3.6. Merge Mode

./juno merge bakes a trained LoRA adapter into a standalone GGUF. It writes a new GGUF where the LoRA-patched projection tensors (wq/wv on every layer) are stored as F32 for full precision. All other tensors are copied verbatim in their original quantized encoding. The resulting file loads with ./juno local or ./juno like any other model, with no sidecar adapter needed.

# Default: reads <model>.lora, writes <model>-merged.gguf
./juno merge --model-path /path/to/TinyLlama.Q4_K_M.gguf

# Explicit paths
./juno merge --model-path /path/to/model.gguf \
             --lora-path  /adapters/my.lora   \
             --output     /path/to/merged.gguf

# Larger heap for big models (rule of thumb: 2x model file size)
./juno merge --model-path /path/to/Mistral-7B.gguf --heap 12g

Windows (Command Prompt):

juno.bat merge --model-path models\TinyLlama.Q4_K_M.gguf

juno.bat merge --model-path models\model.gguf ^
               --lora-path adapters\my.lora ^
               --output merged\merged.gguf

juno.bat merge --model-path models\Mistral-7B.gguf --heap 12g

The LoRA delta per element (roughly 6e-4) is smaller than Q4_K quantization noise (roughly 3e-3). Re-quantizing the merged weights back to Q4_K would erase the training entirely. F32 storage for the 44 patched tensors is the correct trade-off. For TinyLlama 1.1B Q4_K_M (667 MB), the merged file is approximately 1 GB.

Before redistributing a merged GGUF, review the licensing implications: see LoRA and merge licensing.

See also¶