Crate tch

Modules§

data: Dataset iterators.
display
index: Indexing operations
jit: JIT interface to run model trained/saved using PyTorch Python API.
kind: The different kind of elements supported in Torch.
nn: A small neural-network library based on Torch.
utils
vision: The vision module groups functions and models related to computer vision.

CModule: A jit PyTorch module.
COptimizer
NewAxis
NoGradGuard: A RAII guard that prevents gradient tracking until deallocated.
Scalar: A single scalar value.
Tensor: A tensor object.
TrainableCModule: The trainable version of a jit PyTorch module.

Cuda: Cuda related helper functions.
Device: A torch device.
IValue: Argument and output values for JIT models. These represent arbitrary values, e.g. tensors, atomic values, pairs of values, etc.
Kind: The different kind of elements that a Tensor can hold.
Layout: A tensor layout.
QEngine: Quantization engines
Reduction
TchError: Main library error type.
TensorIndexer

autocast: Runs a closure in mixed precision.
get_num_interop_threads: Get the number of threads used by torch for inter-op parallelism.
get_num_threads: Get the number of threads used by torch in parallel regions.
manual_seed: Sets the random seed used by torch.
no_grad: Runs a closure without keeping track of gradients.
no_grad_guard: Disables gradient tracking, this will be enabled back when the returned value gets deallocated. Note that it is important to bind this to a name like _guard and not to _ as the latter would immediately drop the guard. See https://internals.rust-lang.org/t/pre-rfc-must-bind/12658/46 for more details.
set_num_interop_threads: Set the number of threads used by torch for inter-op parallelism.
set_num_threads: Set the number of threads used by torch in parallel regions.
with_grad: Runs a closure explicitly keeping track of gradients, this could be run within a no_grad closure for example.