KittyLM-1B (gemma3-1b)

A kitten. In a language model. This is a LoRA finetune of google/gemma-3-1b-it that answers everything in kitten language (mrrp, nya~, prrr, *actions*, occasional :3) while staying factually correct underneath.

License note: this is a derivative of a Gemma model and is distributed under the Gemma Terms of Use. You must accept Google's license on HuggingFace before downloading.

Training

• Data: 900 ShareGPT-style pairs + 100 held-out eval (see KittyLM/kittylm-data), system prompt baked in
• Method: LoRA SFT (scripts/train.py in the KittyLM project), RTX 3060 12GB
• Files: merged bf16 weights + the LoRA adapter (adapter_*.safetensors) live side by side. AutoModelForCausalLM loads the merged model; PeftModel picks up the adapter.
• GGUF quants for Ollama / llama.cpp / LM Studio: KittyLM/kittylm-gemma3-1b-gguf

Limitations

• Persona is stylistic, not a refusal behavior: an explicit "answer in plain English" can make it drop character — it was never trained to resist.
• Small-model knowledge gaps persist: off-distribution facts may confabulate (tracked per model by the 20-probe suite in eval/).
• Kitten flavor adds no capability: reasoning/coding limits are the base model's limits.

Usage (transformers)

from transformers import AutoModelForCausalLM, AutoTokenizer
tok = AutoTokenizer.from_pretrained("KittyLM/kittylm-gemma3-1b")
model = AutoModelForCausalLM.from_pretrained("KittyLM/kittylm-gemma3-1b", device_map="auto", dtype="bfloat16")
msgs = [{"role": "user", "content": "What's the capital of Japan?"}]
x = tok.apply_chat_template(msgs, tokenize=False, add_generation_prompt=True)
print(tok.decode(model.generate(**tok(x, return_tensors="pt").to(model.device), max_new_tokens=120)[0]))
# mrrp... Tokyo. big city. lots of cats. nya~

Usage (Ollama)

hf download KittyLM/kittylm-gemma3-1b-gguf --include "*.gguf" --local-dir ./gguf
ollama create kittylm-gemma3-1b -f Modelfile   # see Modelfile template in project
ollama run kittylm-gemma3-1b "Good night!"