dgenerate-ultralytics-headless 8.3.187__tar.gz → 8.3.189__tar.gz

{dgenerate_ultralytics_headless-8.3.187 → dgenerate_ultralytics_headless-8.3.189}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dgenerate-ultralytics-headless
-Version: 8.3.187
+Version: 8.3.189
 Summary: Automatically built Ultralytics package with python-opencv-headless dependency instead of python-opencv
 Author-email: Glenn Jocher <glenn.jocher@ultralytics.com>, Jing Qiu <jing.qiu@ultralytics.com>
 Maintainer-email: Ultralytics <hello@ultralytics.com>
@@ -58,7 +58,8 @@ Requires-Dist: mkdocs-ultralytics-plugin>=0.1.29; extra == "dev"
 Requires-Dist: mkdocs-macros-plugin>=1.0.5; extra == "dev"
 Provides-Extra: export
 Requires-Dist: numpy<2.0.0; extra == "export"
-Requires-Dist: onnx<1.18.0,>=1.12.0; extra == "export"
+Requires-Dist: onnx>=1.12.0; platform_system != "Darwin" and extra == "export"
+Requires-Dist: onnx<1.18.0,>=1.12.0; platform_system == "Darwin" and extra == "export"
 Requires-Dist: coremltools>=8.0; (platform_system != "Windows" and python_version <= "3.13") and extra == "export"
 Requires-Dist: scikit-learn>=1.3.2; (platform_system != "Windows" and python_version <= "3.13") and extra == "export"
 Requires-Dist: openvino>=2024.0.0; extra == "export"

{dgenerate_ultralytics_headless-8.3.187 → dgenerate_ultralytics_headless-8.3.189}/dgenerate_ultralytics_headless.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dgenerate-ultralytics-headless
-Version: 8.3.187
+Version: 8.3.189
 Summary: Automatically built Ultralytics package with python-opencv-headless dependency instead of python-opencv
 Author-email: Glenn Jocher <glenn.jocher@ultralytics.com>, Jing Qiu <jing.qiu@ultralytics.com>
 Maintainer-email: Ultralytics <hello@ultralytics.com>
@@ -58,7 +58,8 @@ Requires-Dist: mkdocs-ultralytics-plugin>=0.1.29; extra == "dev"
 Requires-Dist: mkdocs-macros-plugin>=1.0.5; extra == "dev"
 Provides-Extra: export
 Requires-Dist: numpy<2.0.0; extra == "export"
-Requires-Dist: onnx<1.18.0,>=1.12.0; extra == "export"
+Requires-Dist: onnx>=1.12.0; platform_system != "Darwin" and extra == "export"
+Requires-Dist: onnx<1.18.0,>=1.12.0; platform_system == "Darwin" and extra == "export"
 Requires-Dist: coremltools>=8.0; (platform_system != "Windows" and python_version <= "3.13") and extra == "export"
 Requires-Dist: scikit-learn>=1.3.2; (platform_system != "Windows" and python_version <= "3.13") and extra == "export"
 Requires-Dist: openvino>=2024.0.0; extra == "export"

{dgenerate_ultralytics_headless-8.3.187 → dgenerate_ultralytics_headless-8.3.189}/dgenerate_ultralytics_headless.egg-info/requires.txt

@@ -28,7 +28,6 @@ mkdocs-macros-plugin>=1.0.5
 
 [export]
 numpy<2.0.0
-onnx<1.18.0,>=1.12.0
 openvino>=2024.0.0
 tensorflow<=2.19.0,>=2.0.0
 tensorflowjs>=2.0.0
@@ -39,10 +38,16 @@ h5py!=3.11.0
 [export:platform_machine == "aarch64" and python_version >= "3.9"]
 tensorstore>=0.1.63
 
+[export:platform_system != "Darwin"]
+onnx>=1.12.0
+
 [export:platform_system != "Windows" and python_version <= "3.13"]
 coremltools>=8.0
 scikit-learn>=1.3.2
 
