ultralytics 8.2.63__tar.gz → 8.2.65__tar.gz

{ultralytics-8.2.63/ultralytics.egg-info → ultralytics-8.2.65}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: ultralytics
-Version: 8.2.63
+Version: 8.2.65
 Summary: Ultralytics YOLOv8 for SOTA object detection, multi-object tracking, instance segmentation, pose estimation and image classification.
 Author: Glenn Jocher, Ayush Chaurasia, Jing Qiu
 Maintainer: Glenn Jocher, Ayush Chaurasia, Jing Qiu

{ultralytics-8.2.63 → ultralytics-8.2.65}/tests/test_python.py

@@ -95,7 +95,7 @@ def test_predict_img(model_name):
         Image.open(SOURCE),  # PIL
         np.zeros((320, 640, 3), dtype=np.uint8),  # numpy
     ]
-    assert len(model(batch, imgsz=32, augment=True)) == len(batch)  # multiple sources in a batch
+    assert len(model(batch, imgsz=32)) == len(batch)  # multiple sources in a batch
 
 
 @pytest.mark.parametrize("model", MODELS)

{ultralytics-8.2.63 → ultralytics-8.2.65}/ultralytics/__init__.py

@@ -1,6 +1,6 @@
 # Ultralytics YOLO 🚀, AGPL-3.0 license
 
-__version__ = "8.2.63"
+__version__ = "8.2.65"
 
 import os
 

{ultralytics-8.2.63 → ultralytics-8.2.65}/ultralytics/data/augment.py

@@ -2322,7 +2322,7 @@ def classify_transforms(
     size=224,
     mean=DEFAULT_MEAN,
     std=DEFAULT_STD,
-    interpolation=Image.BILINEAR,
+    interpolation="BILINEAR",
     crop_fraction: float = DEFAULT_CROP_FRACTION,
 ):
     """
@@ -2337,7 +2337,7 @@ def classify_transforms(
             tuple, it defines (height, width).
         mean (tuple): Mean values for each RGB channel used in normalization.
         std (tuple): Standard deviation values for each RGB channel used in normalization.
-        interpolation (int): Interpolation method for resizing.
+        interpolation (str): Interpolation method of either 'NEAREST', 'BILINEAR' or 'BICUBIC'.
         crop_fraction (float): Fraction of the image to be cropped.
 
     Returns:
@@ -2360,7 +2360,7 @@ def classify_transforms(
     # Aspect ratio is preserved, crops center within image, no borders are added, image is lost
     if scale_size[0] == scale_size[1]:
         # Simple case, use torchvision built-in Resize with the shortest edge mode (scalar size arg)
-        tfl = [T.Resize(scale_size[0], interpolation=interpolation)]
+        tfl = [T.Resize(scale_size[0], interpolation=getattr(T.InterpolationMode, interpolation))]
     else:
         # Resize the shortest edge to matching target dim for non-square target
         tfl = [T.Resize(scale_size)]
@@ -2389,7 +2389,7 @@ def classify_augmentations(
     hsv_v=0.4,  # image HSV-Value augmentation (fraction)
     force_color_jitter=False,
     erasing=0.0,
-    interpolation=Image.BILINEAR,
+    interpolation="BILINEAR",
 ):
     """
     Creates a composition of image augmentation transforms for classification tasks.
@@ -2411,7 +2411,7 @@ def classify_augmentations(
         hsv_v (float): Image HSV-Value augmentation factor.
         force_color_jitter (bool): Whether to apply color jitter even if auto augment is enabled.
         erasing (float): Probability of random erasing.
-        interpolation (int): Interpolation method.
+        interpolation (str): Interpolation method of either 'NEAREST', 'BILINEAR' or 'BICUBIC'.
 
     Returns:
         (torchvision.transforms.Compose): A composition of image augmentation transforms.
@@ -2427,6 +2427,7 @@ def classify_augmentations(
         raise TypeError(f"classify_transforms() size {size} must be integer, not (list, tuple)")
     scale = tuple(scale or (0.08, 1.0))  # default imagenet scale range
     ratio = tuple(ratio or (3.0 / 4.0, 4.0 / 3.0))  # default imagenet ratio range
+    interpolation = getattr(T.InterpolationMode, interpolation)
     primary_tfl = [T.RandomResizedCrop(size, scale=scale, ratio=ratio, interpolation=interpolation)]
     if hflip > 0.0:
         primary_tfl.append(T.RandomHorizontalFlip(p=hflip))

