bash2nl-qwen2.5-coder-7b-GGUF

bash2nl — Qwen2.5-Coder-7B QLoRA, GGUF q4_K_M

Explains a Bash command line in exactly one English sentence, phrased as an instruction starting with a verb. QLoRA fine-tune of Qwen/Qwen2.5-Coder-7B-Instruct, merged and quantized to q4_K_M.

find . -name "*.py"
-> Display the names of all files in the current directory and recursively into subdirectories with names that end in .py

Usage

hf download C-H-Liu/bash2nl-qwen2.5-coder-7b-GGUF --local-dir bash2nl
cd bash2nl && ollama create bash2nl -f Modelfile
ollama run bash2nl 'ps -ef | grep nginx | awk "{print \$2}" | xargs kill -9'

The Modelfile pins the system prompt and this decoding:

Evaluation

920 held-out commands Judge columns are claude-sonnet-5 scoring a fixed 200-command subset.

License

Apache-2.0, matching the base model, whose weights this GGUF contains. Training data comes from the nl2bash project; consult it for the terms attached to that corpus.