dgenerate-ultralytics-headless 8.3.209__py3-none-any.whl → 8.3.213__py3-none-any.whl

{dgenerate_ultralytics_headless-8.3.209.dist-info → dgenerate_ultralytics_headless-8.3.213.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: dgenerate-ultralytics-headless
-Version: 8.3.209
+Version: 8.3.213
 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-8.3.209.dist-info → dgenerate_ultralytics_headless-8.3.213.dist-info}/RECORD

@@ -1,14 +1,14 @@
-dgenerate_ultralytics_headless-8.3.209.dist-info/licenses/LICENSE,sha256=DZak_2itbUtvHzD3E7GNUYSRK6jdOJ-GqncQ2weavLA,34523
+dgenerate_ultralytics_headless-8.3.213.dist-info/licenses/LICENSE,sha256=DZak_2itbUtvHzD3E7GNUYSRK6jdOJ-GqncQ2weavLA,34523
 tests/__init__.py,sha256=b4KP5_q-2IO8Br8YHOSLYnn7IwZS81l_vfEF2YPa2lM,894
 tests/conftest.py,sha256=LXtQJcFNWPGuzauTGkiXgsvVC3llJKfg22WcmhRzuQc,2593
 tests/test_cli.py,sha256=0jqS6RfzmJeqgjozUqfT4AoP2d_IhUR0Ej-5ToQBK7A,5463
 tests/test_cuda.py,sha256=6zUSwu3xaYiO3RRNyDkNsuyeq47b1e9f6JNhPZVeDL4,8142
-tests/test_engine.py,sha256=8W4_D48ZBUp-DsUlRYxHTXzougycY8yggvpbVwQDLPg,5025
+tests/test_engine.py,sha256=80S2SwcybVZUKNyAXQAR763rRIQUVly2lmP096azoz0,5730
 tests/test_exports.py,sha256=3o-qqPrPqjD1a_U6KBvwAusZ_Wy6S1WzmuvgRRUXmcA,11099
-tests/test_integrations.py,sha256=kl_AKmE_Qs1GB0_91iVwbzNxofm_hFTt0zzU6JF-pg4,6323
-tests/test_python.py,sha256=0D0VVzIk8Bdc-0qRUv5Wq3ASGOWl7fwWUtYzT9kyZ1I,28210
+tests/test_integrations.py,sha256=ehRcYMpGvUI3KvgsaT1pkN1rXkr7tDSlYYMqIcXyGbg,6220
+tests/test_python.py,sha256=x2q5Wx3eOl32ymmr_4p6srz7ebO-O8zFttuerys_OWg,28083
 tests/test_solutions.py,sha256=oaTz5BttPDIeHkQh9oEaw-O73L4iYDP3Lfe82V7DeKM,13416
-ultralytics/__init__.py,sha256=VzOgK7z21m8S3V2PbacldF7dGIiveV4sYpshEMU8F9Q,1302
+ultralytics/__init__.py,sha256=1UAecgYYPQuMuPeNmOT8HW8hSQkiu5Z6jQ9yehBWUqo,1302
 ultralytics/py.typed,sha256=la67KBlbjXN-_-DfGNcdOcjYumVpKG_Tkw-8n5dnGB4,8
 ultralytics/assets/bus.jpg,sha256=wCAZxJecGR63Od3ZRERe9Aja1Weayrb9Ug751DS_vGM,137419
 ultralytics/assets/zidane.jpg,sha256=Ftc4aeMmen1O0A3o6GCDO9FlfBslLpTAw0gnetx7bts,50427
@@ -20,15 +20,15 @@ ultralytics/cfg/datasets/DOTAv1.yaml,sha256=JrDuYcQ0JU9lJlCA-dCkMNko_jaj6MAVGHjs
 ultralytics/cfg/datasets/GlobalWheat2020.yaml,sha256=dnr_loeYSE6Eo_f7V1yubILsMRBMRm1ozyC5r7uT-iY,2144
 ultralytics/cfg/datasets/HomeObjects-3K.yaml,sha256=xEtSqEad-rtfGuIrERjjhdISggmPlvaX-315ZzKz50I,934
 ultralytics/cfg/datasets/ImageNet.yaml,sha256=GvDWypLVG_H3H67Ai8IC1pvK6fwcTtF5FRhzO1OXXDU,42530
-ultralytics/cfg/datasets/Objects365.yaml,sha256=eMQuA8B4ZGp_GsmMNKFP4CziMSVduyuAK1IANkAZaJw,9367
+ultralytics/cfg/datasets/Objects365.yaml,sha256=8Bl-NAm0mlMW8EfMsz39JZo-HCvmp0ejJXaMeoHTpqw,9649
 ultralytics/cfg/datasets/SKU-110K.yaml,sha256=xvRkq3SdDOwBA91U85bln7HTXkod5MvFX6pt1PxTjJE,2609
-ultralytics/cfg/datasets/VOC.yaml,sha256=NhVLvsmLOwMIteW4DPKxetURP5bTaJvYc7w08-HYAUs,3785
-ultralytics/cfg/datasets/VisDrone.yaml,sha256=vIEBrCJLrKg8zYu5imnA5XQKrXwOpVKyaLvoz5oKAG8,3581
+ultralytics/cfg/datasets/VOC.yaml,sha256=84BaL-iwG03M_W9hNzjgEQi918dZgSHbCgf9DShjwLA,3747
+ultralytics/cfg/datasets/VisDrone.yaml,sha256=PfudojW5av_5q-dC9VsG_xhvuv9cTGEpRp4loXCJ4Ng,3397
 ultralytics/cfg/datasets/african-wildlife.yaml,sha256=SuloMp9WAZBigGC8az-VLACsFhTM76_O29yhTvUqdnU,915
 ultralytics/cfg/datasets/brain-tumor.yaml,sha256=qrxPO_t9wxbn2kHFwP3vGTzSWj2ELTLelUwYL3_b6nc,800
 ultralytics/cfg/datasets/carparts-seg.yaml,sha256=A4e9hM1unTY2jjZIXGiKSarF6R-Ad9R99t57OgRJ37w,1253
-ultralytics/cfg/datasets/coco-pose.yaml,sha256=UYEY90XjHxTEYsUMXZXXaxzxs31zRun-PLTMRo1i334,1623
-ultralytics/cfg/datasets/coco.yaml,sha256=iptVWzO1gLRPs76Mrs1Sp4yjYAR4f3AYeoUwP0r4UKw,2606
+ultralytics/cfg/datasets/coco-pose.yaml,sha256=9qc7Fwvt5Qz4hWCMvIRQX4sEYkMLfLpvc-SLpsy_ySc,1601
+ultralytics/cfg/datasets/coco.yaml,sha256=woUMk6L3G3DMQDcThIKouZMcjTI5vP9XUdEVrzYGL50,2584
 ultralytics/cfg/datasets/coco128-seg.yaml,sha256=knBS2enqHzQj5R5frU4nJdxKsFFBhq8TQ1G1JNiaz9s,1982
 ultralytics/cfg/datasets/coco128.yaml,sha256=ok_dzaBUzSd0DWfe531GT_uYTEoF5mIQcgoMHZyIVIA,1965
 ultralytics/cfg/datasets/coco8-grayscale.yaml,sha256=8v6G6mOzZHQNdQM1YwdTBW_lsWWkLRnAimwZBHKtJg8,1961
@@ -42,7 +42,7 @@ ultralytics/cfg/datasets/dog-pose.yaml,sha256=sRU1JDtEC4nLVf2vkn7lxbp4ILWNcgE-ok
 ultralytics/cfg/datasets/dota8-multispectral.yaml,sha256=2lMBi1Q3_pc0auK00yX80oF7oUMo0bUlwjkOrp33hvs,1216
 ultralytics/cfg/datasets/dota8.yaml,sha256=5n4h_4zdrtUSkmH5DHJ-JLPvfiATcieIkgP3NeOP5nI,1060
 ultralytics/cfg/datasets/hand-keypoints.yaml,sha256=6JF2wwrfAfaVb5M_yLmXyv7iIFXtAt91FqS-Q3kJda0,990
-ultralytics/cfg/datasets/lvis.yaml,sha256=nEQgUdSdBcTYW3LzdK2ba3k8SK-p7NNgZ-SoCXf5vns,29703
+ultralytics/cfg/datasets/lvis.yaml,sha256=lMvPfuiDv_o2qLxAWoh9WMrvjKJ5moLrcx1gr3RG_pM,29680
 ultralytics/cfg/datasets/medical-pills.yaml,sha256=RK7iQFpDDkUS6EsEGqlbFjoohi3cgSsUIbsk7UItyds,792
 ultralytics/cfg/datasets/open-images-v7.yaml,sha256=wK9v3OAGdHORkFdqoBi0hS0fa1b74LLroAzUSWjxEqw,12119
 ultralytics/cfg/datasets/package-seg.yaml,sha256=V4uyTDWWzgft24y9HJWuELKuZ5AndAHXbanxMI6T8GU,849
@@ -111,12 +111,12 @@ ultralytics/data/annotator.py,sha256=f15TCDEM8SuuzHiFB8oyhTy9vfywKmPTLSPAgsZQP9I
 ultralytics/data/augment.py,sha256=7NsRCYu_uM6KkpU0F03NC9Ra_GQVGp2dRO1RksrrU38,132897
 ultralytics/data/base.py,sha256=gWoGFifyNe1TCwtGdGp5jzKOQ9sh4b-XrfyN0PPvRaY,19661
 ultralytics/data/build.py,sha256=cdhD1Z4Gv9KLi5n9OchDRBH8rfMQ1NyDja_D7DmAS00,11879
