--- license: apache-2.0 base_model: ggml-org-dev/tiny-lev-for-testing tags: - gguf - testing --- # tiny-lev-for-testing-GGUF > [!WARNING] > This is a **modified, damaged copy** of [interfaze-ai/lev](https://huggingface.co/interfaze-ai/lev), made only to test software. Its output is gibberish. Do not use it for anything else. GGUF of [ggml-org-dev/tiny-lev-for-testing](https://huggingface.co/ggml-org-dev/tiny-lev-for-testing) merged into its base model [ggml-org-dev/tiny-Qwen3.5-4B-for-testing](https://huggingface.co/ggml-org-dev/tiny-Qwen3.5-4B-for-testing), converted with `convert_hf_to_gguf.py` of llama.cpp (F16, then quantized to Q4_0 with `llama-quantize`, token embedding included). Used by the llama.cpp server tests of the `lev` decision type. The vocabulary of the GGUF is cut from 248320 to 3348 tokens (the HF slice keeps the full tokenizer): the special tokens, all single characters, the BPE path of the decision label words (A..Z, AA..ZZ, digits, yes/no) and the first 2000 merges are kept, the token ids are renumbered and the rows of the token embedding are sliced to match. Text is still fully encodable, but into more tokens than with the original tokenizer. ```sh llama-server -hf ggml-org-dev/tiny-lev-for-testing-GGUF:Q4_0 ``` ## License Derived from lev by Interfaze AI, under the Apache License 2.0 (as is the base model Qwen3.5-4B).