{ultralytics-8.2.63 → ultralytics-8.2.65}/ultralytics/data/dataset.py

@@ -431,6 +431,12 @@ class ClassificationDataset:
             self.samples = self.samples[: round(len(self.samples) * args.fraction)]
         self.prefix = colorstr(f"{prefix}: ") if prefix else ""
         self.cache_ram = args.cache is True or str(args.cache).lower() == "ram"  # cache images into RAM
+        if self.cache_ram:
+            LOGGER.warning(
+                "WARNING ⚠️ Classification `cache_ram` training has known memory leak in "
+                "https://github.com/ultralytics/ultralytics/issues/9824, setting `cache_ram=False`."
+            )
+            self.cache_ram = False
         self.cache_disk = str(args.cache).lower() == "disk"  # cache images on hard drive as uncompressed *.npy files
         self.samples = self.verify_images()  # filter out bad images
         self.samples = [list(x) + [Path(x[0]).with_suffix(".npy"), None] for x in self.samples]  # file, index, npy, im

{ultralytics-8.2.63 → ultralytics-8.2.65}/ultralytics/data/loaders.py

@@ -545,7 +545,7 @@ def get_best_youtube_url(url, method="pytube"):
     """
     if method == "pytube":
         # Switched from pytube to pytubefix to resolve https://github.com/pytube/pytube/issues/1954
-        check_requirements("pytubefix==6.3.4")  # bug in 6.4.2 https://github.com/JuanBindez/pytubefix/issues/123
+        check_requirements("pytubefix>=6.5.2")
         from pytubefix import YouTube
 
         streams = YouTube(url).streams.filter(file_extension="mp4", only_video=True)

{ultralytics-8.2.63 → ultralytics-8.2.65}/ultralytics/engine/exporter.py

@@ -885,6 +885,8 @@ class Exporter:
             output_integer_quantized_tflite=self.args.int8,
             quant_type="per-tensor",  # "per-tensor" (faster) or "per-channel" (slower but more accurate)
             custom_input_op_name_np_data_path=np_data,
+            disable_group_convolution=True,  # for end-to-end model compatibility
+            enable_batchmatmul_unfold=True,  # for end-to-end model compatibility
         )
         yaml_save(f / "metadata.yaml", self.metadata)  # add metadata.yaml
 

{ultralytics-8.2.63 → ultralytics-8.2.65}/ultralytics/engine/trainer.py

@@ -43,6 +43,7 @@ from ultralytics.utils.files import get_latest_run
 from ultralytics.utils.torch_utils import (
     EarlyStopping,
     ModelEMA,
+    autocast,
     convert_optimizer_state_dict_to_fp16,
     init_seeds,
     one_cycle,
@@ -376,7 +377,7 @@ class BaseTrainer:
                             x["momentum"] = np.interp(ni, xi, [self.args.warmup_momentum, self.args.momentum])
 
                 # Forward
-                with torch.cuda.amp.autocast(self.amp):
+                with autocast(self.amp):
                     batch = self.preprocess_batch(batch)
                     self.loss, self.loss_items = self.model(batch)
                     if RANK != -1:

{ultralytics-8.2.63 → ultralytics-8.2.65}/ultralytics/models/nas/model.py

@@ -16,7 +16,8 @@ from pathlib import Path
 import torch
 
 from ultralytics.engine.model import Model
-from ultralytics.utils.torch_utils import model_info, smart_inference_mode
+from ultralytics.utils.downloads import attempt_download_asset
+from ultralytics.utils.torch_utils import model_info
 
 from .predict import NASPredictor
 from .val import NASValidator
@@ -49,16 +50,25 @@ class NAS(Model):
         assert Path(model).suffix not in {".yaml", ".yml"}, "YOLO-NAS models only support pre-trained models."
         super().__init__(model, task="detect")
 
-    @smart_inference_mode()
-    def _load(self, weights: str, task: str):
+    def _load(self, weights: str, task=None) -> None:
         """Loads an existing NAS model weights or creates a new NAS model with pretrained weights if not provided."""
         import super_gradients
 
         suffix = Path(weights).suffix
         if suffix == ".pt":
-            self.model = torch.load(weights)
+            self.model = torch.load(attempt_download_asset(weights))
+
         elif suffix == "":
             self.model = super_gradients.training.models.get(weights, pretrained_weights="coco")
+
+        # Override the forward method to ignore additional arguments
+        def new_forward(x, *args, **kwargs):
+            """Ignore additional __call__ arguments."""
+            return self.model._original_forward(x)
+
+        self.model._original_forward = self.model.forward
+        self.model.forward = new_forward
+
         # Standardize model
         self.model.fuse = lambda verbose=True: self.model
         self.model.stride = torch.tensor([32])