+[export:platform_system == "Darwin"]
+onnx<1.18.0,>=1.12.0
+
 [extra]
 hub-sdk>=0.0.12
 ipython

{dgenerate_ultralytics_headless-8.3.187 → dgenerate_ultralytics_headless-8.3.189}/pyproject.toml

@@ -92,7 +92,8 @@ dev = [
 ]
 export = [
     "numpy<2.0.0", # TF 2.20 compatibility
-    "onnx>=1.12.0,<1.18.0", # ONNX export
+    "onnx>=1.12.0; platform_system != 'Darwin'", # ONNX export
+    "onnx>=1.12.0,<1.18.0; platform_system == 'Darwin'",  # TF inference hanging on MacOS
     "coremltools>=8.0; platform_system != 'Windows' and python_version <= '3.13'", # CoreML supported on macOS and Linux
     "scikit-learn>=1.3.2; platform_system != 'Windows' and python_version <= '3.13'", # CoreML k-means quantization
     "openvino>=2024.0.0",  # OpenVINO export

{dgenerate_ultralytics_headless-8.3.187 → dgenerate_ultralytics_headless-8.3.189}/ultralytics/__init__.py

@@ -1,6 +1,6 @@
 # Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
 
-__version__ = "8.3.187"
+__version__ = "8.3.189"
 
 import os
 

{dgenerate_ultralytics_headless-8.3.187 → dgenerate_ultralytics_headless-8.3.189}/ultralytics/engine/exporter.py

@@ -582,9 +582,9 @@ class Exporter:
     @try_export
     def export_onnx(self, prefix=colorstr("ONNX:")):
         """Export YOLO model to ONNX format."""
-        requirements = ["onnx>=1.12.0,<1.18.0"]
+        requirements = ["onnx>=1.12.0"]
         if self.args.simplify:
-            requirements += ["onnxslim>=0.1.59", "onnxruntime" + ("-gpu" if torch.cuda.is_available() else "")]
+            requirements += ["onnxslim>=0.1.65", "onnxruntime" + ("-gpu" if torch.cuda.is_available() else "")]
         check_requirements(requirements)
         import onnx  # noqa
 
@@ -960,9 +960,9 @@ class Exporter:
                 "sng4onnx>=1.0.1",  # required by 'onnx2tf' package
                 "onnx_graphsurgeon>=0.3.26",  # required by 'onnx2tf' package
                 "ai-edge-litert>=1.2.0,<1.4.0",  # required by 'onnx2tf' package
-                "onnx>=1.12.0,<1.18.0",
+                "onnx>=1.12.0",
                 "onnx2tf>=1.26.3",
-                "onnxslim>=0.1.59",
+                "onnxslim>=0.1.65",
                 "onnxruntime-gpu" if cuda else "onnxruntime",
                 "protobuf>=5",
             ),

{dgenerate_ultralytics_headless-8.3.187 → dgenerate_ultralytics_headless-8.3.189}/ultralytics/models/rtdetr/val.py

@@ -196,7 +196,8 @@ class RTDETRValidator(DetectionValidator):
                 with bounding box coordinates, confidence scores, and class predictions.
             pbatch (Dict[str, Any]): Batch dictionary containing 'imgsz', 'ori_shape', 'ratio_pad', and 'im_file'.
         """
-        stem = Path(pbatch["im_file"]).stem
+        path = Path(pbatch["im_file"])
+        stem = path.stem
         image_id = int(stem) if stem.isnumeric() else stem
         box = predn["bboxes"].clone()
         box[..., [0, 2]] *= pbatch["ori_shape"][1] / self.args.imgsz  # native-space pred
@@ -207,6 +208,7 @@ class RTDETRValidator(DetectionValidator):
             self.jdict.append(
                 {
                     "image_id": image_id,
+                    "file_name": path.name,
                     "category_id": self.class_map[int(c)],
                     "bbox": [round(x, 3) for x in b],
                     "score": round(s, 5),

{dgenerate_ultralytics_headless-8.3.187 → dgenerate_ultralytics_headless-8.3.189}/ultralytics/models/yolo/detect/val.py