-ultralytics/data/converter.py,sha256=N1YFD0mG7uwL12wMcuVtF2zbISBIzTsGiy1QioDTDGs,32049
+ultralytics/data/converter.py,sha256=HMJ5H7nvHkeeSYNEwcWrSDkPJykVVg3kLmTC_V8adqg,31967
 ultralytics/data/dataset.py,sha256=GL6J_fvluaF2Ck1in3W5q3Xm7lRcUd6Amgd_uu6r_FM,36772
 ultralytics/data/loaders.py,sha256=sfQ0C86uBg9QQbN3aU0W8FIjGQmMdJTQAMK4DA1bjk8,31748
 ultralytics/data/split.py,sha256=5ubnL_wsEutFQOj4I4K01L9UpZrrO_vO3HrydSLJyIY,5107
 ultralytics/data/split_dota.py,sha256=Lz04qVufTvHn4cTyo3VkqoIM93rb-Ymr8uOIXeSsaJI,12910
-ultralytics/data/utils.py,sha256=rrHphhNcAT29Xpulg2RqvU4UlcLN3cPmsXvT7UvAXb0,36979
+ultralytics/data/utils.py,sha256=uXnt0yo0kUSIKS1uR6_vSjuVRIijdLxPmnDmiPevKUA,36923
 ultralytics/data/scripts/download_weights.sh,sha256=0y8XtZxOru7dVThXDFUXLHBuICgOIqZNUwpyL4Rh6lg,595
 ultralytics/data/scripts/get_coco.sh,sha256=UuJpJeo3qQpTHVINeOpmP0NYmg8PhEFE3A8J3jKrnPw,1768
 ultralytics/data/scripts/get_coco128.sh,sha256=qmRQl_hOKrsdHrTrnyQuFIH01oDz3lfaz138OgGfLt8,650
@@ -126,7 +126,7 @@ ultralytics/engine/exporter.py,sha256=BFzmv7tn2e9zUPwFspb677o1QzzJlOfcVyl3gXmVGW
 ultralytics/engine/model.py,sha256=uX6cTFdlLllGRbz8Lr90IZGb4OrtMDIHQEg7DxUqwe8,53449
 ultralytics/engine/predictor.py,sha256=4lfw2RbBDE7939011FcSCuznscrcnMuabZtc8GXaKO4,22735
 ultralytics/engine/results.py,sha256=uQ_tgvdxKAg28pRgb5WCHiqx9Ktu7wYiVbwZy_IJ5bo,71499
-ultralytics/engine/trainer.py,sha256=OQZWfG2PFm8O6N6fwFmTOgkGeRSR5gSGjfy9NWNnKnQ,41178
+ultralytics/engine/trainer.py,sha256=cd1Qq0SxToCLh7NWIRKKTyWZ-rGQGi3TjwKZ0u02gWk,43529
 ultralytics/engine/tuner.py,sha256=8uiZ9DSYdjHmbhfiuzbMPw--1DLS3cpfZPeSzJ9dGEA,21664
 ultralytics/engine/validator.py,sha256=s7cKMqj2HgVm-GL9bUc76QBeue2jb4cKPk-uQQG5nck,16949
 ultralytics/hub/__init__.py,sha256=xCF02lzlPKbdmGfO3NxLuXl5Kb0MaBZp_-fAWDHZ8zw,6698
@@ -198,7 +198,7 @@ ultralytics/models/yolo/yoloe/train_seg.py,sha256=aCV7M8oQOvODFnU4piZdJh3tIrBJYA
 ultralytics/models/yolo/yoloe/val.py,sha256=5Gd9EoFH0FmKKvWXBl4J7gBe9DVxIczN-s3ceHwdUDo,9458
 ultralytics/nn/__init__.py,sha256=PJgOn2phQTTBR2P3s_JWvGeGXQpvw1znsumKow4tCuE,545
 ultralytics/nn/autobackend.py,sha256=Fs4gjgfCzR9mSpvZpnNXh1V1WWaUEap6oEZeSg5R4Hw,41270
-ultralytics/nn/tasks.py,sha256=1hz7w60SNYk7T5TRWBOPup-mbAqCJDgZ91rv9cheqdc,70379
+ultralytics/nn/tasks.py,sha256=r01JGRa9bgGdOHXycN6TSK30I_Ip4GHO9dZ8LtpkmYk,70846
 ultralytics/nn/text_model.py,sha256=pHqnKe8UueR1MuwJcIE_IvrnYIlt68QL796xjcRJs2A,15275
 ultralytics/nn/modules/__init__.py,sha256=BPMbEm1daI7Tuds3zph2_afAX7Gq1uAqK8BfiCfKTZs,3198
 ultralytics/nn/modules/activation.py,sha256=75JcIMH2Cu9GTC2Uf55r_5YLpxcrXQDaVoeGQ0hlUAU,2233
@@ -240,11 +240,11 @@ ultralytics/trackers/utils/matching.py,sha256=I8SX0sBaBgr4GBJ9uDGOy5LnotgNZHpB2p
 ultralytics/utils/__init__.py,sha256=whSIuj-0lV0SAp4YjOeBJZ2emP1Qa8pqLnrhRiwl2Qs,53503
 ultralytics/utils/autobatch.py,sha256=i6KYLLSItKP1Q2IUlTPHrZhjcxl7UOjs0Seb8bF8pvM,5124
 ultralytics/utils/autodevice.py,sha256=d9yq6eEn05fdfzfpxeSECd0YEO61er5f7T-0kjLdofg,8843
-ultralytics/utils/benchmarks.py,sha256=L0EAWMTYmH-vvPp-mGkxaMXzKghmuWW766DSipm7wJM,31504
-ultralytics/utils/checks.py,sha256=R3Fm3N3B5k4a_ATWMnOVuwFAFeSYaWmQm2yoNYh3MvU,36304
+ultralytics/utils/benchmarks.py,sha256=wZQBGfQdIUcTSzz9jV8iGeRY6QGA1H_YUnbhoZCLV1E,32118
+ultralytics/utils/checks.py,sha256=X9UpXHPaMjqX37CGVdtK1ZsT25yszWc1s7eIRuqxZzU,36270
 ultralytics/utils/cpu.py,sha256=OPlVxROWhQp-kEa9EkeNRKRQ-jz0KwySu5a-h91JZjk,3634
 ultralytics/utils/dist.py,sha256=5xQhWK0OLORvseAL08UmG1LYdkiDVLquxmaGSnqiSqo,4151
-ultralytics/utils/downloads.py,sha256=mlebo09sFXKp1s_BsPza3bQXIj07pV2VnAeEdj4SX0k,23038
+ultralytics/utils/downloads.py,sha256=VmWgrcywhnIUcZYZOeZW0maksh0veo5WjorG7HwWKeE,22928
 ultralytics/utils/errors.py,sha256=XT9Ru7ivoBgofK6PlnyigGoa7Fmf5nEhyHtnD-8TRXI,1584
 ultralytics/utils/events.py,sha256=v2RmLlx78_K6xQfOAuUTJMOexAgNdiuiOvvnsH65oDA,4679
 ultralytics/utils/files.py,sha256=kxE2rkBuZL288nSN7jxLljmDnBgc16rekEXeRjhbUoo,8213
@@ -276,8 +276,8 @@ ultralytics/utils/callbacks/tensorboard.py,sha256=_4nfGK1dDLn6ijpvphBDhc-AS8qhS3
 ultralytics/utils/callbacks/wb.py,sha256=ngQO8EJ1kxJDF1YajScVtzBbm26jGuejA0uWeOyvf5A,7685
 ultralytics/utils/export/__init__.py,sha256=jQtf716PP0jt7bMoY9FkqmjG26KbvDzuR84jGhaBi2U,9901
 ultralytics/utils/export/imx.py,sha256=Jl5nuNxqaP_bY5yrV2NypmoJSrexHE71TxR72SDdjcg,11394
-dgenerate_ultralytics_headless-8.3.209.dist-info/METADATA,sha256=YrSWWjO2-hcF3REbM4efT_YGRGFJom1HSWaDTv1GYpc,38763
-dgenerate_ultralytics_headless-8.3.209.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-dgenerate_ultralytics_headless-8.3.209.dist-info/entry_points.txt,sha256=YM_wiKyTe9yRrsEfqvYolNO5ngwfoL4-NwgKzc8_7sI,93
-dgenerate_ultralytics_headless-8.3.209.dist-info/top_level.txt,sha256=XP49TwiMw4QGsvTLSYiJhz1xF_k7ev5mQ8jJXaXi45Q,12
-dgenerate_ultralytics_headless-8.3.209.dist-info/RECORD,,
+dgenerate_ultralytics_headless-8.3.213.dist-info/METADATA,sha256=mk_6ygTQTklQy5ZH3VW6WrkHkyy15KKFRsPNI-6oZzs,38763
+dgenerate_ultralytics_headless-8.3.213.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+dgenerate_ultralytics_headless-8.3.213.dist-info/entry_points.txt,sha256=YM_wiKyTe9yRrsEfqvYolNO5ngwfoL4-NwgKzc8_7sI,93
+dgenerate_ultralytics_headless-8.3.213.dist-info/top_level.txt,sha256=XP49TwiMw4QGsvTLSYiJhz1xF_k7ev5mQ8jJXaXi45Q,12
+dgenerate_ultralytics_headless-8.3.213.dist-info/RECORD,,

