Expand description
Dynamic non-AD tensor compute facade for tenferro-rs.
This crate is the public home for runtime-selected dtypes without autodiff. It re-exports the shared dynamic tensor substrate plus runtime management.
§Examples
use tenferro_dynamic_compute::{ScalarType, Tensor};
use tenferro_tensor::{MemoryOrder, Tensor as DenseTensor};
let dense = DenseTensor::<f64>::from_slice(&[1.0, 2.0], &[2], MemoryOrder::ColumnMajor)?;
let value: Tensor = dense.into();
assert_eq!(value.scalar_type(), ScalarType::F64);Modules§
- snapshot
- Dynamic primal snapshots shared across tenferro surface crates.
Structs§
- Default
Runtime Guard - Guard returned by
set_default_runtime. - Structured
Tensor - AD-capable structured tensor wrapper shared by dynamic tenferro frontends.
Enums§
- Compute
Device - Compute device that can execute tensor operations.
- Error
- Shared error type for dynamic
tenferrosurface crates. - Logical
Memory Space - Logical memory space where tensor data resides.
- Memory
Order - Memory ordering for new allocations.
- Runtime
Context - Runtime execution context used by builder
.run()entry points. - Scalar
Type - Runtime scalar type tag used by dynamic tensor wrappers.
- Scalar
Value - Dynamic scalar value extracted from a rank-0 dynamic tensor.
- Tensor
- Runtime tensor wrapper for a fixed supported dtype set.
Functions§
- set_
default_ runtime - Sets the default runtime context for builder
.run(). - with_
default_ runtime - Runs
fwith the default runtime context. - with_
runtime - Runs
fwith an explicitly supplied runtime installed for the duration of the closure.
Type Aliases§
- Result
- Convenience result alias for
tenferrosurface errors.