@@ -372,8 +372,18 @@ class DetectionValidator(BaseValidator):
             predn (Dict[str, torch.Tensor]): Predictions dictionary containing 'bboxes', 'conf', and 'cls' keys
                 with bounding box coordinates, confidence scores, and class predictions.
             pbatch (Dict[str, Any]): Batch dictionary containing 'imgsz', 'ori_shape', 'ratio_pad', and 'im_file'.
-        """
-        stem = Path(pbatch["im_file"]).stem
+
+        Examples:
+             >>> result = {
+             ...     "image_id": 42,
+             ...     "file_name": "42.jpg",
+             ...     "category_id": 18,
+             ...     "bbox": [258.15, 41.29, 348.26, 243.78],
+             ...     "score": 0.236,
+             ... }
+        """
+        path = Path(pbatch["im_file"])
+        stem = path.stem
         image_id = int(stem) if stem.isnumeric() else stem
         box = ops.xyxy2xywh(predn["bboxes"])  # xywh
         box[:, :2] -= box[:, 2:] / 2  # xy center to top-left corner
@@ -381,6 +391,7 @@ class DetectionValidator(BaseValidator):
             self.jdict.append(
                 {
                     "image_id": image_id,
+                    "file_name": path.name,
                     "category_id": self.class_map[int(c)],
                     "bbox": [round(x, 3) for x in b],
                     "score": round(s, 5),

{dgenerate_ultralytics_headless-8.3.187 → dgenerate_ultralytics_headless-8.3.189}/ultralytics/models/yolo/obb/val.py

@@ -176,7 +176,8 @@ class OBBValidator(DetectionValidator):
             (x, y, w, h, angle) and polygon format (x1, y1, x2, y2, x3, y3, x4, y4) before adding them
             to the JSON dictionary.
         """
-        stem = Path(pbatch["im_file"]).stem
+        path = Path(pbatch["im_file"])
+        stem = path.stem
         image_id = int(stem) if stem.isnumeric() else stem
         rbox = predn["bboxes"]
         poly = ops.xywhr2xyxyxyxy(rbox).view(-1, 8)
@@ -184,6 +185,7 @@ class OBBValidator(DetectionValidator):
             self.jdict.append(
                 {
                     "image_id": image_id,
+                    "file_name": path.name,
                     "category_id": self.class_map[int(c)],
                     "score": round(s, 5),
                     "rbox": [round(x, 3) for x in r],

{dgenerate_ultralytics_headless-8.3.187 → dgenerate_ultralytics_headless-8.3.189}/ultralytics/models/yolo/segment/val.py

@@ -217,9 +217,6 @@ class SegmentationValidator(DetectionValidator):
         Args:
             predn (Dict[str, torch.Tensor]): Predictions containing bboxes, masks, confidence scores, and classes.
             pbatch (Dict[str, Any]): Batch dictionary containing 'imgsz', 'ori_shape', 'ratio_pad', and 'im_file'.
-
-        Examples:
-             >>> result = {"image_id": 42, "category_id": 18, "bbox": [258.15, 41.29, 348.26, 243.78], "score": 0.236}
         """
         from faster_coco_eval.core.mask import encode  # noqa
 

{dgenerate_ultralytics_headless-8.3.187 → dgenerate_ultralytics_headless-8.3.189}/ultralytics/nn/autobackend.py

@@ -14,7 +14,7 @@ import torch
 import torch.nn as nn
 from PIL import Image
 
-from ultralytics.utils import ARM64, IS_JETSON, LINUX, LOGGER, PYTHON_VERSION, ROOT, YAML
+from ultralytics.utils import ARM64, IS_JETSON, LINUX, LOGGER, PYTHON_VERSION, ROOT, YAML, is_jetson
 from ultralytics.utils.checks import check_requirements, check_suffix, check_version, check_yaml, is_rockchip
 from ultralytics.utils.downloads import attempt_download_asset, is_url
 
@@ -194,9 +194,12 @@ class AutoBackend(nn.Module):
 
