidence values on images, e.g. '0.99'
+show_boxes: True # (bool) draw bounding boxes on images
+line_width: # (int, optional) line width of boxes; auto-scales with image size if not set
 
 # Export settings ------------------------------------------------------------------------------------------------------
-format: torchscript # (str) format to export to, choices at https://docs.ultralytics.com/modes/export/#export-formats
-keras: False # (bool) use Kera=s
-optimize: False # (bool) TorchScript: optimize for mobile
-int8: False # (bool) CoreML/TF INT8 quantization
-dynamic: False # (bool) ONNX/TF/TensorRT: dynamic axes
-simplify: True # (bool) ONNX: simplify model using `onnxslim`
-opset: # (int, optional) ONNX: opset version
-workspace: # (float, optional) TensorRT: workspace size (GiB), `None` will let TensorRT auto-allocate memory
-nms: False # (bool) CoreML: add NMS
+format: torchscript # (str) target format, e.g. torchscript|onnx|openvino|engine|coreml|saved_model|pb|tflite|edgetpu|tfjs|paddle|mnn|ncnn|imx|rknn
+keras: False # (bool) TF SavedModel only (format=saved_model); enable Keras layers during export
+optimize: False # (bool) TorchScript only; apply mobile optimizations to the scripted model
+int8: False # (bool) INT8/PTQ where supported (openvino, tflite, tfjs, engine, imx); needs calibration data/fraction
+dynamic: False # (bool) dynamic shapes for torchscript, onnx, openvino, engine; enable variable image sizes
+simplify: True # (bool) ONNX/engine only; run graph simplifier for cleaner ONNX before runtime conversion
+opset: # (int, optional) ONNX/engine only; opset version for export; leave unset to use a tested default
+workspace: # (float, optional) engine (TensorRT) only; workspace size in GiB, e.g. 4
+nms: False # (bool) fuse NMS into exported model when backend supports; if True, conf/iou apply (agnostic_nms except coreml)
 
