Views
No views yet
ZeroXClem/Qwen3.5-9B-Fable-5-Quad-Stock using llama.cpp (master) on Modal.
Refer to the original model card for more details on the model.Rebuilt 2026-07-03: the original conversion inherited a stalemtp_num_hidden_layers = 1from the merge config (mergekit drops themtp.*tensors of models that not all donors share, but copies the base config verbatim), so the GGUF declared a 33rd block with no tensors behind it and failed to load in every llama.cpp-based runtime (LM Studio included:missing tensor 'blk.32.attn_norm.weight'). Rebuilt from the corrected config — inference weights are unchanged.
brew install llama.cppllama-cli --hf-repo ZeroXClem/Qwen3.5-9B-Fable-5-Quad-Stock-Q5_0-GGUF --hf-file qwen3.5-9b-fable-5-quad-stock-q5_0.gguf -p "The meaning to life and the universe is"llama-server --hf-repo ZeroXClem/Qwen3.5-9B-Fable-5-Quad-Stock-Q5_0-GGUF --hf-file qwen3.5-9b-fable-5-quad-stock-q5_0.gguf -c 2048mmproj-qwen3.5-9b-fable-5-quad-stock-f16.gguf in this repo is the model's vision encoder + projector,
extracted from the same merge. LM Studio and llama.cpp pair it with
the text GGUF automatically when both sit in the model folder — download
both files and the model accepts images.1llama-mtmd-cli --hf-repo ZeroXClem/Qwen3.5-9B-Fable-5-Quad-Stock-Q5_0-GGUF --hf-file qwen3.5-9b-fable-5-quad-stock-q5_0.gguf \
2 --mmproj mmproj -p "Describe this image" --image photo.jpg