         # In-memory PyTorch model
         if nn_module:
-            model = weights.to(device)
             if fuse:
-                model = model.fuse(verbose=verbose)
+                if IS_JETSON and is_jetson(jetpack=5):
+                    # Jetson Jetpack5 requires device before fuse https://github.com/ultralytics/ultralytics/pull/21028
+                    weights = weights.to(device)
+                weights = weights.fuse(verbose=verbose)
+            model = weights.to(device)
             if hasattr(model, "kpt_shape"):
                 kpt_shape = model.kpt_shape  # pose-only
             stride = max(int(model.stride.max()), 32)  # model stride

{dgenerate_ultralytics_headless-8.3.187 → dgenerate_ultralytics_headless-8.3.189}/ultralytics/nn/tasks.py

@@ -1500,7 +1500,7 @@ def attempt_load_weights(weights, device=None, inplace=True, fuse=False):
     for w in weights if isinstance(weights, list) else [weights]:
         ckpt, w = torch_safe_load(w)  # load ckpt
         args = {**DEFAULT_CFG_DICT, **ckpt["train_args"]} if "train_args" in ckpt else None  # combined args
-        model = (ckpt.get("ema") or ckpt["model"]).to(device).float()  # FP32 model
+        model = (ckpt.get("ema") or ckpt["model"]).float()  # FP32 model
 
         # Model compatibility updates
         model.args = args  # attach args to model
@@ -1510,7 +1510,7 @@ def attempt_load_weights(weights, device=None, inplace=True, fuse=False):
             model.stride = torch.tensor([32.0])
 
         # Append
-        ensemble.append(model.fuse().eval() if fuse and hasattr(model, "fuse") else model.eval())  # model in eval mode
+        ensemble.append((model.fuse().eval() if fuse and hasattr(model, "fuse") else model.eval()).to(device))
 
     # Module updates
     for m in ensemble.modules():

{dgenerate_ultralytics_headless-8.3.187 → dgenerate_ultralytics_headless-8.3.189}/ultralytics/utils/__init__.py

@@ -12,6 +12,7 @@ import subprocess
 import sys
 import threading
 import time
+from functools import lru_cache
 from pathlib import Path
 from threading import Lock
 from types import SimpleNamespace
@@ -727,14 +728,26 @@ def is_raspberrypi() -> bool:
     return "rpi" in DEVICE_MODEL
 
 
-def is_jetson() -> bool:
+@lru_cache(maxsize=3)
+def is_jetson(jetpack=None) -> bool:
     """
     Determine if the Python environment is running on an NVIDIA Jetson device.
 
+    Args:
+        jetpack (int | None): If specified, check for specific JetPack version (4, 5, 6).
+
     Returns:
         (bool): True if running on an NVIDIA Jetson device, False otherwise.
     """
-    return "tegra" in DEVICE_MODEL
+    if jetson := ("tegra" in DEVICE_MODEL):
+        if jetpack:
+            try:
+                content = open("/etc/nv_tegra_release").read()
+                version_map = {4: "R32", 5: "R35", 6: "R36"}  # JetPack to L4T major version mapping
+                return jetpack in version_map and version_map[jetpack] in content
+            except Exception:
+                return False
+    return jetson
 
 
 def is_online() -> bool:

{dgenerate_ultralytics_headless-8.3.187 → dgenerate_ultralytics_headless-8.3.189}/ultralytics/utils/benchmarks.py

@@ -90,9 +90,13 @@ def benchmark(
 
     import polars as pl  # scope for faster 'import ultralytics'
 
-    pl.Config.set_tbl_cols(10)
-    pl.Config.set_tbl_width_chars(120)
-    pl.Config.set_tbl_hide_dataframe_shape(True)
+    pl.Config.set_tbl_cols(-1)  # Show all columns
+    pl.Config.set_tbl_rows(-1)  # Show all rows
+    pl.Config.set_tbl_width_chars(-1)  # No width limit
+    pl.Config.set_tbl_hide_column_data_types(True)  # Hide data types
+    pl.Config.set_tbl_hide_dataframe_shape(True)  # Hide shape info
+    pl.Config.set_tbl_formatting("ASCII_BORDERS_ONLY_CONDENSED")
+
     device = select_device(device, verbose=False)
     if isinstance(model, (str, Path)):
         model = YOLO(model)
@@ -194,12 +198,14 @@ def benchmark(
 