 # Hyperparameters ------------------------------------------------------------------------------------------------------
-lr0: 0.01 # (float) initial learning rate (i.e. SGD=1E-2, Adam=1E-3)
-lrf: 0.01 # (float) final learning rate (lr0 * lrf)
-momentum: 0.937 # (float) SGD momentum/Adam beta1
-weight_decay: 0.0005 # (float) optimizer weight decay 5e-4
-warmup_epochs: 3.0 # (float) warmup epochs (fractions ok)
-warmup_momentum: 0.8 # (float) warmup initial momentum
-warmup_bias_lr: 0.1 # (float) warmup initial bias lr
+lr0: 0.01 # (float) initial learning rate (SGD=1e-2, Adam/AdamW=1e-3)
+lrf: 0.01 # (float) final LR fraction; final LR = lr0 * lrf
+momentum: 0.937 # (float) SGD momentum or Adam beta1
+weight_decay: 0.0005 # (float) weight decay (L2 regularization)
+warmup_epochs: 3.0 # (float) warmup epochs (fractions allowed)
+warmup_momentum: 0.8 # (float) initial momentum during warmup
+warmup_bias_lr: 0.1 # (float) bias learning rate during warmup
 box: 7.5 # (float) box loss gain
-cls: 0.5 # (float) cls loss gain (scale with pixels)
-dfl: 1.5 # (float) dfl loss gain
-pose: 12.0 # (float) pose loss gain
-kobj: 1.0 # (float) keypoint obj loss gain
-nbs: 64 # (int) nominal batch size
-hsv_h: 0.015 # (float) image HSV-Hue augmentation (fraction)
-hsv_s: 0.7 # (float) image HSV-Saturation augmentation (fraction)
-hsv_v: 0.4 # (float) image HSV-Value augmentation (fraction)
-degrees: 0.0 # (float) image rotation (+/- deg)
-translate: 0.1 # (float) image translation (+/- fraction)
-scale: 0.5 # (float) image scale (+/- gain)
-shear: 0.0 # (float) image shear (+/- deg)
-perspective: 0.0 # (float) image perspective (+/- fraction), range 0-0.001
-flipud: 0.0 # (float) image flip up-down (probability)
-fliplr: 0.5 # (float) image flip left-right (probability)
-bgr: 0.0 # (float) image channel BGR (probability)
-mosaic: 1.0 # (float) image mosaic (probability)
-mixup: 0.0 # (float) image mixup (probability)
-cutmix: 0.0 # (float) image cutmix (probability)
-copy_paste: 0.0 # (float) segment copy-paste (probability)
-copy_paste_mode: "flip" # (str) the method to do copy_paste augmentation (flip, mixup)
-auto_augment: randaugment # (str) auto augmentation policy for classification (randaugment, autoaugment, augmix)
-erasing: 0.4 # (float) probability of random erasing during classification training (0-0.9), 0 means no erasing, must be less than 1.0.
+cls: 0.5 # (float) classification loss gain
+dfl: 1.5 # (float) distribution focal loss gain
+pose: 12.0 # (float) pose loss gain (pose tasks)
+kobj: 1.0 # (float) keypoint objectness loss gain (pose tasks)
+nbs: 64 # (int) nominal batch size used for loss normalization
+hsv_h: 0.015 # (float) HSV hue augmentation fraction
+hsv_s: 0.7 # (float) HSV saturation augmentation fraction
+hsv_v: 0.4 # (float) HSV value (brightness) augmentation fraction
+degrees: 0.0 # (float) rotation degrees (+/-)
+translate: 0.1 # (float) translation fraction (+/-)
+scale: 0.5 # (float) scale gain (+/-)
+shear: 0.0 # (float) shear degrees (+/-)
+perspective: 0.0 # (float) perspective fraction (0–0.001 typical)
+flipud: 0.0 # (float) vertical flip probability
+fliplr: 0.5 # (float) horizontal flip probability
+bgr: 0.0 # (float) RGB↔BGR channel swap probability
+mosaic: 1.0 # (float) mosaic augmentation probability
+mixup: 0.0 # (float) MixUp augmentation probability
+cutmix: 0.0 # (float) CutMix augmentation probability
+copy_paste: 0.0 # (float) segmentation copy-paste probability
+copy_paste_mode: flip # (str) copy-paste strategy for segmentation: flip or mixup
+auto_augment: randaugment # (str) classification auto augmentation policy: randaugment, autoaugment, augmix
+erasing: 0.4 # (float) random erasing probability for classification (0–0.9), <1.0
 
 # Custom config.yaml ---------------------------------------------------------------------------------------------------
-cfg: # (str, optional) for overriding defaults.yaml
+cfg: # (str, optional) path to a config.yaml that overrides defaults
 
 # Tracker settings ------------------------------------------------------------------------------------------------------
-tracker: botsort.yaml # (str) tracker type, choices=[botsort.yaml, bytetrack.yaml]
+tracker: botsort.yaml # (str) tracker config file: botsort.yaml or bytetrack.yaml

ultralytics/cfg/trackers/botsort.yaml

@@ -1,22 +1,21 @@
 # Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
 
-# Default Ultralytics settings for BoT-SORT tracker when using mode="track"
-# For documentation and examples see https://docs.ultralytics.com/modes/track/
-# For BoT-SORT source code see https://github.com/NirAharon/BoT-SORT
+# BoT-SORT tracker defaults for mode="track"
+# Docs: https://docs.ultralytics.com/modes/track/
 
-tracker_type: botsort # tracker type, ['botsort', 'bytetrack']
-track_high_thresh: 0.25 # threshold for the first association
-track_low_thresh: 0.1 # threshold for the second association
-new_track_thresh: 0.25 # threshold for init new track if the detection does not match any tracks
-track_buffer: 30 # buffer to calculate the time when to remove tracks
-match_thresh: 0.8 # threshold for matching tracks
-fuse_score: True # Whether to fuse confidence scores with the iou distances before matching
-# min_box_area: 10  # threshold for min box areas(for tracker evaluation, not used for now)
+tracker_type: botsort # (str) Tracker backend: botsort|bytetrack; choose botsort to enable BoT-SORT features
+track_high_thresh: 0.25 # (float) First-stage match threshold; raise for cleaner tracks, lower to keep more
+track_low_thresh: 0.1 # (float) Second-stage threshold for low-score matches; balances recovery vs drift
+new_track_thresh: 0.25 # (float) Start a new track if no match ≥ this; higher reduces false tracks
+track_buffer: 30 # (int) Frames to keep lost tracks alive; higher handles occlusion, increases ID switches risk
+match_thresh: 0.8 # (float) Association similarity threshold (IoU/cost); tune with detector quality
+fuse_score: True # (bool) Fuse detection score with motion/IoU for matching; stabilizes weak detections
+
+# BoT-SORT specifics
+gmc_method: sparseOptFlow # (str) Global motion compensation: sparseOptFlow|orb|none; helps moving camera scenes
 