{ultralytics-8.2.63 → ultralytics-8.2.65}/ultralytics/models/utils/ops.py

@@ -133,7 +133,7 @@ class HungarianMatcher(nn.Module):
     #     sample_points = torch.cat([a.repeat(b, 1, 1, 1) for a, b in zip(sample_points, num_gts) if b > 0])
     #     tgt_mask = F.grid_sample(tgt_mask, sample_points, align_corners=False).squeeze([1, 2])
     #
-    #     with torch.cuda.amp.autocast(False):
+    #     with torch.amp.autocast("cuda", enabled=False):
     #         # binary cross entropy cost
     #         pos_cost_mask = F.binary_cross_entropy_with_logits(out_mask, torch.ones_like(out_mask), reduction='none')
     #         neg_cost_mask = F.binary_cross_entropy_with_logits(out_mask, torch.zeros_like(out_mask), reduction='none')

{ultralytics-8.2.63 → ultralytics-8.2.65}/ultralytics/models/yolo/detect/val.py

@@ -97,7 +97,7 @@ class DetectionValidator(BaseValidator):
             self.args.iou,
             labels=self.lb,
             multi_label=True,
-            agnostic=self.args.single_cls,
+            agnostic=self.args.single_cls or self.args.agnostic_nms,
             max_det=self.args.max_det,
         )
 

{ultralytics-8.2.63 → ultralytics-8.2.65}/ultralytics/nn/autobackend.py

@@ -587,14 +587,21 @@ class AutoBackend(nn.Module):
                     if x.ndim == 3:  # if task is not classification, excluding masks (ndim=4) as well
                         # Denormalize xywh by image size. See https://github.com/ultralytics/ultralytics/pull/1695
                         # xywh are normalized in TFLite/EdgeTPU to mitigate quantization error of integer models
-                        x[:, [0, 2]] *= w
-                        x[:, [1, 3]] *= h
+                        if x.shape[-1] == 6:  # end-to-end model
+                            x[:, :, [0, 2]] *= w
+                            x[:, :, [1, 3]] *= h
+                        else:
+                            x[:, [0, 2]] *= w
+                            x[:, [1, 3]] *= h
                     y.append(x)
             # TF segment fixes: export is reversed vs ONNX export and protos are transposed
             if len(y) == 2:  # segment with (det, proto) output order reversed
                 if len(y[1].shape) != 4:
                     y = list(reversed(y))  # should be y = (1, 116, 8400), (1, 160, 160, 32)
-                y[1] = np.transpose(y[1], (0, 3, 1, 2))  # should be y = (1, 116, 8400), (1, 32, 160, 160)
+                if y[1].shape[-1] == 6:  # end-to-end model
+                    y = [y[1]]
+                else:
+                    y[1] = np.transpose(y[1], (0, 3, 1, 2))  # should be y = (1, 116, 8400), (1, 32, 160, 160)
             y = [x if isinstance(x, np.ndarray) else x.numpy() for x in y]
 
         # for x in y:

ultralytics-8.2.65/ultralytics/nn/modules/activation.py

@@ -0,0 +1,22 @@
+# Ultralytics YOLO 🚀, AGPL-3.0 license
+"""Activation modules."""
+
+import torch
+import torch.nn as nn
+
+
+class AGLU(nn.Module):
+    """Unified activation function module from https://github.com/kostas1515/AGLU."""
+
+    def __init__(self, device=None, dtype=None) -> None:
+        """Initialize the Unified activation function."""
+        super().__init__()
+        self.act = nn.Softplus(beta=-1.0)
+        self.lambd = nn.Parameter(nn.init.uniform_(torch.empty(1, device=device, dtype=dtype)))  # lambda parameter
+        self.kappa = nn.Parameter(nn.init.uniform_(torch.empty(1, device=device, dtype=dtype)))  # kappa parameter
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        """Compute the forward pass of the Unified activation function."""
+        lam = torch.clamp(self.lambd, min=0.0001)
+        y = torch.exp((1 / lam) * self.act((self.kappa * x) - torch.log(lam)))
+        return y  # for AGLU simply return y * input

{ultralytics-8.2.63 → ultralytics-8.2.65}/ultralytics/solutions/streamlit_inference.py

@@ -69,7 +69,7 @@ def inference(model=None):
     # Add dropdown menu for model selection
     available_models = [x.replace("yolo", "YOLO") for x in GITHUB_ASSETS_STEMS if x.startswith("yolov8")]
     if model:
-        available_models.insert(0, model)
+        available_models.insert(0, model.split(".pt")[0])  # insert model without suffix as *.pt is added later
 
     selected_model = st.sidebar.selectbox("Model", available_models)
     with st.spinner("Model is downloading..."):

