Helper to create DALI pipelines that return GPU tensors with sequential values and are iterating over virtual sharded dataset. For example se

Sets the internal state of the generator. Equivalent to assigning to :attr:`state`. Provided so that an :class:`RNG` exposes the same

Returns the outputs of the pipeline. Main difference to :meth:`outputs` is that share_outputs doesn't release returned buffers, relea

Simple symbol form.

Store a fragment to memory

Hyperbolic tangent implementation. \param arg function argument \return function value stored in single-preicision

Returns the tensor on the specified device. If the tensor already resides on the device specified, the function will return `self`

Convert a ``PIL.Image`` to a float32 CHW ``torch.Tensor`` with values in [0, 1]. Mirrors ``torchvision.transforms.v2.functional.to_tensor``

We expect the tested operation to be generalized over `fn` an `ndd` APIs by taking the appropriate module as the first argument. This and

See :func:use_fn_api

Zero fp32 and fp16 parameter grads.

* @brief Creates a placeholder for an External Source operator with the given name * (and output of given name). * * Equivalent to inserting

* @brief Channel information is necessary only for inner convolution, for outer it's always 1 */

* @brief Get pointer to the start of pixel `idx` in the buffer, taking into account the cyclic * wrapping */

* @brief Prepare TensorListShape based on `in_shape` number of channels and spatial output * dimensions provided in `output_sizes`. */

* @brief Get the start frame for a given sample (default is 0) */

* @brief Get the total number of frames for a given sample */

* @brief true if there is a constant explicitly provided */

* @brief Scalar-like nodes are the Constant nodes and Tensor nodes that consist of batch of * scalars. */

Constructs a DLTensorResource WITHOUT data ownership. */