Macro builder for working with data types
Operator implementations for OxiONNX — 88+ ONNX operators in Pure Rust
Pure Rust ONNX inference engine — zero C/C++ dependencies
Enables type-safe enum variants with shared type tokens across multiple enums, allowing for synchronized variant types and powerful downcasting capabilities between related enums.
Proc macros for dtype_variant
Quantization kernels for all GGUF quantization types
Apple Metal runtime dispatch for MetalTile GPU kernels
Idiomatic Rust wrappers for the NVIDIA CUDA stack (Driver API, Runtime API, NVRTC, cuBLAS, cuDNN, NCCL, NVML, ...). Umbrella crate.
Proc-macro crate providing the #[kernel] DSL for MetalTile GPU kernels
Unified ML op facade for the baracuda CUDA ecosystem. Exposes every primitive an ML framework would expect (union of PyTorch torch.* + nn.functional and JAX lax.* / numpy ops) through a single Plan-based Rust surface, internally dispatching to baracuda-cutlass, the baracuda-* NVIDIA-library wrappers, or bespoke baracuda-kernels-sys kernels.
CUB-backed device-wide reductions, scans, sorts, histograms, and selects, surfaced as an atomr actor compiled per-(op, dtype) via NVRTC against the atomr-accel-cuda Phase 0.6 disk cache.
Backend-agnostic compute-acceleration core. Defines the AccelBackend trait, AccelRef<T> typed pointers, AccelError enum, and CompletionStrategy — the abstraction layer that lets atomr-accel-cuda (NVIDIA), and future ROCm / Metal / oneAPI / Vulkan backends plug into the same actor surface.