MiniCPM-V 4.7 1B Mermaid LoRA GGUF
Lucasgitguy-co/minicpm-v-4-7-1b-mermaid-adapter as a llama.cpp LoRA file. Load it on top of a MiniCPM-V 4.7 1B base GGUF with --lora.
| File | Type | Size (GB) |
|---|---|---|
minicpm-v-4-7-1b-mermaid-LoRA-F16.gguf |
LoRA, rank 8, alpha 16, 150 language-model layers | 0.010 |
Use
Tested base: the official openbmb/MiniCPM-V-4.7-1B-gguf. For images, add that repo's mmproj file. The adapter doesn't change the vision tower.
hf download openbmb/MiniCPM-V-4.7-1B-gguf MiniCPM-V-4_7-1B-Q4_K_M.gguf --local-dir .
hf download Lucasgitguy-co/minicpm-v-4-7-1b-mermaid-adapter-GGUF minicpm-v-4-7-1b-mermaid-LoRA-F16.gguf --local-dir .
llama-cli -m MiniCPM-V-4_7-1B-Q4_K_M.gguf --lora minicpm-v-4-7-1b-mermaid-LoRA-F16.gguf --jinja \
-p "Write Mermaid code for a flowchart: Start, Process, End."
How it was made and tested
- Converted with llama.cpp
convert_lora_to_gguf.py(checkout1b6935b). The base config is fromopenbmb/MiniCPM-V-4.7-1Bat revision286817f, the revision the adapter was trained on. - Tested 2026-11-16 with llama.cpp 1.0.0 (build 12359):
- Base F16 + this LoRA gives exactly the same greedy text as a hand-merged model (W + 2·B·A, converted separately) on 3 Mermaid prompts of 80 tokens each. Both differ from the plain base, so the LoRA is applied.
- The LoRA also loads without errors on openbmb's
MiniCPM-V-4_7-1B-Q4_K_M.ggufand gives Mermaid-only answers there. - Image input with the LoRA was not tested.
Licence
Apache-2.0, the same as the adapter and the base model. See NOTICE.