tests/test_engine.py

@@ -3,6 +3,8 @@
 import sys
 from unittest import mock
 
+import torch
+
 from tests import MODEL
 from ultralytics import YOLO
 from ultralytics.cfg import get_cfg
@@ -136,3 +138,20 @@ def test_classify():
     assert test_func in pred.callbacks["on_predict_start"], "callback test failed"
     result = pred(source=ASSETS, model=trainer.best)
     assert len(result), "predictor test failed"
+
+
+def test_nan_recovery():
+    """Test NaN loss detection and recovery during training."""
+    nan_injected = [False]
+
+    def inject_nan(trainer):
+        """Inject NaN into loss during batch processing to test recovery mechanism."""
+        if trainer.epoch == 1 and trainer.tloss is not None and not nan_injected[0]:
+            trainer.tloss *= torch.tensor(float("nan"))
+            nan_injected[0] = True
+
+    overrides = {"data": "coco8.yaml", "model": "yolo11n.yaml", "imgsz": 32, "epochs": 3}
+    trainer = detect.DetectionTrainer(overrides=overrides)
+    trainer.add_callback("on_train_batch_end", inject_nan)
+    trainer.train()
+    assert nan_injected[0], "NaN injection failed"

tests/test_integrations.py

@@ -10,7 +10,7 @@ import pytest
 
 from tests import MODEL, SOURCE, TMP
 from ultralytics import YOLO, download
-from ultralytics.utils import DATASETS_DIR, SETTINGS
+from ultralytics.utils import ASSETS_URL, DATASETS_DIR, SETTINGS
 from ultralytics.utils.checks import check_requirements
 