-# BoT-SORT settings
-gmc_method: sparseOptFlow # method of global motion compensation
 # ReID model related thresh
-proximity_thresh: 0.5 # minimum IoU for valid match with ReID
-appearance_thresh: 0.8 # minimum appearance similarity for ReID
-with_reid: False
-model: auto # uses native features if detector is YOLO else yolo11n-cls.pt
+proximity_thresh: 0.5 # (float) Min IoU to consider tracks proximate for ReID; higher is stricter
+appearance_thresh: 0.8 # (float) Min appearance similarity for ReID; raise to avoid identity swaps
+with_reid: False # (bool) Enable ReID model use; needs extra model and compute
+model: auto # (str) ReID model name/path; "auto" uses detector features if available

ultralytics/cfg/trackers/bytetrack.yaml

@@ -1,14 +1,12 @@
 # Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
 
-# Default Ultralytics settings for ByteTrack tracker when using mode="track"
-# For documentation and examples see https://docs.ultralytics.com/modes/track/
-# For ByteTrack source code see https://github.com/ifzhang/ByteTrack
+# ByteTrack tracker defaults for mode="track"
+# Docs: https://docs.ultralytics.com/modes/track/
 
-tracker_type: bytetrack # tracker type, ['botsort', 'bytetrack']
-track_high_thresh: 0.25 # threshold for the first association
-track_low_thresh: 0.1 # threshold for the second association
-new_track_thresh: 0.25 # threshold for init new track if the detection does not match any tracks
-track_buffer: 30 # buffer to calculate the time when to remove tracks
-match_thresh: 0.8 # threshold for matching tracks
-fuse_score: True # Whether to fuse confidence scores with the iou distances before matching
-# min_box_area: 10  # threshold for min box areas(for tracker evaluation, not used for now)
+tracker_type: bytetrack # (str) Tracker backend: botsort|bytetrack; choose bytetrack for the classic baseline
+track_high_thresh: 0.25 # (float) First-stage match threshold; raise for cleaner tracks, lower to keep more
+track_low_thresh: 0.1 # (float) Second-stage threshold for low-score matches; balances recovery vs drift
+new_track_thresh: 0.25 # (float) Start a new track if no match ≥ this; higher reduces false tracks
+track_buffer: 30 # (int) Frames to keep lost tracks alive; higher handles occlusion, increases ID switches risk
+match_thresh: 0.8 # (float) Association similarity threshold (IoU/cost); tune with detector quality
+fuse_score: True # (bool) Fuse detection score with motion/IoU for matching; stabilizes weak detections

ultralytics/data/augment.py

@@ -2382,7 +2382,7 @@ class LoadVisualPrompt:
         #     assert len(cls_unique) == cls_unique[-1] + 1, (
         #         f"Expected a continuous range of class indices, but got {cls_unique}"
         #     )
-        visuals = torch.zeros(len(cls_unique), *masksz)
+        visuals = torch.zeros(cls_unique.shape[0], *masksz)
         for idx, mask in zip(inverse_indices, masks):
             visuals[idx] = torch.logical_or(visuals[idx], mask)
         return visuals

ultralytics/data/dataset.py

@@ -172,7 +172,7 @@ class YOLODataset(BaseDataset):
             cache, exists = load_dataset_cache_file(cache_path), True  # attempt to load a *.cache file
             assert cache["version"] == DATASET_CACHE_VERSION  # matches current version
             assert cache["hash"] == get_hash(self.label_files + self.im_files)  # identical hash
-        except (FileNotFoundError, AssertionError, AttributeError):
+        except (FileNotFoundError, AssertionError, AttributeError, ModuleNotFoundError):
             cache, exists = self.cache_labels(cache_path), False  # run cache ops
 
