Return the padded dataset length (always a multiple of the alignment unit).

Return the fixed fake dataset length.

Validate that ratios sum to approximately 1.0 and are non-negative.

Return a readable string representation of the wrapper. Returns: Representation including the wrapped transform and type.

Fall back to ``'auto'`` (with a warning) for an unrecognised or wrong-typed ``amp_dtype``. Mixed precision is a best-effort speed/memory opti

Normalize legacy boolean progress_bar values to the new string/None representation. This preserves compatibility with older configs where ``p

Normalise the resume checkpoint value to ``str`` without resolving it. Unlike ``dataset_dir``/``output_dir``, ``resume`` is forwarded verbati

Decorate RF-DETR instance methods that require lazy model device placement. The wrapped method receives the same arguments and return value as th

Convert Ultralytics ``flip_idx`` permutation metadata to flat swap pairs.

Initialize the weights.

Build a ``K.RandomAffine`` from aug_config params. Albumentations ``translate_percent`` is a ``(min, max)`` signed range (e.g. ``(-0.1, 0.1)``).

Build a ``K.ColorJiggle`` from aug_config ``ColorJitter`` params. Note: Kornia >=0.7 uses ``ColorJiggle``; the ``ColorJitter`` alias was adde

Return a SimpleNamespace shaped like the args passed to _load_pretrain_weights_into.

Build a ``K.RandomGaussianNoise`` from aug_config params. Kornia takes a single ``std`` value; we use the upper bound of ``std_range`` as an acce

Build a ``K.RandomGaussianBlur`` from aug_config params. ``blur_limit`` is rounded up to an odd number for the kernel size.

Build a ``K.RandomHorizontalFlip`` from aug_config params.

Build a ``K.ColorJiggle`` from ``RandomBrightnessContrast`` params.

Build a ``K.RandomRotation`` from aug_config params. The ``limit`` parameter may be a scalar (symmetric range) or a tuple.

Return a minimal TrainConfig for use in load_pretrain_weights. Args: tmp_path: Optional pytest tmp_path fixture value.

Build a ``K.RandomVerticalFlip`` from aug_config params.

Inject a fake ai_edge_litert.interpreter into sys.modules and mask lower-priority backends. Mirrors ``_mock_tflite_runtime`` for the first-pr

Mock ``_get_onnx_input_info`` to return a known shape.

Inject a fake tflite_runtime.interpreter into sys.modules and mask ai_edge_litert. ``_create_interpreter`` probes backends in priority order:

Normalize supported device inputs to a canonical torch-style string. Args: v: Device specifier provided by callers. Supported val

Benchmark ONNX Runtime inference for one image and provider list. Creates a fresh ``InferenceSession`` with the requested providers, preprocesses

Suppress all download and file-existence side effects.

Suppress all download and file-existence side effects.

Suppress all download and file-existence side effects.