| rlx-ir cargo | 0.2.18 | none | none | Tensor IR for the RLX ML compiler — standalone, serializable, optimizable | 7.9K | 4h ago |
| rlx-gguf cargo | | – 0.2.18 did not build | – | GGUF v1/v2/v3 parser + dequantization to f32. Standalone — no rlx-* deps. | 7.4K | 4h ago |
| rlx-fusion cargo | | – 0.2.18 did not build | – | Fusion passes and fused-op decomposition for RLX MIR | 7.3K | 4h ago |
| rlx-autodiff cargo | | – 0.2.18 did not build | – | JAX-shaped transforms for RLX MIR — autodiff, JVP/HVP, vmap | 7.2K | 4h ago |
| rlx-compile cargo | | – 0.2.18 did not build | – | HIR → MIR → LIR compile pipeline for RLX | 7.2K | 4h ago |
| rlx-opt cargo | | – 0.2.18 did not build | – | RLX optimizer facade — fusion, autodiff, and compile pipelines | 7.1K | 3h ago |
| rlx-cpu cargo | | – 0.2.18 did not build | – | CPU backend for RLX — SIMD kernels, BLAS dispatch, thread pool, arena executor | 6.8K | 3h ago |
| rlx-driver cargo | | – 0.2.18 did not build | – | RLX driver layer — device handles, arenas, buffers, command streams | 6.2K | 4h ago |
| rlx-macros cargo | 0.2.18 | none | none | RLX proc macros — #[rlx_model] for AOT-compiled ML models | 5.8K | 4h ago |
| rlx-gpu-kernels cargo | | – 0.2.18 did not build | – | Shared CUDA/HIP C++ kernel sources for RLX GPU backends (NVRTC / hipRTC). | 3.9K | 4h ago |
| rlx-wgpu cargo | | – 0.2.18 did not build | – | Cross-platform GPU backend for RLX via wgpu (Metal/Vulkan/DX12/WebGPU) | 3.8K | 3h ago |
| rlx-mlx cargo | | – 0.2.18 did not build | – | MLX backend for RLX — Apple's array framework via hand-rolled C++ shim, eager + lazy execution | 3.8K | 3h ago |
| rlx-metal cargo | | – 0.2.18 did not build | – | Metal backend for RLX — Apple Silicon GPU via Metal Performance Shaders + custom MSL kernels | 3.8K | 3h ago |
| rlx-flow cargo | | – 0.2.18 did not build | – | Block assembly-line API for RLX model builders — fusion-first, config-driven | 3.7K | 4h ago |
| rlx-mlx-sys cargo | 0.2.18 | high | none | Low-level MLX C++ build + C ABI shim for RLX (vendored mlx, NVRTC-free). | 3.7K | 4h ago |
| rlx-cuda cargo | | – 0.2.18 did not build | – | NVIDIA CUDA backend — cuBLAS for matmul + NVRTC-compiled kernels for everything else, via the pure-Rust `cudarc` crate. | 3.7K | 2h ago |
| rlx-rocm cargo | | – 0.2.18 did not build | – | AMD ROCm/HIP backend — same .cu kernel sources as rlx-cuda, dispatched via HIP. | 3.4K | 2h ago |
| rlx-vulkan cargo | | – 0.2.18 did not build | – | Native Vulkan compute backend for RLX (raw `ash` + embedded SPIR-V compute kernels) | 1.5K | 1h ago |
| rlx-tpu cargo | | – 0.2.18 did not build | – | Google TPU backend — drives libtpu's PJRT plugin from Rust. | 1.3K | 1h ago |
| rlx-coreml cargo | | – 0.2.18 did not build | – | Apple CoreML / Neural Engine (ANE) backend for RLX — lowers the IR to an ML Program (MIL) and runs it through CoreML.framework | 1.2K | 1h ago |
| rlx-mlx-io cargo | | – 0.2.18 did not build | – | RLX — load MLX weight layouts (mlx-community safetensors dirs, .npz/.npy) including affine / mxfp dequant. | 1.2K | 4h ago |
| rlx-nemo cargo | | – 0.2.18 did not build | – | RLX — native loader for NVIDIA NeMo `.nemo` model files (tar + torch.save ckpt + YAML config). | 1.1K | 4h ago |
| rlx-dduf cargo | 0.2.18 | none | none | RLX — load HuggingFace DDUF (.dduf) ZIP packages of nested safetensors + JSON. | 797 | 4h ago |
| rlx-pkg cargo | | – 0.2.18 did not build | – | RLX package format (.rlxp) — flat hybrid mmap packs, optional graph, ZIP/dir | 772 | 4h ago |
| rlx-gpu-host cargo | | – 0.2.18 did not build | – | Backend-agnostic host-fallback kernels (D2H → CPU → H2D) shared by RLX GPU backends (rlx-cuda / rlx-rocm). | 771 | 2h ago |
| rlx-onnx-proto cargo | 0.2.18 | none | none | Vendored pure-Rust ONNX protobuf types (protoc-free) | 713 | 39m ago |
| rlx-unfuse cargo | | – 0.2.18 did not build | – | Shared IR-level unfusion/decompose pass for the GPU-family backends (CUDA/ROCm/wgpu), parameterized by a per-backend DecomposePolicy. | 704 | 4h ago |
| rlxsl cargo | | – 0.2.18 did not build | – | Scalar-expression manifest + per-language emitters for RLX's warm-tier unary-activation kernels (single source for WGSL/CUDA/MSL/GLSL/OpenCL activation math,… | 607 | 4h ago |
| rlx-optim cargo | 0.2.18 | none | none | RLX training-step optimizers — Adam, AdamW, NAdamW, RAdam, QHAdamW, LAMB, Adafactor, Lion, SOAP, Kron-PSGD, Muon, Sophia, MARS | 577 | 4h ago |
| rlx-cortexm cargo | 0.2.18 | none | none | Cortex-M backend for RLX — INT8 kernels for ARMv7E-M microcontrollers (e.g. nRF52840) | 491 | 1h ago |