     # Print results
     check_yolo(device=device)  # print system info
-    df = pl.DataFrame(y, schema=["Format", "Status❔", "Size (MB)", key, "Inference time (ms/im)", "FPS"])
+    df = pl.DataFrame(y, schema=["Format", "Status❔", "Size (MB)", key, "Inference time (ms/im)", "FPS"], orient="row")
+    df = df.with_row_index(" ", offset=1)  # add index info
+    df_display = df.with_columns(pl.all().cast(pl.String).fill_null("-"))
 
     name = model.model_name
     dt = time.time() - t0
     legend = "Benchmarks legend:  - ✅ Success  - ❎ Export passed but validation failed  - ❌️ Export failed"
-    s = f"\nBenchmarks complete for {name} on {data} at imgsz={imgsz} ({dt:.2f}s)\n{legend}\n{df.fill_null('-')}\n"
+    s = f"\nBenchmarks complete for {name} on {data} at imgsz={imgsz} ({dt:.2f}s)\n{legend}\n{df_display}\n"
     LOGGER.info(s)
     with open("benchmarks.log", "a", errors="ignore", encoding="utf-8") as f:
         f.write(s)
@@ -209,7 +215,7 @@ def benchmark(
         floor = verbose  # minimum metric floor to pass, i.e. = 0.29 mAP for YOLOv5n
         assert all(x > floor for x in metrics if not np.isnan(x)), f"Benchmark failure: metric(s) < floor {floor}"
 
-    return df
+    return df_display
 
 
 class RF100Benchmark:

{dgenerate_ultralytics_headless-8.3.187 → dgenerate_ultralytics_headless-8.3.189}/ultralytics/utils/downloads.py

@@ -1,12 +1,13 @@
 # Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
 
+from __future__ import annotations
+
 import re
 import shutil
 import subprocess
 from itertools import repeat
 from multiprocessing.pool import ThreadPool
 from pathlib import Path
-from typing import List, Tuple
 from urllib import parse, request
 
 from ultralytics.utils import LOGGER, TQDM, checks, clean_url, emojis, is_online, url2file
@@ -41,7 +42,7 @@ GITHUB_ASSETS_NAMES = frozenset(
 GITHUB_ASSETS_STEMS = frozenset(k.rpartition(".")[0] for k in GITHUB_ASSETS_NAMES)
 
 
-def is_url(url, check: bool = False) -> bool:
+def is_url(url: str | Path, check: bool = False) -> bool:
     """
     Validate if the given string is a URL and optionally check if the URL exists online.
 
@@ -68,7 +69,7 @@ def is_url(url, check: bool = False) -> bool:
         return False
 
 
-def delete_dsstore(path, files_to_delete=(".DS_Store", "__MACOSX")):
+def delete_dsstore(path: str | Path, files_to_delete: tuple[str, ...] = (".DS_Store", "__MACOSX")) -> None:
     """
     Delete all specified system files in a directory.
 
@@ -91,7 +92,12 @@ def delete_dsstore(path, files_to_delete=(".DS_Store", "__MACOSX")):
             f.unlink()
 
 
-def zip_directory(directory, compress: bool = True, exclude=(".DS_Store", "__MACOSX"), progress: bool = True) -> Path:
+def zip_directory(
+    directory: str | Path,
+    compress: bool = True,
+    exclude: tuple[str, ...] = (".DS_Store", "__MACOSX"),
+    progress: bool = True,
+) -> Path:
     """
     Zip the contents of a directory, excluding specified files.
 