 
@@ -129,26 +129,23 @@ def test_faster_coco_eval():
     from ultralytics.models.yolo.pose import PoseValidator
     from ultralytics.models.yolo.segment import SegmentationValidator
 
-    # Download annotations after each dataset downloads first
-    url = "https://github.com/ultralytics/assets/releases/download/v0.0.0/"
-
     args = {"model": "yolo11n.pt", "data": "coco8.yaml", "save_json": True, "imgsz": 64}
     validator = DetectionValidator(args=args)
     validator()
     validator.is_coco = True
-    download(f"{url}instances_val2017.json", dir=DATASETS_DIR / "coco8/annotations")
+    download(f"{ASSETS_URL}/instances_val2017.json", dir=DATASETS_DIR / "coco8/annotations")
     _ = validator.eval_json(validator.stats)
 
     args = {"model": "yolo11n-seg.pt", "data": "coco8-seg.yaml", "save_json": True, "imgsz": 64}
     validator = SegmentationValidator(args=args)
     validator()
     validator.is_coco = True
-    download(f"{url}instances_val2017.json", dir=DATASETS_DIR / "coco8-seg/annotations")
+    download(f"{ASSETS_URL}/instances_val2017.json", dir=DATASETS_DIR / "coco8-seg/annotations")
     _ = validator.eval_json(validator.stats)
 
     args = {"model": "yolo11n-pose.pt", "data": "coco8-pose.yaml", "save_json": True, "imgsz": 64}
     validator = PoseValidator(args=args)
     validator()
     validator.is_coco = True
-    download(f"{url}person_keypoints_val2017.json", dir=DATASETS_DIR / "coco8-pose/annotations")
+    download(f"{ASSETS_URL}/person_keypoints_val2017.json", dir=DATASETS_DIR / "coco8-pose/annotations")
     _ = validator.eval_json(validator.stats)

tests/test_python.py

@@ -20,8 +20,11 @@ from ultralytics.data.utils import check_det_dataset
 from ultralytics.utils import (
     ARM64,
     ASSETS,
+    ASSETS_URL,
     DEFAULT_CFG,
     DEFAULT_CFG_PATH,
+    IS_JETSON,
+    IS_RASPBERRYPI,
     LINUX,
     LOGGER,
     ONLINE,
@@ -125,9 +128,7 @@ def test_predict_img(model_name):
     batch = [
         str(SOURCE),  # filename
         Path(SOURCE),  # Path
-        "https://github.com/ultralytics/assets/releases/download/v0.0.0/zidane.jpg?token=123"
-        if ONLINE
-        else SOURCE,  # URI
+        f"{ASSETS_URL}/zidane.jpg?token=123" if ONLINE else SOURCE,  # URI
         im,  # OpenCV
         Image.open(SOURCE),  # PIL
         np.zeros((320, 640, channels), dtype=np.uint8),  # numpy
@@ -190,7 +191,7 @@ def test_track_stream(model):
     """
     if model == "yolo11n-cls.pt":  # classification model not supported for tracking
         return
-    video_url = "https://github.com/ultralytics/assets/releases/download/v0.0.0/decelera_portrait_min.mov"
+    video_url = f"{ASSETS_URL}/decelera_portrait_min.mov"
     model = YOLO(model)
     model.track(video_url, imgsz=160, tracker="bytetrack.yaml")
     model.track(video_url, imgsz=160, tracker="botsort.yaml", save_frames=True)  # test frame saving also
@@ -218,6 +219,7 @@ def test_val(task: str, weight: str, data: str) -> None:
         metrics.confusion_matrix.to_json()
 
 
+@pytest.mark.skipif(IS_JETSON or IS_RASPBERRYPI, reason="Edge devices not intended for training")
 def test_train_scratch():
     """Test training the YOLO model from scratch using the provided configuration."""
     model = YOLO(CFG)
@@ -229,9 +231,7 @@ def test_train_scratch():
 def test_train_ndjson():
     """Test training the YOLO model using NDJSON format dataset."""
     model = YOLO(WEIGHTS_DIR / "yolo11n.pt")
-    model.train(
-        data="https://github.com/ultralytics/assets/releases/download/v0.0.0/coco8-ndjson.ndjson", epochs=1, imgsz=32
-    )
+    model.train(data=f"{ASSETS_URL}/coco8-ndjson.ndjson", epochs=1, imgsz=32)
 
 
 @pytest.mark.parametrize("scls", [False, True])
@@ -290,8 +290,8 @@ def test_predict_callback_and_setup():
 @pytest.mark.parametrize("model", MODELS)
 def test_results(model: str):
     """Test YOLO model results processing and output in various formats."""
-    temp_s = "https://ultralytics.com/images/boats.jpg" if model == "yolo11n-obb.pt" else SOURCE
-    results = YOLO(WEIGHTS_DIR / model)([temp_s, temp_s], imgsz=160)
+    im = f"{ASSETS_URL}/boats.jpg" if model == "yolo11n-obb.pt" else SOURCE
+    results = YOLO(WEIGHTS_DIR / model)([im, im], imgsz=160)
     for r in results:
         assert len(r), f"'{model}' results should not be empty!"
         r = r.cpu().numpy()
@@ -357,8 +357,7 @@ def test_data_converter():
     """Test dataset conversion functions from COCO to YOLO format and class mappings."""
     from ultralytics.data.converter import coco80_to_coco91_class, convert_coco
 
-    file = "instances_val2017.json"
-    download(f"https://github.com/ultralytics/assets/releases/download/v0.0.0/{file}", dir=TMP)
+    download(f"{ASSETS_URL}/instances_val2017.json", dir=TMP)
     convert_coco(labels_dir=TMP, save_dir=TMP / "yolo_labels", use_segments=True, use_keypoints=False, cls91to80=True)
     coco80_to_coco91_class()
 

ultralytics/__init__.py

@@ -1,6 +1,6 @@
 # Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
 
-__version__ = "8.3.209"
+__version__ = "8.3.213"
 
 import importlib
 import os

ultralytics/cfg/datasets/Objects365.yaml

@@ -384,6 +384,7 @@ names:
 
 # Download script/URL (optional) ---------------------------------------------------------------------------------------
 download: |
+  from concurrent.futures import ThreadPoolExecutor
   from pathlib import Path
 
   import numpy as np
@@ -396,31 +397,28 @@ download: |
   check_requirements("faster-coco-eval")
   from faster_coco_eval import COCO
 
-  # Make Directories
-  dir = Path(yaml["path"])  # dataset root dir
-  for p in "images", "labels":
-      (dir / p).mkdir(parents=True, exist_ok=True)
-      for q in "train", "val":
-          (dir / p / q).mkdir(parents=True, exist_ok=True)
-
   # Train, Val Splits
+  dir = Path(yaml["path"])
   for split, patches in [("train", 50 + 1), ("val", 43 + 1)]:
       print(f"Processing {split} in {patches} patches ...")
       images, labels = dir / "images" / split, dir / "labels" / split
+      images.mkdir(parents=True, exist_ok=True)
+      labels.mkdir(parents=True, exist_ok=True)
 
