Gentoo Packages

Search

Install

Install this package:

emerge -a sci-ml/flm-q4nx-converter

Package Information

Description:

Converts the weights of a GGUF model file to Q4NX, the safetensors layout FastFlowLM loads on AMD Ryzen AI (XDNA2) NPUs. The converter repacks weights only. The NPU kernels are compiled per model and ship with FastFlowLM, so the source must be a fine-tune of a model that FastFlowLM already supports, with the same architecture and size.

Homepage:

https://github.com/ROCm/FLM_Q4NX_Converter

License:

Apache-2.0

Versions

Metadata

Description

• Converts the weights of a GGUF model file to Q4NX, the safetensors layout FastFlowLM loads on AMD Ryzen AI (XDNA2) NPUs. The converter repacks weights only. The NPU kernels are compiled per model and ship with FastFlowLM, so the source must be a fine-tune of a model that FastFlowLM already supports, with the same architecture and size.

Maintainers

• Ivan S. Titov (iohann.s.titov@gmail.com) - Type: person

Upstream

• Bugs: https://github.com/ROCm/FLM_Q4NX_Converter/issues
• Remote IDs:
  • github: ROCm/FLM_Q4NX_Converter

<pkgmetadata>
	<maintainer type="person">
		<email>iohann.s.titov@gmail.com</email>
		<name>Ivan S. Titov</name>
	</maintainer>
	<longdescription lang="en">
    Converts the weights of a GGUF model file to Q4NX, the safetensors
    layout FastFlowLM loads on AMD Ryzen AI (XDNA2) NPUs. The converter
    repacks weights only. The NPU kernels are compiled per model and ship
    with FastFlowLM, so the source must be a fine-tune of a model that
    FastFlowLM already supports, with the same architecture and size.
  </longdescription>
	<upstream>
		<bugs-to>https://github.com/ROCm/FLM_Q4NX_Converter/issues</bugs-to>
		<remote-id type="github">ROCm/FLM_Q4NX_Converter</remote-id>
	</upstream>
</pkgmetadata>

Lint Warnings

• [Warning] Ebuild 0_pre20260827 has a line exceeding 80 positions on line 13.
• [Warning] Ebuild 0_pre20260827 has a line exceeding 80 positions on line 93.
• [Warning] Ebuild 0_pre20260827 has a line exceeding 80 positions on line 94.
• [Notice] Package has only unstable keywords for arch(es): [ amd64 ], all versions are unstable: [ 0_pre20260827 ]

Files

• flm-q4nx-check
• flm-q4nx-converter-0_pre20260827-configs-path.patch
• flm-q4nx-converter-0_pre20260827-detect-size-label.patch
• flm-q4nx-converter-0_pre20260827-qwen35-prefill-copies.patch
• flm-q4nx-converter-0_pre20260827-qwen35-value-heads.patch.xz
• flm-q4nx-converter-0_pre20260827-qwen38-27b.patch.xz

Manifest