dgenerate-ultralytics-headless 8.3.198__tar.gz → 8.3.199__tar.gz

{dgenerate_ultralytics_headless-8.3.198 → dgenerate_ultralytics_headless-8.3.199}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dgenerate-ultralytics-headless
-Version: 8.3.198
+Version: 8.3.199
 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>

{dgenerate_ultralytics_headless-8.3.198 → dgenerate_ultralytics_headless-8.3.199}/dgenerate_ultralytics_headless.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dgenerate-ultralytics-headless
-Version: 8.3.198
+Version: 8.3.199
 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>

{dgenerate_ultralytics_headless-8.3.198 → dgenerate_ultralytics_headless-8.3.199}/tests/test_cuda.py

@@ -68,7 +68,7 @@ def test_export_onnx_matrix(task, dynamic, int8, half, batch, simplify, nms):
         half=half,
         batch=batch,
         simplify=simplify,
-        nms=nms and task != "obb",  # disable NMS for OBB task for now on T4 instance
+        nms=nms,
         device=DEVICES[0],
     )
     YOLO(file)([SOURCE] * batch, imgsz=64 if dynamic else 32, device=DEVICES[0])  # exported model inference
@@ -76,7 +76,6 @@ def test_export_onnx_matrix(task, dynamic, int8, half, batch, simplify, nms):
 
 
 @pytest.mark.slow
-@pytest.mark.skipif(True, reason="CUDA export tests disabled pending additional Ultralytics GPU server availability")
 @pytest.mark.skipif(not DEVICES, reason="No CUDA devices available")
 @pytest.mark.parametrize(
     "task, dynamic, int8, half, batch",
@@ -164,7 +163,7 @@ def test_autobatch():
 
 
 @pytest.mark.slow
-@pytest.mark.skipif(True, reason="Skip for now since T4 instance does not support TensorRT > 10.0")
+@pytest.mark.skipif(not DEVICES, reason="No CUDA devices available")
 def test_utils_benchmarks():
     """Profile YOLO models for performance benchmarks."""
     from ultralytics.utils.benchmarks import ProfileModels

dgenerate_ultralytics_headless-8.3.199/ultralytics/__init__.py

@@ -0,0 +1,43 @@
+# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
+
+__version__ = "8.3.199"
+
+import importlib
+import os
+
+# Set ENV variables (place before imports)
+if not os.environ.get("OMP_NUM_THREADS"):
+    os.environ["OMP_NUM_THREADS"] = "1"  # default for reduced CPU utilization during training
+
+from ultralytics.utils import ASSETS, SETTINGS
+from ultralytics.utils.checks import check_yolo as checks
+from ultralytics.utils.downloads import download
+
+settings = SETTINGS
+
+MODELS = ("YOLO", "YOLOWorld", "YOLOE", "NAS", "SAM", "FastSAM", "RTDETR")
+
+__all__ = (
+    "__version__",
+    "ASSETS",
+    *MODELS,
+    "checks",
+    "download",
+    "settings",
+)
+
+
+def __getattr__(name: str):
+    """Lazy-import model classes on first access."""
+    if name in MODELS:
+        return getattr(importlib.import_module("ultralytics.models"), name)
+    raise AttributeError(f"module {__name__} has no attribute {name}")
+
+
+def __dir__():
+    """Extend dir() to include lazily available model names for IDE autocompletion."""
+    return sorted(set(globals()) | set(MODELS))
+
+
+if __name__ == "__main__":
+    print(__version__)

{dgenerate_ultralytics_headless-8.3.198 → dgenerate_ultralytics_headless-8.3.199}/ultralytics/cfg/default.yaml

@@ -37,7 +37,7 @@ 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"
+compile: False # (bool | str) enable torch.compile() backend='inductor'; True="default", False=off, or "default|reduce-overhead|max-autotune-no-cudagraphs"
 
 # Segmentation
 overlap_mask: True # (bool) merge instance masks into one mask during training (segment only)

{dgenerate_ultralytics_headless-8.3.198 → dgenerate_ultralytics_headless-8.3.199}/ultralytics/engine/exporter.py

@@ -922,7 +922,8 @@ class Exporter:
             import tensorrt as trt  # noqa
         except ImportError:
             if LINUX:
-                check_requirements("tensorrt>7.0.0,!=10.1.0")
+                cuda_version = torch.version.cuda.split(".")[0]
+                check_requirements(f"tensorrt-cu{cuda_version}>7.0.0,!=10.1.0")
             import tensorrt as trt  # noqa
         check_version(trt.__version__, ">=7.0.0", hard=True)
         check_version(trt.__version__, "!=10.1.0", msg="https://github.com/ultralytics/ultralytics/pull/14239")
@@ -1306,7 +1307,7 @@ class Exporter:
                     kpts = outputs[2]  # (bs, max_detections, kpts 17*3)
                     out_kpts = torch.gather(kpts, 1, nms_outputs.indices.unsqueeze(-1).expand(-1, -1, kpts.size(-1)))
                     return nms_outputs.boxes, nms_outputs.scores, nms_outputs.labels, out_kpts
-                return nms_outputs
+                return nms_outputs.boxes, nms_outputs.scores, nms_outputs.labels, nms_outputs.n_valid
 
         quant_model = NMSWrapper(
             model=quant_model,

{dgenerate_ultralytics_headless-8.3.198 → dgenerate_ultralytics_headless-8.3.199}/ultralytics/nn/tasks.py