@@ -129,9 +135,9 @@ def zip_directory(directory, compress: bool = True, exclude=(".DS_Store", "__MAC
 
 
 def unzip_file(
-    file,
-    path=None,
-    exclude=(".DS_Store", "__MACOSX"),
+    file: str | Path,
+    path: str | Path | None = None,
+    exclude: tuple[str, ...] = (".DS_Store", "__MACOSX"),
     exist_ok: bool = False,
     progress: bool = True,
 ) -> Path:
@@ -198,8 +204,8 @@ def unzip_file(
 
 
 def check_disk_space(
-    url: str = "https://ultralytics.com/assets/coco8.zip",
-    path=Path.cwd(),
+    file_bytes: int,
+    path: str | Path = Path.cwd(),
     sf: float = 1.5,
     hard: bool = True,
 ) -> bool:
@@ -207,7 +213,7 @@ def check_disk_space(
     Check if there is sufficient disk space to download and store a file.
 
     Args:
-        url (str, optional): The URL to the file.
+        file_bytes (int): The file size in bytes.
         path (str | Path, optional): The path or drive to check the available free space on.
         sf (float, optional): Safety factor, the multiplier for the required free space.
         hard (bool, optional): Whether to throw an error or not on insufficient disk space.
@@ -215,26 +221,14 @@ def check_disk_space(
     Returns:
         (bool): True if there is sufficient disk space, False otherwise.
     """
-    import requests  # slow import
-
-    try:
-        r = requests.head(url)  # response
-        assert r.status_code < 400, f"URL error for {url}: {r.status_code} {r.reason}"  # check response
-    except Exception:
-        return True  # requests issue, default to True
-
-    # Check file size
-    gib = 1 << 30  # bytes per GiB
-    data = int(r.headers.get("Content-Length", 0)) / gib  # file size (GB)
-    total, used, free = (x / gib for x in shutil.disk_usage(path))  # bytes
-
-    if data * sf < free:
+    total, used, free = shutil.disk_usage(path)  # bytes
+    if file_bytes * sf < free:
         return True  # sufficient space
 
     # Insufficient space
     text = (
-        f"Insufficient free disk space {free:.1f} GB < {data * sf:.3f} GB required, "
-        f"Please free {data * sf - free:.1f} GB additional disk space and try again."
+        f"Insufficient free disk space {free >> 30:.3f} GB < {int(file_bytes * sf) >> 30:.3f} GB required, "
+        f"Please free {int(file_bytes * sf - free) >> 30:.3f} GB additional disk space and try again."
     )
     if hard:
         raise MemoryError(text)
@@ -242,7 +236,7 @@ def check_disk_space(
     return False
 
 
-def get_google_drive_file_info(link: str) -> Tuple[str, str]:
+def get_google_drive_file_info(link: str) -> tuple[str, str | None]:
     """
     Retrieve the direct download link and filename for a shareable Google Drive file link.
 
@@ -283,9 +277,9 @@ def get_google_drive_file_info(link: str) -> Tuple[str, str]:
 
 
 def safe_download(
-    url,
-    file=None,
-    dir=None,
+    url: str | Path,
+    file: str | Path | None = None,
+    dir: str | Path | None = None,
     unzip: bool = True,
     delete: bool = False,
     curl: bool = False,
@@ -293,7 +287,7 @@ def safe_download(
     min_bytes: float = 1e0,
     exist_ok: bool = False,
     progress: bool = True,
-):
+) -> Path | str:
     """
     Download files from a URL with options for retrying, unzipping, and deleting the downloaded file. Enhanced with
     robust partial download detection using Content-Length validation.
@@ -335,7 +329,6 @@ def safe_download(
         )
         desc = f"Downloading {uri} to '{f}'"
         f.parent.mkdir(parents=True, exist_ok=True)  # make directory if missing
-        check_disk_space(url, path=f.parent)
         curl_installed = shutil.which("curl")
         for i in range(retry + 1):
             try:
@@ -347,6 +340,9 @@ def safe_download(
                 else:  # urllib download
                     with request.urlopen(url) as response:
                         expected_size = int(response.getheader("Content-Length", 0))
+                        if i == 0 and expected_size > 1048576:
+                            check_disk_space(expected_size, path=f.parent)
+                        buffer_size = max(8192, min(1048576, expected_size // 1000)) if expected_size else 8192
                         with TQDM(
                             total=expected_size,
                             desc=desc,
@@ -356,7 +352,10 @@ def safe_download(
                             unit_divisor=1024,
                         ) as pbar:
                             with open(f, "wb") as f_opened:
-                                for data in response:
+                                while True:
+                                    data = response.read(buffer_size)
+                                    if not data:
+                                        break
                                     f_opened.write(data)
                                     pbar.update(len(data))
 
