mxfp8

High-performance implementations of AI kernels for CPUs (x64, AArch64, RISC-V) and Intel GPUs. Powers PyTorch, TensorFlow, OpenVINO, and ONNX Runtime.

cuDNN Frontend is NVIDIA's modern, open-source entry point to the cuDNN library and a growing collection of high-performance open-source kernels.

LLM fine-tuning with LoRA + NVFP4/MXFP8 on NVIDIA DGX Spark (Blackwell GB10)

Easy recipes to speed up latency of Flux, QwenImage, and LTX-2 with NVFP4 and MXFP8 on Blackwell.

Patches + recipe to deploy festr2/MiMo-V2.5-Pro-NVFP4-MXFP8-attn-TP8 on 8-node DGX Spark sm_121 (Ray + vLLM, TP=8). Fixes the fused-qkv loader bug that mis-slotted Q values as K/V on 7 of 8 ranks.

Three ComfyUI nodes that show the real runtime precision of MODEL, CLIP, and VAE objects.

ComfyUI custom node for converting diffusion checkpoints to FP8, NVFP4, MXFP8 and INT8 with native per-layer quantization metadata.

A dependency-light, transformers-free toolkit for data-free weight-only quantization (FP8 blockwise, MXFP8, MXFP4), with file-level multi-node support and HF / compressed-tensors output.

Blackwell-optimized llama.cpp bundle with DFlash2, MXFP6/MXFP8/NVFP4, TurboQuant KV, and GPU-resident speculative handoff.

Dual DGX Spark TP=2 serving for GLM-5.3 Flash with an NVFP4 target and MXFP8 DFlash2 draft

Block-scaled FP8 / FP4 / INT4 tensor primitive with Triton scaled-matmul at FP32 parity on H100. NumPy / PyTorch / MLX / JAX backends.

MXFP/NVFP/FP16 to GGUF conversion with automatic safe routing