@@ -1543,8 +1543,8 @@ def parse_model(d, ch, verbose=True):
     max_channels = float("inf")
     nc, act, scales = (d.get(x) for x in ("nc", "activation", "scales"))
     depth, width, kpt_shape = (d.get(x, 1.0) for x in ("depth_multiple", "width_multiple", "kpt_shape"))
+    scale = d.get("scale")
     if scales:
-        scale = d.get("scale")
         if not scale:
             scale = tuple(scales.keys())[0]
             LOGGER.warning(f"no model scale passed. Assuming scale='{scale}'.")

{dgenerate_ultralytics_headless-8.3.198 → dgenerate_ultralytics_headless-8.3.199}/ultralytics/utils/plotting.py

@@ -934,13 +934,14 @@ def plt_color_scatter(v, f, bins: int = 20, cmap: str = "viridis", alpha: float
 
 
 @plt_settings()
-def plot_tune_results(csv_file: str = "tune_results.csv"):
+def plot_tune_results(csv_file: str = "tune_results.csv", exclude_zero_fitness_points: bool = True):
     """
     Plot the evolution results stored in a 'tune_results.csv' file. The function generates a scatter plot for each key
     in the CSV, color-coded based on fitness scores. The best-performing configurations are highlighted on the plots.
 
     Args:
         csv_file (str, optional): Path to the CSV file containing the tuning results.
+        exclude_zero_fitness_points (bool, optional): Don't include points with zero fitness in tuning plots.
 
     Examples:
         >>> plot_tune_results("path/to/tune_results.csv")
@@ -962,6 +963,9 @@ def plot_tune_results(csv_file: str = "tune_results.csv"):
     keys = [x.strip() for x in data.columns][num_metrics_columns:]
     x = data.to_numpy()
     fitness = x[:, 0]  # fitness
+    if exclude_zero_fitness_points:
+        mask = fitness > 0  # exclude zero-fitness points
+        x, fitness = x[mask], fitness[mask]
     j = np.argmax(fitness)  # max fitness index
     n = math.ceil(len(keys) ** 0.5)  # columns and rows in plot
     plt.figure(figsize=(10, 10), tight_layout=True)

{dgenerate_ultralytics_headless-8.3.198 → dgenerate_ultralytics_headless-8.3.199}/ultralytics/utils/torch_utils.py

@@ -1028,7 +1028,7 @@ def attempt_compile(
         use_autocast (bool, optional): Whether to run warmup under autocast on CUDA or MPS devices.
         warmup (bool, optional): Whether to execute a single dummy forward pass to warm up the compiled model.
         mode (bool | str, optional): torch.compile mode. True → "default", False → no compile, or a string like
-            "default", "reduce-overhead", "max-autotune".
+            "default", "reduce-overhead", "max-autotune-no-cudagraphs".
 
     Returns:
         model (torch.nn.Module): Compiled model if compilation succeeds, otherwise the original unmodified model.
@@ -1050,6 +1050,9 @@ def attempt_compile(
         mode = "default"
     prefix = colorstr("compile:")
     LOGGER.info(f"{prefix} starting torch.compile with '{mode}' mode...")
+    if mode == "max-autotune":
+        LOGGER.warning(f"{prefix} mode='{mode}' not recommended, using mode='max-autotune-no-cudagraphs' instead")
+        mode = "max-autotune-no-cudagraphs"
     t0 = time.perf_counter()
     try:
         model = torch.compile(model, mode=mode, backend="inductor")

dgenerate_ultralytics_headless-8.3.198/ultralytics/__init__.py

@@ -1,30 +0,0 @@
-# Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
-
-__version__ = "8.3.198"
-
-import os
-
-# Set ENV variables (place before imports)
-if not os.environ.get("OMP_NUM_THREADS"):
-    os.environ["OMP_NUM_THREADS"] = "1"  # default for reduced CPU utilization during training
-
-from ultralytics.models import NAS, RTDETR, SAM, YOLO, YOLOE, FastSAM, YOLOWorld
-from ultralytics.utils import ASSETS, SETTINGS
-from ultralytics.utils.checks import check_yolo as checks
-from ultralytics.utils.downloads import download
-
-settings = SETTINGS
-__all__ = (
-    "__version__",
-    "ASSETS",
-    "YOLO",
-    "YOLOWorld",
-    "YOLOE",
-    "NAS",
-    "SAM",
-    "FastSAM",
-    "RTDETR",
-    "checks",
-    "download",
-    "settings",
-)