fp4

4 projects share this GitHub topic

Lines connect members that are measurably related to each other. Dot size reflects stars.

🧬 Members

TransformerEngine

A library for accelerating Transformer models on NVIDIA GPUs, including using 8-bit and 4-bit floating point…

★ 3.4k

neural-compressor

SOTA low-bit LLM quantization (INT8/FP8/MXFP8/INT4/MXFP4/NVFP4) & sparsity; leading model compression…

★ 2.7k

AngelSlim

Model compression toolkit engineered for enhanced usability, comprehensiveness, and efficiency.

★ 1.6k

halofpx

Run Ornith, Qwen, Nemotron & DeepSeek optimized on AMD Strix Halo — unified OpenAI-compatible server with…

★ 41 · GitHub ↗

Measured from GitHub topics shared by both projects, weighted by how rare each topic is.