Re-exports§
pub use config::PrepareConfig;pub use tensor_registry::apply_map_to_tensors;pub use variable::BoundVariable;pub use variable::Variable;
Modules§
- activation
- Activation functions for neural networks.
- arithmetic
- bitwise
- Bitwise operations for integer tensors.
- broadcast
- Broadcasting operations for tensors.
- conditional
- Conditional and selection operations for tensors.
- config
- data
- einsum
- Einstein summation convention.
- error
- indexing
- Indexing operations for Tensors.
- math
- Mathematical operations for tensors.
- matmul
- Matrix multiplication and linear transformations.
- memory_
planner - Memory planner for buffer reuse optimization.
- nn
- Neural network operations: convolution, pooling, normalization.
- rand
- THREEFRY-backed deterministic random tensor generation.
- realize
- Tensor realization (execution) API.
- reduce
- Reduction operations for tensors.
- schedule
- Callable scheduling types and execution.
- shape_
ops - Shape manipulation operations for Tensors.
- tensor_
registry - Global tensor registry for atomic graph substitution.
- traits
- transformer
- Transformer building blocks: embedding, attention, rotary position embeddings.
- variable
- Symbolic variables for dynamic tensor dimensions.
Macros§
- codegen_
tests - Generate one test per codegen backend (Clang, LLVM) from a single test body.
Structs§
- Kernel
Info - Information about a rendered kernel.
- Tensor
- Tensor represents a multi-dimensional array with lazy evaluation.
- Tensor
Arange With Dtype Builder - Use builder syntax to set the inputs and finish with
call(). - Tensor
Cumprod With Builder - Use builder syntax to set the inputs and finish with
call(). - Tensor
Cumsum With Builder - Use builder syntax to set the inputs and finish with
call().
Enums§
- CpuBackend
- CPU backend selection.