Collected molecules will appear here. Add from search or explore.
A specialized quantization script to convert GGUF-formatted models (specifically Qwen3.5-27B Gated-DeltaNet hybrids) into NVIDIA's 4-bit floating point (NVFP4) format.
stars
4
forks
1
This is a highly niche utility tool with very low adoption (4 stars). It bridges a gap between the community GGUF format and NVIDIA's proprietary FP4 format for a specific experimental model architecture. This functionality is likely to be subsumed by official NVIDIA tools (TensorRT-LLM) or mainstream community projects like llama.cpp as FP4 hardware (Blackwell) becomes more common.
TECH STACK
INTEGRATION
cli_tool
READINESS