         # Display cache

ultralytics/engine/exporter.py

@@ -194,9 +194,9 @@ def try_export(inner_func):
         dt = 0.0
         try:
             with Profile() as dt:
-                f, model = inner_func(*args, **kwargs)
+                f = inner_func(*args, **kwargs)
             LOGGER.info(f"{prefix} export success ✅ {dt.t:.1f}s, saved as '{f}' ({file_size(f):.1f} MB)")
-            return f, model
+            return f
         except Exception as e:
             LOGGER.error(f"{prefix} export failure {dt.t:.1f}s: {e}")
             raise e
@@ -486,36 +486,36 @@ class Exporter:
         # Exports
         f = [""] * len(fmts)  # exported filenames
         if jit or ncnn:  # TorchScript
-            f[0], _ = self.export_torchscript()
+            f[0] = self.export_torchscript()
         if engine:  # TensorRT required before ONNX
-            f[1], _ = self.export_engine(dla=dla)
+            f[1] = self.export_engine(dla=dla)
         if onnx:  # ONNX
-            f[2], _ = self.export_onnx()
+            f[2] = self.export_onnx()
         if xml:  # OpenVINO
-            f[3], _ = self.export_openvino()
+            f[3] = self.export_openvino()
         if coreml:  # CoreML
-            f[4], _ = self.export_coreml()
+            f[4] = self.export_coreml()
         if is_tf_format:  # TensorFlow formats
             self.args.int8 |= edgetpu
             f[5], keras_model = self.export_saved_model()
             if pb or tfjs:  # pb prerequisite to tfjs
-                f[6], _ = self.export_pb(keras_model=keras_model)
+                f[6] = self.export_pb(keras_model=keras_model)
             if tflite:
-                f[7], _ = self.export_tflite()
+                f[7] = self.export_tflite()
             if edgetpu:
-                f[8], _ = self.export_edgetpu(tflite_model=Path(f[5]) / f"{self.file.stem}_full_integer_quant.tflite")
+                f[8] = self.export_edgetpu(tflite_model=Path(f[5]) / f"{self.file.stem}_full_integer_quant.tflite")
             if tfjs:
-                f[9], _ = self.export_tfjs()
+                f[9] = self.export_tfjs()
         if paddle:  # PaddlePaddle
-            f[10], _ = self.export_paddle()
+            f[10] = self.export_paddle()
         if mnn:  # MNN
-            f[11], _ = self.export_mnn()
+            f[11] = self.export_mnn()
         if ncnn:  # NCNN
-            f[12], _ = self.export_ncnn()
+            f[12] = self.export_ncnn()
         if imx:
-            f[13], _ = self.export_imx()
+            f[13] = self.export_imx()
         if rknn:
-            f[14], _ = self.export_rknn()
+            f[14] = self.export_rknn()
 
         # Finish
         f = [str(x) for x in f if x]  # filter out '' and None
@@ -580,7 +580,7 @@ class Exporter:
             optimize_for_mobile(ts)._save_for_lite_interpreter(str(f), _extra_files=extra_files)
         else:
             ts.save(str(f), _extra_files=extra_files)
-        return f, None
+        return f
 
     @try_export
     def export_onnx(self, prefix=colorstr("ONNX:")):
@@ -639,7 +639,7 @@ class Exporter:
             meta.key, meta.value = k, str(v)
 
         onnx.save(model_onnx, f)
-        return f, model_onnx
+        return f
 
     @try_export
     def export_openvino(self, prefix=colorstr("OpenVINO:")):
@@ -708,13 +708,13 @@ class Exporter:
                 ignored_scope=ignored_scope,
             )
             serialize(quantized_ov_model, fq_ov)
-            return fq, None
+            return fq
 
         f = str(self.file).replace(self.file.suffix, f"_openvino_model{os.sep}")
         f_ov = str(Path(f) / self.file.with_suffix(".xml").name)
 
         serialize(ov_model, f_ov)
-        return f, None
+        return f
 
     @try_export
     def export_paddle(self, prefix=colorstr("PaddlePaddle:")):
@@ -738,12 +738,12 @@ class Exporter:
 
         pytorch2paddle(module=self.model, save_dir=f, jit_type="trace", input_examples=[self.im])  # export
         YAML.save(Path(f) / "metadata.yaml", self.metadata)  # add metadata.yaml
-        return f, None
+        return f
 
