Results (parse = decodes; fidelity = byte-exact round-trip)

• valid = held-out split of in-training shapes; holdout = shapes withheld from training entirely.
• Token cost vs minified JSON: not re-measured for this base's tokenizer (the −14% figure is from the Llama-3.2/cl100k bench).

Training

Data: synthetic RAIF examples (with mechanism-carrier shapes) augmented with real tool-call argument objects from glaiveai/glaive-function-calling-v2 (Apache-2.0), kept only where they round-trip losslessly. Full recipe: RECIPE.md.

Usage

python
Copy code
from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

base = AutoModelForCausalLM.from_pretrained("unsloth/Qwen2.5-0.5B-Instruct")
tok = AutoTokenizer.from_pretrained("skrrt-sh/raif-qwen2.5-0.5b-lora")
model = PeftModel.from_pretrained(base, "skrrt-sh/raif-qwen2.5-0.5b-lora")

License & attribution

Derivative of Qwen2.5 — Apache-2.0 (the Qwen2.5 small bases are Apache-2.0 licensed). Trained in part on glaiveai/glaive-function-calling-v2 (Apache-2.0) — attribute Glaive AI.