@@ -371,6 +370,8 @@ def safe_download(
                         else:
                             break  # success
                     f.unlink()  # remove partial downloads
+            except MemoryError:
+                raise  # Re-raise immediately - no point retrying if insufficient disk space
             except Exception as e:
                 if i == 0 and not is_online():
                     raise ConnectionError(emojis(f"❌  Download failure for {uri}. Environment is not online.")) from e
@@ -397,7 +398,7 @@ def get_github_assets(
     repo: str = "ultralytics/assets",
     version: str = "latest",
     retry: bool = False,
-) -> Tuple[str, List[str]]:
+) -> tuple[str, list[str]]:
     """
     Retrieve the specified version's tag and assets from a GitHub repository.
 
@@ -430,7 +431,12 @@ def get_github_assets(
     return data["tag_name"], [x["name"] for x in data["assets"]]  # tag, assets i.e. ['yolo11n.pt', 'yolov8s.pt', ...]
 
 
-def attempt_download_asset(file, repo: str = "ultralytics/assets", release: str = "v8.3.0", **kwargs) -> str:
+def attempt_download_asset(
+    file: str | Path,
+    repo: str = "ultralytics/assets",
+    release: str = "v8.3.0",
+    **kwargs,
+) -> str:
     """
     Attempt to download a file from GitHub release assets if it is not found locally.
 
@@ -482,15 +488,15 @@ def attempt_download_asset(file, repo: str = "ultralytics/assets", release: str
 
 
 def download(
-    url,
-    dir=Path.cwd(),
+    url: str | list[str] | Path,
+    dir: Path = Path.cwd(),
     unzip: bool = True,
     delete: bool = False,
     curl: bool = False,
     threads: int = 1,
     retry: int = 3,
     exist_ok: bool = False,
-):
+) -> None:
     """
     Download files from specified URLs to a given directory.
 

{dgenerate_ultralytics_headless-8.3.187 → dgenerate_ultralytics_headless-8.3.189}/ultralytics/utils/torch_utils.py

@@ -250,68 +250,71 @@ def time_sync():
 
 
 def fuse_conv_and_bn(conv, bn):
-    """Fuse Conv2d() and BatchNorm2d() layers."""
-    fusedconv = (
-        nn.Conv2d(
-            conv.in_channels,
-            conv.out_channels,
-            kernel_size=conv.kernel_size,
-            stride=conv.stride,
-            padding=conv.padding,
-            dilation=conv.dilation,
-            groups=conv.groups,
-            bias=True,
-        )
-        .requires_grad_(False)
-        .to(conv.weight.device)
-    )
+    """
+    Fuse Conv2d and BatchNorm2d layers for inference optimization.
+
+    Args:
+        conv (nn.Conv2d): Convolutional layer to fuse.
+        bn (nn.BatchNorm2d): Batch normalization layer to fuse.
+
+    Returns:
+        (nn.Conv2d): The fused convolutional layer with gradients disabled.
 