       # Download
       url = f"https://dorc.ks3-cn-beijing.ksyun.com/data-set/2020Objects365%E6%95%B0%E6%8D%AE%E9%9B%86/{split}/"
       if split == "train":
           download([f"{url}zhiyuan_objv2_{split}.tar.gz"], dir=dir)  # annotations json
-          download([f"{url}patch{i}.tar.gz" for i in range(patches)], dir=images, curl=True, threads=8)
+          download([f"{url}patch{i}.tar.gz" for i in range(patches)], dir=images, threads=17)  # 51 patches / 17 threads = 3
       elif split == "val":
           download([f"{url}zhiyuan_objv2_{split}.json"], dir=dir)  # annotations json
-          download([f"{url}images/v1/patch{i}.tar.gz" for i in range(15 + 1)], dir=images, curl=True, threads=8)
-          download([f"{url}images/v2/patch{i}.tar.gz" for i in range(16, patches)], dir=images, curl=True, threads=8)
+          download([f"{url}images/v1/patch{i}.tar.gz" for i in range(15 + 1)], dir=images, threads=16)
+          download([f"{url}images/v2/patch{i}.tar.gz" for i in range(16, patches)], dir=images, threads=16)
 
       # Move
-      for f in TQDM(images.rglob("*.jpg"), desc=f"Moving {split} images"):
-          f.rename(images / f.name)  # move to /images/{split}
+      files = list(images.rglob("*.jpg"))
+      with ThreadPoolExecutor(max_workers=16) as executor:
+          list(TQDM(executor.map(lambda f: f.rename(images / f.name), files), total=len(files), desc=f"Moving {split} images"))
 
       # Labels
       coco = COCO(dir / f"zhiyuan_objv2_{split}.json")
@@ -428,10 +426,12 @@ download: |
       for cid, cat in enumerate(names):
           catIds = coco.getCatIds(catNms=[cat])
           imgIds = coco.getImgIds(catIds=catIds)
-          for im in TQDM(coco.loadImgs(imgIds), desc=f"Class {cid + 1}/{len(names)} {cat}"):
-              width, height = im["width"], im["height"]
-              path = Path(im["file_name"])  # image filename
+
+          def process_annotation(im):
+              """Process and write annotations for a single image."""
               try:
+                  width, height = im["width"], im["height"]
+                  path = Path(im["file_name"])
                   with open(labels / path.with_suffix(".txt").name, "a", encoding="utf-8") as file:
                       annIds = coco.getAnnIds(imgIds=im["id"], catIds=catIds, iscrowd=None)
                       for a in coco.loadAnns(annIds):
@@ -441,3 +441,7 @@ download: |
                           file.write(f"{cid} {x:.5f} {y:.5f} {w:.5f} {h:.5f}\n")
               except Exception as e:
                   print(e)
+
+          images_list = coco.loadImgs(imgIds)
+          with ThreadPoolExecutor(max_workers=16) as executor:
+              list(TQDM(executor.map(process_annotation, images_list), total=len(images_list), desc=f"Class {cid + 1}/{len(names)} {cat}"))

ultralytics/cfg/datasets/VOC.yaml

@@ -49,7 +49,7 @@ download: |
   from pathlib import Path
 
   from ultralytics.utils.downloads import download
-  from ultralytics.utils import TQDM
+  from ultralytics.utils import ASSETS_URL, TQDM
 
   def convert_label(path, lb_path, year, image_id):
       """Converts XML annotations from VOC format to YOLO format by extracting bounding boxes and class IDs."""
@@ -79,11 +79,10 @@ download: |
 
   # Download
   dir = Path(yaml["path"])  # dataset root dir
-  url = "https://github.com/ultralytics/assets/releases/download/v0.0.0/"
   urls = [
-      f"{url}VOCtrainval_06-Nov-2007.zip",  # 446MB, 5012 images
-      f"{url}VOCtest_06-Nov-2007.zip",  # 438MB, 4953 images
-      f"{url}VOCtrainval_11-May-2012.zip",  # 1.95GB, 17126 images
+      f"{ASSETS_URL}/VOCtrainval_06-Nov-2007.zip",  # 446MB, 5012 images
+      f"{ASSETS_URL}/VOCtest_06-Nov-2007.zip",  # 438MB, 4953 images
+      f"{ASSETS_URL}/VOCtrainval_11-May-2012.zip",  # 1.95GB, 17126 images
   ]
   download(urls, dir=dir / "images", threads=3, exist_ok=True)  # download and unzip over existing (required)
 

ultralytics/cfg/datasets/VisDrone.yaml

@@ -34,7 +34,7 @@ download: |
   import shutil
 
   from ultralytics.utils.downloads import download
-  from ultralytics.utils import TQDM
+  from ultralytics.utils import ASSETS_URL, TQDM
 
 
   def visdrone2yolo(dir, split, source_name=None):
@@ -73,10 +73,10 @@ download: |
   # Download (ignores test-challenge split)
   dir = Path(yaml["path"])  # dataset root dir
   urls = [
-      "https://github.com/ultralytics/assets/releases/download/v0.0.0/VisDrone2019-DET-train.zip",
-      "https://github.com/ultralytics/assets/releases/download/v0.0.0/VisDrone2019-DET-val.zip",
-      "https://github.com/ultralytics/assets/releases/download/v0.0.0/VisDrone2019-DET-test-dev.zip",
-      # "https://github.com/ultralytics/assets/releases/download/v0.0.0/VisDrone2019-DET-test-challenge.zip",
+      f"{ASSETS_URL}/VisDrone2019-DET-train.zip",
+      f"{ASSETS_URL}/VisDrone2019-DET-val.zip",
+      f"{ASSETS_URL}/VisDrone2019-DET-test-dev.zip",
+      # f"{ASSETS_URL}/VisDrone2019-DET-test-challenge.zip",
   ]
   download(urls, dir=dir, threads=4)
 

ultralytics/cfg/datasets/coco-pose.yaml

@@ -26,12 +26,13 @@ names:
 download: |
   from pathlib import Path
 
+  from ultralytics.utils import ASSETS_URL
   from ultralytics.utils.downloads import download
 
   # Download labels
   dir = Path(yaml["path"])  # dataset root dir
-  url = "https://github.com/ultralytics/assets/releases/download/v0.0.0/"
-  urls = [f"{url}coco2017labels-pose.zip"]
+
+  urls = [f"{ASSETS_URL}/coco2017labels-pose.zip"]
   download(urls, dir=dir.parent)
   # Download data
   urls = [

ultralytics/cfg/datasets/coco.yaml

@@ -101,13 +101,13 @@ names:
 download: |
   from pathlib import Path
 
+  from ultralytics.utils import ASSETS_URL
   from ultralytics.utils.downloads import download
 
