Demucs Mdx Q

Demucs-mdx_q provides pretrained weights for the mdx_q model within the Demucs family, designed for music source separation in the waveform domain. The resource consists of a set of four models, with their outputs averaged as described in the mdx_q.yaml file. Each model's weights are distributed in .safetensors files, which include class and initialization arguments stored as JSON metadata, and are accompanied by a sidecar .json file containing full training metadata. The tool is made available under the MIT license. It is hosted on Hugging Face, and users are directed to consult the Demucs repository for instructions on how to utilize these weights. No information is provided regarding specific user roles, supported platforms, or additional features beyond the provision of model weights for music source separation tasks. There is no mention of pricing beyond the open MIT licensing, nor are download metrics or deployment details by inference providers included in the available information.