cstr/adair-GGUF

AdaIR All-in-One Restoration (GGUF)

Restormer + Adaptive Frequency Learning Blocks for 5-task restoration: denoise, derain, dehaze, deblur, low-light. 28.8M params, 110 MB F32. FFT-based spectral decomposition with adaptive masking and cross-attention.

Parity: cos=0.999924 vs Python reference. Source: c-yn/AdaIR (MIT, ICLR 2025).

Provenance and EU AI Act Art. 53 note

• Upstream model: c-yn/AdaIR.
• Upstream licence: mit. This repository redistributes under the same terms; it grants no rights the upstream licence does not.
• What was done here: format conversion and/or quantisation only (GGUF). No training, no fine-tuning, no merging, no distillation, no change to architecture, vocabulary or capability. Only the numeric representation of the upstream weights differs.
• Training data: documented — where it is documented at all — by the upstream provider; see the upstream model card. No training data was used, added or selected by this repository.
• Provider status: under Regulation (EU) 2024/1689 the upstream authors remain the provider of this model. Converting the serialisation format does not make this repository the provider of a new general-purpose AI model, and no such claim is made. Questions about training content, copyright policy or model capability belong upstream.