-    # Prepare filters
+    Example:
+        >>> conv = nn.Conv2d(3, 16, 3)
+        >>> bn = nn.BatchNorm2d(16)
+        >>> fused_conv = fuse_conv_and_bn(conv, bn)
+    """
+    # Compute fused weights
     w_conv = conv.weight.view(conv.out_channels, -1)
     w_bn = torch.diag(bn.weight.div(torch.sqrt(bn.eps + bn.running_var)))
-    fusedconv.weight.copy_(torch.mm(w_bn, w_conv).view(fusedconv.weight.shape))
+    conv.weight.data = torch.mm(w_bn, w_conv).view(conv.weight.shape)
 
-    # Prepare spatial bias
-    b_conv = (
-        torch.zeros(conv.weight.shape[0], dtype=conv.weight.dtype, device=conv.weight.device)
-        if conv.bias is None
-        else conv.bias
-    )
+    # Compute fused bias
+    b_conv = torch.zeros(conv.out_channels, device=conv.weight.device) if conv.bias is None else conv.bias
     b_bn = bn.bias - bn.weight.mul(bn.running_mean).div(torch.sqrt(bn.running_var + bn.eps))
-    fusedconv.bias.copy_(torch.mm(w_bn, b_conv.reshape(-1, 1)).reshape(-1) + b_bn)
+    fused_bias = torch.mm(w_bn, b_conv.reshape(-1, 1)).reshape(-1) + b_bn
+
+    if conv.bias is None:
+        conv.register_parameter("bias", nn.Parameter(fused_bias))
+    else:
+        conv.bias.data = fused_bias
 
-    return fusedconv
+    return conv.requires_grad_(False)
 
 
 def fuse_deconv_and_bn(deconv, bn):
-    """Fuse ConvTranspose2d() and BatchNorm2d() layers."""
-    fuseddconv = (
-        nn.ConvTranspose2d(
-            deconv.in_channels,
-            deconv.out_channels,
-            kernel_size=deconv.kernel_size,
-            stride=deconv.stride,
-            padding=deconv.padding,
-            output_padding=deconv.output_padding,
-            dilation=deconv.dilation,
-            groups=deconv.groups,
-            bias=True,
-        )
-        .requires_grad_(False)
-        .to(deconv.weight.device)
-    )
+    """
+    Fuse ConvTranspose2d and BatchNorm2d layers for inference optimization.
 
-    # Prepare filters
+    Args:
+        deconv (nn.ConvTranspose2d): Transposed convolutional layer to fuse.
+        bn (nn.BatchNorm2d): Batch normalization layer to fuse.
+
+    Returns:
+        (nn.ConvTranspose2d): The fused transposed convolutional layer with gradients disabled.
+
+    Example:
+        >>> deconv = nn.ConvTranspose2d(16, 3, 3)
+        >>> bn = nn.BatchNorm2d(3)
+        >>> fused_deconv = fuse_deconv_and_bn(deconv, bn)
+    """
+    # Compute fused weights
     w_deconv = deconv.weight.view(deconv.out_channels, -1)
     w_bn = torch.diag(bn.weight.div(torch.sqrt(bn.eps + bn.running_var)))
-    fuseddconv.weight.copy_(torch.mm(w_bn, w_deconv).view(fuseddconv.weight.shape))
+    deconv.weight.data = torch.mm(w_bn, w_deconv).view(deconv.weight.shape)
 
-    # Prepare spatial bias
-    b_conv = torch.zeros(deconv.weight.shape[1], device=deconv.weight.device) if deconv.bias is None else deconv.bias
+    # Compute fused bias
+    b_conv = torch.zeros(deconv.out_channels, device=deconv.weight.device) if deconv.bias is None else deconv.bias
     b_bn = bn.bias - bn.weight.mul(bn.running_mean).div(torch.sqrt(bn.running_var + bn.eps))
-    fuseddconv.bias.copy_(torch.mm(w_bn, b_conv.reshape(-1, 1)).reshape(-1) + b_bn)
+    fused_bias = torch.mm(w_bn, b_conv.reshape(-1, 1)).reshape(-1) + b_bn
+
+    if deconv.bias is None:
+        deconv.register_parameter("bias", nn.Parameter(fused_bias))
+    else:
+        deconv.bias.data = fused_bias
 
-    return fuseddconv
+    return deconv.requires_grad_(False)
 
 
 def model_info(model, detailed=False, verbose=True, imgsz=640):