{ultralytics-8.2.63 → ultralytics-8.2.65}/ultralytics/utils/__init__.py

@@ -1066,8 +1066,9 @@ TESTS_RUNNING = is_pytest_running() or is_github_action_running()
 set_sentry()
 
 # Apply monkey patches
-from ultralytics.utils.patches import imread, imshow, imwrite, torch_save
+from ultralytics.utils.patches import imread, imshow, imwrite, torch_load, torch_save
 
+torch.load = torch_load
 torch.save = torch_save
 if WINDOWS:
     # Apply cv2 patches for non-ASCII and non-UTF characters in image paths

{ultralytics-8.2.63 → ultralytics-8.2.65}/ultralytics/utils/autobatch.py

@@ -7,7 +7,7 @@ import numpy as np
 import torch
 
 from ultralytics.utils import DEFAULT_CFG, LOGGER, colorstr
-from ultralytics.utils.torch_utils import profile
+from ultralytics.utils.torch_utils import autocast, profile
 
 
 def check_train_batch_size(model, imgsz=640, amp=True, batch=-1):
@@ -23,7 +23,7 @@ def check_train_batch_size(model, imgsz=640, amp=True, batch=-1):
         (int): Optimal batch size computed using the autobatch() function.
     """
 
-    with torch.cuda.amp.autocast(amp):
+    with autocast(enabled=amp):
         return autobatch(deepcopy(model).train(), imgsz, fraction=batch if 0.0 < batch < 1.0 else 0.6)
 
 

{ultralytics-8.2.63 → ultralytics-8.2.65}/ultralytics/utils/benchmarks.py

@@ -100,9 +100,11 @@ def benchmark(
                 assert not is_end2end, "End-to-end models not supported by CoreML and TF.js yet"
             if i in {3, 5}:  # CoreML and OpenVINO
                 assert not IS_PYTHON_3_12, "CoreML and OpenVINO not supported on Python 3.12"
-            if i in {6, 7, 8, 9, 10}:  # All TF formats
+            if i in {6, 7, 8}:  # TF SavedModel, TF GraphDef, and TFLite
                 assert not isinstance(model, YOLOWorld), "YOLOWorldv2 TensorFlow exports not supported by onnx2tf yet"
-                assert not is_end2end, "End-to-end models not supported by onnx2tf yet"
+            if i in {9, 10}:  # TF EdgeTPU and TF.js
+                assert not isinstance(model, YOLOWorld), "YOLOWorldv2 TensorFlow exports not supported by onnx2tf yet"
+                assert not is_end2end, "End-to-end models not supported by TF EdgeTPU and TF.js yet"
             if i in {11}:  # Paddle
                 assert not isinstance(model, YOLOWorld), "YOLOWorldv2 Paddle exports not supported yet"
                 assert not is_end2end, "End-to-end models not supported by PaddlePaddle yet"

{ultralytics-8.2.63 → ultralytics-8.2.65}/ultralytics/utils/checks.py

@@ -641,6 +641,8 @@ def check_amp(model):
     Returns:
         (bool): Returns True if the AMP functionality works correctly with YOLOv8 model, else False.
     """
+    from ultralytics.utils.torch_utils import autocast
+
     device = next(model.parameters()).device  # get model device
     if device.type in {"cpu", "mps"}:
         return False  # AMP only used on CUDA devices
@@ -648,7 +650,7 @@ def check_amp(model):
     def amp_allclose(m, im):
         """All close FP32 vs AMP results."""
         a = m(im, device=device, verbose=False)[0].boxes.data  # FP32 inference
-        with torch.cuda.amp.autocast(True):
+        with autocast(enabled=True):
             b = m(im, device=device, verbose=False)[0].boxes.data  # AMP inference
         del m
         return a.shape == b.shape and torch.allclose(a, b.float(), atol=0.5)  # close to 0.5 absolute tolerance

{ultralytics-8.2.63 → ultralytics-8.2.65}/ultralytics/utils/loss.py

@@ -7,6 +7,7 @@ import torch.nn.functional as F
 from ultralytics.utils.metrics import OKS_SIGMA
 from ultralytics.utils.ops import crop_mask, xywh2xyxy, xyxy2xywh
 from ultralytics.utils.tal import RotatedTaskAlignedAssigner, TaskAlignedAssigner, dist2bbox, dist2rbox, make_anchors
+from ultralytics.utils.torch_utils import autocast
 
 from .metrics import bbox_iou, probiou
 from .tal import bbox2dist
@@ -27,7 +28,7 @@ class VarifocalLoss(nn.Module):
     def forward(pred_score, gt_score, label, alpha=0.75, gamma=2.0):
         """Computes varfocal loss."""
         weight = alpha * pred_score.sigmoid().pow(gamma) * (1 - label) + gt_score * label
-        with torch.cuda.amp.autocast(enabled=False):
+        with autocast(enabled=False):
             loss = (
                 (F.binary_cross_entropy_with_logits(pred_score.float(), gt_score.float(), reduction="none") * weight)
                 .mean(1)