   # Download labels
   segments = True  # segment or box labels
   dir = Path(yaml["path"])  # dataset root dir
-  url = "https://github.com/ultralytics/assets/releases/download/v0.0.0/"
-  urls = [url + ("coco2017labels-segments.zip" if segments else "coco2017labels.zip")]  # labels
+  urls = [ASSETS_URL + ("/coco2017labels-segments.zip" if segments else "/coco2017labels.zip")]  # labels
   download(urls, dir=dir.parent)
   # Download data
   urls = [

ultralytics/cfg/datasets/lvis.yaml

@@ -1223,12 +1223,12 @@ names:
 download: |
   from pathlib import Path
 
+  from ultralytics.utils import ASSETS_URL
   from ultralytics.utils.downloads import download
 
   # Download labels
   dir = Path(yaml["path"])  # dataset root dir
-  url = "https://github.com/ultralytics/assets/releases/download/v0.0.0/"
-  urls = [f"{url}lvis-labels-segments.zip"]
+  urls = [f"{ASSETS_URL}/lvis-labels-segments.zip"]
   download(urls, dir=dir.parent)
 
   # Download data

ultralytics/data/converter.py

@@ -14,7 +14,7 @@ import cv2
 import numpy as np
 from PIL import Image
 
-from ultralytics.utils import DATASETS_DIR, LOGGER, NUM_THREADS, TQDM, YAML
+from ultralytics.utils import ASSETS_URL, DATASETS_DIR, LOGGER, NUM_THREADS, TQDM, YAML
 from ultralytics.utils.checks import check_file, check_requirements
 from ultralytics.utils.downloads import download, zip_directory
 from ultralytics.utils.files import increment_path
@@ -678,9 +678,7 @@ def create_synthetic_coco_dataset():
 
     # Download labels
     dir = DATASETS_DIR / "coco"
-    url = "https://github.com/ultralytics/assets/releases/download/v0.0.0/"
-    label_zip = "coco2017labels-segments.zip"
-    download([url + label_zip], dir=dir.parent)
+    download([f"{ASSETS_URL}/coco2017labels-segments.zip"], dir=dir.parent)
 
     # Create synthetic images
     shutil.rmtree(dir / "labels" / "test2017", ignore_errors=True)  # Remove test2017 directory as not needed

ultralytics/data/utils.py

@@ -19,6 +19,7 @@ from PIL import Image, ImageOps
 
 from ultralytics.nn.autobackend import check_class_names
 from ultralytics.utils import (
+    ASSETS_URL,
     DATASETS_DIR,
     LOGGER,
     NUM_THREADS,
@@ -523,8 +524,7 @@ def check_cls_dataset(dataset: str | Path, split: str = "") -> dict[str, Any]:
         if str(dataset) == "imagenet":
             subprocess.run(["bash", str(ROOT / "data/scripts/get_imagenet.sh")], check=True)
         else:
-            url = f"https://github.com/ultralytics/assets/releases/download/v0.0.0/{dataset}.zip"
-            download(url, dir=data_dir.parent)
+            download(f"{ASSETS_URL}/{dataset}.zip", dir=data_dir.parent)
         LOGGER.info(f"Dataset download success ✅ ({time.time() - t:.1f}s), saved to {colorstr('bold', data_dir)}\n")
     train_set = data_dir / "train"
     if not train_set.is_dir():

ultralytics/engine/trainer.py

@@ -171,6 +171,7 @@ class BaseTrainer:
         self.loss_names = ["Loss"]
         self.csv = self.save_dir / "results.csv"
         self.plot_idx = [0, 1, 2]
+        self.nan_recovery_attempts = 0
 
         # Callbacks
         self.callbacks = _callbacks or callbacks.get_default_callbacks()
@@ -420,14 +421,10 @@ class BaseTrainer:
                     self.loss = loss.sum()
                     if RANK != -1:
                         self.loss *= self.world_size
-                    self.tloss = (
-                        (self.tloss * i + self.loss_items) / (i + 1) if self.tloss is not None else self.loss_items
-                    )
+                    self.tloss = self.loss_items if self.tloss is None else (self.tloss * i + self.loss_items) / (i + 1)
 
                 # Backward
                 self.scaler.scale(self.loss).backward()
-
-                # Optimize - https://pytorch.org/docs/master/notes/amp_examples.html
                 if ni - last_opt_step >= self.accumulate:
                     self.optimizer_step()
                     last_opt_step = ni
@@ -462,6 +459,7 @@ class BaseTrainer:
                 self.run_callbacks("on_train_batch_end")
 
             self.lr = {f"lr/pg{ir}": x["lr"] for ir, x in enumerate(self.optimizer.param_groups)}  # for loggers
+
             self.run_callbacks("on_train_epoch_end")
             if RANK in {-1, 0}:
                 final_epoch = epoch + 1 >= self.epochs
@@ -471,6 +469,13 @@ class BaseTrainer:
                 if self.args.val or final_epoch or self.stopper.possible_stop or self.stop:
                     self._clear_memory(threshold=0.5)  # prevent VRAM spike
                     self.metrics, self.fitness = self.validate()
+
+            # NaN recovery
+            if self._handle_nan_recovery(epoch):
+                continue
+
+            self.nan_recovery_attempts = 0
+            if RANK in {-1, 0}:
                 self.save_metrics(metrics={**self.label_loss_items(self.tloss), **self.metrics, **self.lr})
                 self.stop |= self.stopper(epoch + 1, self.fitness) or final_epoch
                 if self.args.time:
@@ -556,7 +561,10 @@ class BaseTrainer:
         """Read results.csv into a dictionary using polars."""
         import polars as pl  # scope for faster 'import ultralytics'
 
-        return pl.read_csv(self.csv, infer_schema_length=None).to_dict(as_series=False)
+        try:
+            return pl.read_csv(self.csv, infer_schema_length=None).to_dict(as_series=False)
+        except Exception:
+            return {}
 
     def _model_train(self):
         """Set model in training mode."""
@@ -600,6 +608,7 @@ class BaseTrainer:
         serialized_ckpt = buffer.getvalue()  # get the serialized content to save
 
         # Save checkpoints
+        self.wdir.mkdir(parents=True, exist_ok=True)  # ensure weights directory exists
         self.last.write_bytes(serialized_ckpt)  # save last.pt
         if self.best_fitness == self.fitness:
             self.best.write_bytes(serialized_ckpt)  # save best.pt
@@ -665,7 +674,7 @@ class BaseTrainer:
     def optimizer_step(self):
         """Perform a single step of the training optimizer with gradient clipping and EMA update."""
         self.scaler.unscale_(self.optimizer)  # unscale gradients
-        torch.nn.utils.clip_grad_norm_(self.model.parameters(), max_norm=10.0)  # clip gradients
+        torch.nn.utils.clip_grad_norm_(self.model.parameters(), max_norm=10.0)
         self.scaler.step(self.optimizer)
         self.scaler.update()
         self.optimizer.zero_grad()
@@ -740,8 +749,9 @@ class BaseTrainer:
         """Save training metrics to a CSV file."""
         keys, vals = list(metrics.keys()), list(metrics.values())
         n = len(metrics) + 2  # number of cols
-        s = "" if self.csv.exists() else (("%s," * n % tuple(["epoch", "time"] + keys)).rstrip(",") + "\n")  # header
         t = time.time() - self.train_time_start
+        self.csv.parent.mkdir(parents=True, exist_ok=True)  # ensure parent directory exists
+        s = "" if self.csv.exists() else (("%s," * n % tuple(["epoch", "time"] + keys)).rstrip(",") + "\n")  # header
         with open(self.csv, "a", encoding="utf-8") as f:
             f.write(s + ("%.6g," * n % tuple([self.epoch + 1, t] + vals)).rstrip(",") + "\n")
 
