ultralytics/yolov5 · error · RuntimeError

Dataset not found ❌

Error message

Dataset not found ❌

What it means

check_dataset in utils/general.py raises RuntimeError('Dataset not found') when the yaml's val path(s) do not exist on disk AND there is no 'download:' entry (or autodownload=False) to fetch them. The preceding log line lists exactly which paths are missing. This is the top-level dataset gate for train.py/val.py; it only autodownloads when download: is an http..zip URL, a 'bash ...' script, or inline python.

Source

Thrown at utils/general.py:393

        data["path"] = path  # download scripts
    for k in "train", "val", "test":
        if data.get(k):  # prepend path
            if isinstance(data[k], str):
                x = (path / data[k]).resolve()
                if not x.exists() and data[k].startswith("../"):
                    x = (path / data[k][3:]).resolve()
                data[k] = str(x)
            else:
                data[k] = [str((path / x).resolve()) for x in data[k]]

    # Parse yaml
    _train, val, _test, s = (data.get(x) for x in ("train", "val", "test", "download"))
    if val:
        val = [Path(x).resolve() for x in (val if isinstance(val, list) else [val])]  # val path
        if not all(x.exists() for x in val):
            LOGGER.info("\nDataset not found ⚠️, missing paths %s" % [str(x) for x in val if not x.exists()])
            if not s or not autodownload:
                raise RuntimeError("Dataset not found ❌")
            t = time.time()
            if s.startswith("http") and s.endswith(".zip"):  # URL
                download(s, dir=DATASETS_DIR, curl=True)
                r = None  # success
            elif s.startswith("bash "):  # bash script
                LOGGER.info(f"Running {s} ...")
                r = subprocess.run(s, shell=True, check=False).returncode
            else:  # python script
                r = exec(s, {"yaml": data})  # noqa: S102  # return None
            dt = f"({round(time.time() - t, 1)}s)"
            s = f"success ✅ {dt}, saved to {colorstr('bold', DATASETS_DIR)}" if r in (0, None) else f"failure {dt} ❌"
            LOGGER.info(f"Dataset download {s}")
    check_font("Arial.ttf" if is_ascii(data["names"]) else "Arial.Unicode.ttf", progress=True)  # download fonts
    return data  # dictionary


def check_amp(model):
    """Checks PyTorch AMP functionality for a model, returns True if AMP operates correctly, otherwise False."""

View on GitHub (pinned to 20d1d78a08)

Solutions

1. Add a download: URL to the yaml so check_dataset can fetch it (see coco.yaml for the pattern).
2. Or download/extract the dataset manually and correct the train/val/test paths to the extracted locations.
3. Rerun with the same command; the missing paths are logged just above the exception.

Example fix

# before (mydata.yaml)
path: ../datasets/mydata
train: images/train
val: images/val

# after (mydata.yaml)
path: ../datasets/mydata
download: https://example.com/mydata.zip
train: images/train
val: images/val

from pathlib import Path
import yaml

def dataset_ready(data_yaml: str) -> bool:
    p = Path(data_yaml).resolve()
    d = yaml.safe_load(p.read_text())
    base = Path(d.get('path') or p.parent)
    val = d.get('val')
    if not val:
        return False
    vals = val if isinstance(val, list) else [val]
    return all((Path(x) if str(x).startswith('/') else base / x).exists() for x in vals) or bool(d.get('download'))

try:
    data = check_dataset(data_yaml)
except RuntimeError as e:
    if 'Dataset not found' in str(e):
        raise SystemExit('download the dataset or add a download: URL to the yaml') from e

When it happens

Trigger: data yaml with val: ../coco/val2017.txt that is absent and no download: key; passing autodownload=False programmatically while the dataset is missing; a download: field that is an empty string (falsy) so the guard still trips.

Common situations: First-time training runs; yamls for custom datasets created without a download step; CI caches that exclude the datasets dir; users moving DATASETS_DIR without updating yamls.

Related errors

• {prefix}{p} does not exist
• {prefix}Error loading data from {path}: {e}\n{HELP_URL}
• More than one yaml file was found in the dataset root, canno
• No yaml definition found in dataset root path, check that th
• Could not resolve hostname '{hostname}': {e}

AI-assisted analysis of ultralytics/yolov5@20d1d78a08 (2026-08-15). Data as JSON: /api/errors/c96d8b1d6fa4cb38. Report an issue: GitHub.