{ultralytics-8.2.63 → ultralytics-8.2.65}/ultralytics/utils/patches.py

@@ -57,7 +57,33 @@ def imshow(winname: str, mat: np.ndarray):
 
 
 # PyTorch functions ----------------------------------------------------------------------------------------------------
-_torch_save = torch.save  # copy to avoid recursion errors
+_torch_load = torch.load  # copy to avoid recursion errors
+_torch_save = torch.save
+
+
+def torch_load(*args, **kwargs):
+    """
+    Load a PyTorch model with updated arguments to avoid warnings.
+
+    This function wraps torch.load and adds the 'weights_only' argument for PyTorch 1.13.0+ to prevent warnings.
+
+    Args:
+        *args (Any): Variable length argument list to pass to torch.load.
+        **kwargs (Any): Arbitrary keyword arguments to pass to torch.load.
+
+    Returns:
+        (Any): The loaded PyTorch object.
+
+    Note:
+        For PyTorch versions 2.0 and above, this function automatically sets 'weights_only=False'
+        if the argument is not provided, to avoid deprecation warnings.
+    """
+    from ultralytics.utils.torch_utils import TORCH_1_13
+
+    if TORCH_1_13 and "weights_only" not in kwargs:
+        kwargs["weights_only"] = False
+
+    return _torch_load(*args, **kwargs)
 
 
 def torch_save(*args, use_dill=True, **kwargs):
@@ -68,7 +94,7 @@ def torch_save(*args, use_dill=True, **kwargs):
     Args:
         *args (tuple): Positional arguments to pass to torch.save.
         use_dill (bool): Whether to try using dill for serialization if available. Defaults to True.
-        **kwargs (any): Keyword arguments to pass to torch.save.
+        **kwargs (Any): Keyword arguments to pass to torch.save.
     """
     try:
         assert use_dill

{ultralytics-8.2.63 → ultralytics-8.2.65}/ultralytics/utils/torch_utils.py

@@ -68,6 +68,37 @@ def smart_inference_mode():
     return decorate
 
 
+def autocast(enabled: bool, device: str = "cuda"):
+    """
+    Get the appropriate autocast context manager based on PyTorch version and AMP setting.
+
+    This function returns a context manager for automatic mixed precision (AMP) training that is compatible with both
+    older and newer versions of PyTorch. It handles the differences in the autocast API between PyTorch versions.
+
+    Args:
+        enabled (bool): Whether to enable automatic mixed precision.
+        device (str, optional): The device to use for autocast. Defaults to 'cuda'.
+
+    Returns:
+        (torch.amp.autocast): The appropriate autocast context manager.
+
+    Note:
+        - For PyTorch versions 1.13 and newer, it uses `torch.amp.autocast`.
+        - For older versions, it uses `torch.cuda.autocast`.
+
+    Example:
+        ```python
+        with autocast(amp=True):
+            # Your mixed precision operations here
+            pass
+        ```
+    """
+    if TORCH_1_13:
+        return torch.amp.autocast(device, enabled=enabled)
+    else:
+        return torch.cuda.amp.autocast(enabled)
+
+
 def get_cpu_info():
     """Return a string with system CPU information, i.e. 'Apple M2'."""
     import cpuinfo  # pip install py-cpuinfo

{ultralytics-8.2.63 → ultralytics-8.2.65/ultralytics.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: ultralytics
-Version: 8.2.63
+Version: 8.2.65
 Summary: Ultralytics YOLOv8 for SOTA object detection, multi-object tracking, instance segmentation, pose estimation and image classification.
 Author: Glenn Jocher, Ayush Chaurasia, Jing Qiu
 Maintainer: Glenn Jocher, Ayush Chaurasia, Jing Qiu

{ultralytics-8.2.63 → ultralytics-8.2.65}/ultralytics.egg-info/SOURCES.txt

@@ -185,6 +185,7 @@ ultralytics/nn/__init__.py
 ultralytics/nn/autobackend.py
 ultralytics/nn/tasks.py
 ultralytics/nn/modules/__init__.py
+ultralytics/nn/modules/activation.py
 ultralytics/nn/modules/block.py
 ultralytics/nn/modules/conv.py
 ultralytics/nn/modules/head.py