@@ -803,20 +813,52 @@ class BaseTrainer:
                 ) from e
         self.resume = resume
 
-    def resume_training(self, ckpt):
-        """Resume YOLO training from given epoch and best fitness."""
-        if ckpt is None or not self.resume:
-            return
-        best_fitness = 0.0
-        start_epoch = ckpt.get("epoch", -1) + 1
+    def _load_checkpoint_state(self, ckpt):
+        """Load optimizer, scaler, EMA, and best_fitness from checkpoint."""
         if ckpt.get("optimizer") is not None:
-            self.optimizer.load_state_dict(ckpt["optimizer"])  # optimizer
-            best_fitness = ckpt["best_fitness"]
+            self.optimizer.load_state_dict(ckpt["optimizer"])
         if ckpt.get("scaler") is not None:
             self.scaler.load_state_dict(ckpt["scaler"])
         if self.ema and ckpt.get("ema"):
-            self.ema.ema.load_state_dict(ckpt["ema"].float().state_dict())  # EMA
+            self.ema.ema.load_state_dict(ckpt["ema"].float().state_dict())
             self.ema.updates = ckpt["updates"]
+        self.best_fitness = ckpt.get("best_fitness", 0.0)
+
+    def _handle_nan_recovery(self, epoch):
+        """Detect and recover from NaN/Inf loss or fitness collapse by loading last checkpoint."""
+        loss_nan = self.tloss is not None and not torch.isfinite(self.tloss).all()
+        fitness_nan = self.fitness is not None and not np.isfinite(self.fitness)
+        fitness_collapse = self.best_fitness and self.best_fitness > 0 and self.fitness == 0
+        corrupted = RANK in {-1, 0} and (loss_nan or fitness_nan or fitness_collapse)
+        reason = "Loss NaN/Inf" if loss_nan else "Fitness NaN/Inf" if fitness_nan else "Fitness collapse"
+        if RANK != -1:  # DDP: broadcast to all ranks
+            broadcast_list = [corrupted if RANK == 0 else None]
+            dist.broadcast_object_list(broadcast_list, 0)
+            corrupted = broadcast_list[0]
+        if not corrupted:
+            return False
+        if epoch == self.start_epoch or not self.last.exists():
+            LOGGER.warning(f"{reason} detected but can not recover from last.pt...")
+            return False  # Cannot recover on first epoch, let training continue
+        self.nan_recovery_attempts += 1
+        if self.nan_recovery_attempts > 3:
+            raise RuntimeError(f"Training failed: NaN persisted for {self.nan_recovery_attempts} epochs")
+        LOGGER.warning(f"{reason} detected (attempt {self.nan_recovery_attempts}/3), recovering from last.pt...")
+        _, ckpt = load_checkpoint(self.last)
+        ema_state = ckpt["ema"].float().state_dict()
+        if not all(torch.isfinite(v).all() for v in ema_state.values() if isinstance(v, torch.Tensor)):
+            raise RuntimeError(f"Checkpoint {self.last} is corrupted with NaN/Inf weights")
+        unwrap_model(self.model).load_state_dict(ema_state)  # Load EMA weights into model
+        self._load_checkpoint_state(ckpt)  # Load optimizer/scaler/EMA/best_fitness
+        del ckpt, ema_state
+        self.scheduler.last_epoch = epoch - 1
+        return True
+
+    def resume_training(self, ckpt):
+        """Resume YOLO training from given epoch and best fitness."""
+        if ckpt is None or not self.resume:
+            return
+        start_epoch = ckpt.get("epoch", -1) + 1
         assert start_epoch > 0, (
             f"{self.args.model} training to {self.epochs} epochs is finished, nothing to resume.\n"
             f"Start a new training without resuming, i.e. 'yolo train model={self.args.model}'"
@@ -827,7 +869,7 @@ class BaseTrainer:
                 f"{self.model} has been trained for {ckpt['epoch']} epochs. Fine-tuning for {self.epochs} more epochs."
             )
             self.epochs += ckpt["epoch"]  # finetune additional epochs
-        self.best_fitness = best_fitness
+        self._load_checkpoint_state(ckpt)
         self.start_epoch = start_epoch
         if start_epoch > (self.epochs - self.args.close_mosaic):
             self._close_dataloader_mosaic()

ultralytics/nn/tasks.py

@@ -743,6 +743,22 @@ class RTDETRDetectionModel(DetectionModel):
         """
         super().__init__(cfg=cfg, ch=ch, nc=nc, verbose=verbose)
 
+    def _apply(self, fn):
+        """
+        Apply a function to all tensors in the model that are not parameters or registered buffers.
+
+        Args:
+            fn (function): The function to apply to the model.
+
+        Returns:
+            (RTDETRDetectionModel): An updated BaseModel object.
+        """
+        self = super()._apply(fn)
+        m = self.model[-1]
+        m.anchors = fn(m.anchors)
+        m.valid_mask = fn(m.valid_mask)
+        return self
+
     def init_criterion(self):
         """Initialize the loss criterion for the RTDETRDetectionModel."""
         from ultralytics.models.utils.loss import RTDETRDetectionLoss

ultralytics/utils/benchmarks.py

@@ -43,7 +43,7 @@ import torch.cuda
 from ultralytics import YOLO, YOLOWorld
 from ultralytics.cfg import TASK2DATA, TASK2METRIC
 from ultralytics.engine.exporter import export_formats
-from ultralytics.utils import ARM64, ASSETS, IS_JETSON, LINUX, LOGGER, MACOS, TQDM, WEIGHTS_DIR, YAML
+from ultralytics.utils import ARM64, ASSETS, ASSETS_URL, IS_JETSON, LINUX, LOGGER, MACOS, TQDM, WEIGHTS_DIR, YAML
 from ultralytics.utils.checks import IS_PYTHON_3_13, check_imgsz, check_requirements, check_yolo, is_rockchip
 from ultralytics.utils.downloads import safe_download
 from ultralytics.utils.files import file_size
@@ -281,7 +281,7 @@ class RF100Benchmark:
         (shutil.rmtree("rf-100"), os.mkdir("rf-100")) if os.path.exists("rf-100") else os.mkdir("rf-100")
         os.chdir("rf-100")
         os.mkdir("ultralytics-benchmarks")
-        safe_download("https://github.com/ultralytics/assets/releases/download/v0.0.0/datasets_links.txt")
+        safe_download(f"{ASSETS_URL}/datasets_links.txt")
 
         with open(ds_link_txt, encoding="utf-8") as file:
             for line in file:
@@ -583,6 +583,11 @@ class ProfileModels:
         run_times = self.iterative_sigma_clipping(np.array(run_times), sigma=2, max_iters=3)  # sigma clipping
         return np.mean(run_times), np.std(run_times)
 