     @try_export
     def export_mnn(self, prefix=colorstr("MNN:")):
         """Export YOLO model to MNN format using MNN https://github.com/alibaba/MNN."""
-        f_onnx, _ = self.export_onnx()  # get onnx model first
+        f_onnx = self.export_onnx()  # get onnx model first
 
         check_requirements("MNN>=2.9.6")
         import MNN  # noqa
@@ -763,7 +763,7 @@ class Exporter:
         convert_scratch = Path(self.file.parent / ".__convert_external_data.bin")
         if convert_scratch.exists():
             convert_scratch.unlink()
-        return f, None
+        return f
 
     @try_export
     def export_ncnn(self, prefix=colorstr("NCNN:")):
@@ -831,7 +831,7 @@ class Exporter:
             Path(f_debug).unlink(missing_ok=True)
 
         YAML.save(f / "metadata.yaml", self.metadata)  # add metadata.yaml
-        return str(f), None
+        return str(f)
 
     @try_export
     def export_coreml(self, prefix=colorstr("CoreML:")):
@@ -910,13 +910,13 @@ class Exporter:
             )
             f = f.with_suffix(".mlmodel")
             ct_model.save(str(f))
-        return f, ct_model
+        return f
 
     @try_export
     def export_engine(self, dla=None, prefix=colorstr("TensorRT:")):
         """Export YOLO model to TensorRT format https://developer.nvidia.com/tensorrt."""
         assert self.im.device.type != "cpu", "export running on CPU but must be on GPU, i.e. use 'device=0'"
-        f_onnx, _ = self.export_onnx()  # run before TRT import https://github.com/ultralytics/ultralytics/issues/7016
+        f_onnx = self.export_onnx()  # run before TRT import https://github.com/ultralytics/ultralytics/issues/7016
 
         try:
             import tensorrt as trt  # noqa
@@ -946,7 +946,7 @@ class Exporter:
             prefix=prefix,
         )
 
-        return f, None
+        return f
 
     @try_export
     def export_saved_model(self, prefix=colorstr("TensorFlow SavedModel:")):
@@ -991,7 +991,7 @@ class Exporter:
 
         # Export to ONNX
         self.args.simplify = True
-        f_onnx, _ = self.export_onnx()
+        f_onnx = self.export_onnx()
 
         # Export to TF
         np_data = None
@@ -1051,7 +1051,7 @@ class Exporter:
         frozen_func = convert_variables_to_constants_v2(m)
         frozen_func.graph.as_graph_def()
         tf.io.write_graph(graph_or_graph_def=frozen_func.graph, logdir=str(f.parent), name=f.name, as_text=False)
-        return f, None
+        return f
 
     @try_export
     def export_tflite(self, prefix=colorstr("TensorFlow Lite:")):
@@ -1067,7 +1067,7 @@ class Exporter:
             f = saved_model / f"{self.file.stem}_float16.tflite"  # fp32 in/out
         else:
             f = saved_model / f"{self.file.stem}_float32.tflite"
-        return str(f), None
+        return str(f)
 
     @try_export
     def export_edgetpu(self, tflite_model="", prefix=colorstr("Edge TPU:")):
@@ -1102,7 +1102,7 @@ class Exporter:
         LOGGER.info(f"{prefix} running '{cmd}'")
         subprocess.run(cmd, shell=True)
         self._add_tflite_metadata(f)
-        return f, None
+        return f
 
     @try_export
     def export_tfjs(self, prefix=colorstr("TensorFlow.js:")):
@@ -1135,7 +1135,7 @@ class Exporter:
 
         # Add metadata
         YAML.save(Path(f) / "metadata.yaml", self.metadata)  # add metadata.yaml
-        return f, None
+        return f
 
