base_model, base_model_relation, license, pipeline_tag, tags

base_model	base_model_relation	license	pipeline_tag	tags
Mawdistical/Kuwutu-7B	quantized	cc-by-nc-4.0	text-generation	gguf quantization q8_0 q6_k q5_k_m q4_k_m q3_k_m iq4_xs iq3_m iq2_m qwen2 chatml text-generation creative-writing interactive-fiction cyoa furry adult not-for-all-audiences

Kuwutu-7B-CYOA-GGUF

This repository contains mainline-compatible GGUF quantizations of KeinNiemand/Kuwutu-7B-CYOA, a narrow fine-tune of Mawdistical/Kuwutu-7B for adult furry CYOA / interactive-fiction style continuation.

These files are intended for regular llama.cpp-compatible runtimes.

Important Replacement Notice

Older GGUF files in this repository named like run0067-checkpoint-274-*.gguf were generated with incorrect RoPE metadata and should be considered broken. In particular, those exports could load with qwen2.rope.freq_base / freq_base_train set to 10000 instead of the correct 640000, causing severe generation degradation.

Use the Kuwutu-7B-CYOA-*.gguf files instead. The corrected files should report qwen2.rope.freq_base = 640000 when loaded.

Other Versions

• Canonical LoRA adapter: KeinNiemand/Kuwutu-7B-CYOA-LoRA
• Merged HF model: KeinNiemand/Kuwutu-7B-CYOA
• Adapter GGUF: KeinNiemand/Kuwutu-7B-CYOA-LoRA-GGUF
• ik_llama.cpp merged GGUF: KeinNiemand/Kuwutu-7B-CYOA-IK-GGUF

Files

• Kuwutu-7B-CYOA-bf16.gguf
• Kuwutu-7B-CYOA-Q8_0.gguf
• Kuwutu-7B-CYOA-Q6_K.gguf
• Kuwutu-7B-CYOA-Q5_K_M.gguf
• Kuwutu-7B-CYOA-Q4_K_M.gguf
• Kuwutu-7B-CYOA-Q3_K_M.gguf
• Kuwutu-7B-CYOA-IQ4_XS.gguf
• Kuwutu-7B-CYOA-IQ3_M.gguf
• Kuwutu-7B-CYOA-IQ2_M.gguf

Model Summary

This is a narrow adult furry CYOA / interactive-fiction model. It is not intended as a general writing model or broad instruction model.

The training data is permissioned adult furry CYOA / interactive-fiction writing. The dataset is not redistributed here.

Quantization Notes

These quants were produced from the merged model using an imatrix generated from project calibration text. They have not been formally benchmarked beyond local sanity checks.