+    @staticmethod
+    def check_dynamic(tensor_shape):
+        """Check whether the tensor shape in the ONNX model is dynamic."""
+        return not all(isinstance(dim, int) and dim >= 0 for dim in tensor_shape)
+
     def profile_onnx_model(self, onnx_file: str, eps: float = 1e-3):
         """
         Profile an ONNX model, measuring average inference time and standard deviation across multiple runs.
@@ -604,27 +609,36 @@ class ProfileModels:
         sess_options.intra_op_num_threads = 8  # Limit the number of threads
         sess = ort.InferenceSession(onnx_file, sess_options, providers=["CPUExecutionProvider"])
 
-        input_tensor = sess.get_inputs()[0]
-        input_type = input_tensor.type
-        dynamic = not all(isinstance(dim, int) and dim >= 0 for dim in input_tensor.shape)  # dynamic input shape
-        input_shape = (1, 3, self.imgsz, self.imgsz) if dynamic else input_tensor.shape
-
-        # Mapping ONNX datatype to numpy datatype
-        if "float16" in input_type:
-            input_dtype = np.float16
-        elif "float" in input_type:
-            input_dtype = np.float32
-        elif "double" in input_type:
-            input_dtype = np.float64
-        elif "int64" in input_type:
-            input_dtype = np.int64
-        elif "int32" in input_type:
-            input_dtype = np.int32
-        else:
-            raise ValueError(f"Unsupported ONNX datatype {input_type}")
+        input_data_dict = dict()
+        for input_tensor in sess.get_inputs():
+            input_type = input_tensor.type
+            if self.check_dynamic(input_tensor.shape):
+                if len(input_tensor.shape) != 4 and self.check_dynamic(input_tensor.shape[1:]):
+                    raise ValueError(f"Unsupported dynamic shape {input_tensor.shape} of {input_tensor.name}")
+                input_shape = (
+                    (1, 3, self.imgsz, self.imgsz) if len(input_tensor.shape) == 4 else (1, *input_tensor.shape[1:])
+                )
+            else:
+                input_shape = input_tensor.shape
+
+            # Mapping ONNX datatype to numpy datatype
+            if "float16" in input_type:
+                input_dtype = np.float16
+            elif "float" in input_type:
+                input_dtype = np.float32
+            elif "double" in input_type:
+                input_dtype = np.float64
+            elif "int64" in input_type:
+                input_dtype = np.int64
+            elif "int32" in input_type:
+                input_dtype = np.int32
+            else:
+                raise ValueError(f"Unsupported ONNX datatype {input_type}")
+
+            input_data = np.random.rand(*input_shape).astype(input_dtype)
+            input_name = input_tensor.name
+            input_data_dict.update({input_name: input_data})
 
-        input_data = np.random.rand(*input_shape).astype(input_dtype)
-        input_name = input_tensor.name
         output_name = sess.get_outputs()[0].name
 
         # Warmup runs
@@ -632,7 +646,7 @@ class ProfileModels:
         for _ in range(3):
             start_time = time.time()
             for _ in range(self.num_warmup_runs):
-                sess.run([output_name], {input_name: input_data})
+                sess.run([output_name], input_data_dict)
             elapsed = time.time() - start_time
 
         # Compute number of runs as higher of min_time or num_timed_runs
@@ -642,7 +656,7 @@ class ProfileModels:
         run_times = []
         for _ in TQDM(range(num_runs), desc=onnx_file):
             start_time = time.time()
-            sess.run([output_name], {input_name: input_data})
+            sess.run([output_name], input_data_dict)
             run_times.append((time.time() - start_time) * 1000)  # Convert to milliseconds
 
         run_times = self.iterative_sigma_clipping(np.array(run_times), sigma=2, max_iters=5)  # sigma clipping

ultralytics/utils/checks.py

@@ -23,6 +23,7 @@ import torch
 from ultralytics.utils import (
     ARM64,
     ASSETS,
+    ASSETS_URL,
     AUTOINSTALL,
     GIT,
     IS_COLAB,
@@ -336,7 +337,7 @@ def check_font(font="Arial.ttf"):
         return matches[0]
 
     # Download to USER_CONFIG_DIR if missing
-    url = f"https://github.com/ultralytics/assets/releases/download/v0.0.0/{name}"
+    url = f"{ASSETS_URL}/{name}"
     if downloads.is_url(url, check=True):
         downloads.safe_download(url=url, file=file)
         return file

ultralytics/utils/downloads.py

@@ -10,7 +10,7 @@ from multiprocessing.pool import ThreadPool
 from pathlib import Path
 from urllib import parse, request
 
-from ultralytics.utils import LOGGER, TQDM, checks, clean_url, emojis, is_online, url2file
+from ultralytics.utils import ASSETS_URL, LOGGER, TQDM, checks, clean_url, emojis, is_online, url2file
 
 # Define Ultralytics GitHub assets maintained at https://github.com/ultralytics/assets
 GITHUB_ASSETS_REPO = "ultralytics/assets"
@@ -60,10 +60,11 @@ def is_url(url: str | Path, check: bool = False) -> bool:
     try:
         url = str(url)
         result = parse.urlparse(url)
-        assert all([result.scheme, result.netloc])  # check if is url
+        if not (result.scheme and result.netloc):
+            return False
         if check:
-            with request.urlopen(url) as response:
-                return response.getcode() == 200  # check if exists online
+            r = request.urlopen(request.Request(url, method="HEAD"), timeout=3)
+            return 200 <= r.getcode() < 400
         return True
     except Exception:
         return False
@@ -323,10 +324,7 @@ def safe_download(
     if "://" not in str(url) and Path(url).is_file():  # URL exists ('://' check required in Windows Python<3.10)
         f = Path(url)  # filename
     elif not f.is_file():  # URL and file do not exist
-        uri = (url if gdrive else clean_url(url)).replace(  # cleaned and aliased url
-            "https://github.com/ultralytics/assets/releases/download/v0.0.0/",
-            "https://ultralytics.com/assets/",  # assets alias
-        )
+        uri = (url if gdrive else clean_url(url)).replace(ASSETS_URL, "https://ultralytics.com/assets")  # clean
         desc = f"Downloading {uri} to '{f}'"
         f.parent.mkdir(parents=True, exist_ok=True)  # make directory if missing
         curl_installed = shutil.which("curl")