     @try_export
     def export_rknn(self, prefix=colorstr("RKNN:")):
@@ -1151,7 +1151,7 @@ class Exporter:
 
         from rknn.api import RKNN
 
-        f, _ = self.export_onnx()
+        f = self.export_onnx()
         export_path = Path(f"{Path(f).stem}_rknn_model")
         export_path.mkdir(exist_ok=True)
 
@@ -1162,7 +1162,7 @@ class Exporter:
         f = f.replace(".onnx", f"-{self.args.name}.rknn")
         rknn.export_rknn(f"{export_path / f}")
         YAML.save(export_path / "metadata.yaml", self.metadata)
-        return export_path, None
+        return export_path
 
     @try_export
     def export_imx(self, prefix=colorstr("IMX:")):
@@ -1339,7 +1339,7 @@ class Exporter:
         with open(f / "labels.txt", "w", encoding="utf-8") as file:
             file.writelines([f"{name}\n" for _, name in self.model.names.items()])
 
-        return f, None
+        return f
 
     def _add_tflite_metadata(self, file):
         """Add metadata to *.tflite models per https://ai.google.dev/edge/litert/models/metadata."""

ultralytics/engine/predictor.py

@@ -409,8 +409,7 @@ class BasePredictor:
         if hasattr(self.model, "imgsz") and not getattr(self.model, "dynamic", False):
             self.args.imgsz = self.model.imgsz  # reuse imgsz from export metadata
         self.model.eval()
-        if self.args.compile:
-            self.model = attempt_compile(self.model, device=self.device)
+        self.model = attempt_compile(self.model, device=self.device, mode=self.args.compile)
 
     def write_results(self, i: int, p: Path, im: torch.Tensor, s: list[str]) -> str:
         """

ultralytics/engine/results.py

@@ -900,7 +900,7 @@ class Boxes(BaseTensor):
         Args:
             boxes (torch.Tensor | np.ndarray): A tensor or numpy array with detection boxes of shape
                 (num_boxes, 6) or (num_boxes, 7). Columns should contain
-                [x1, y1, x2, y2, confidence, class, (optional) track_id].
+                [x1, y1, x2, y2, (optional) track_id, confidence, class].
             orig_shape (tuple[int, int]): The original image shape as (height, width). Used for normalization.
 
         Attributes:

ultralytics/engine/trainer.py

@@ -42,6 +42,7 @@ from ultralytics.utils.autobatch import check_train_batch_size
 from ultralytics.utils.checks import check_amp, check_file, check_imgsz, check_model_file_from_stem, print_args
 from ultralytics.utils.dist import ddp_cleanup, generate_ddp_command
 from ultralytics.utils.files import get_latest_run
+from ultralytics.utils.plotting import plot_results
 from ultralytics.utils.torch_utils import (
     TORCH_2_4,
     EarlyStopping,
@@ -261,8 +262,7 @@ class BaseTrainer:
             self.model.criterion = self.model.init_criterion()
 
         # Compile model
-        if self.args.compile:
-            self.model = attempt_compile(self.model, device=self.device)
+        self.model = attempt_compile(self.model, device=self.device, mode=self.args.compile)
 
         # Freeze layers
         freeze_list = (
@@ -414,7 +414,7 @@ class BaseTrainer:
                     batch = self.preprocess_batch(batch)
                     # decouple inference and loss calculations for torch.compile convenience
                     preds = self.model(batch["img"])
-                    loss, self.loss_items = self.model.loss(batch, preds)
+                    loss, self.loss_items = unwrap_model(self.model).loss(batch, preds)
                     self.loss = loss.sum()
                     if RANK != -1:
                         self.loss *= world_size
@@ -743,8 +743,8 @@ class BaseTrainer:
             f.write(s + ("%.6g," * n % tuple([self.epoch + 1, t] + vals)).rstrip(",") + "\n")
 
     def plot_metrics(self):
-        """Plot and display metrics visually."""
-        pass
+        """Plot metrics from a CSV file."""
+        plot_results(file=self.csv, on_plot=self.on_plot)  # save results.png
 
     def on_plot(self, name, data=None):
         """Register plots (e.g. to be consumed in callbacks)."""