raif-qwen3-4b-lora

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

• valid = held-out split of in-training shapes; holdout = shapes withheld from training entirely.
• Token cost: ~10% fewer than minified JSON on real function-call data (cross-tokenizer).

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/Qwen3-4B-Instruct-2507")
tok = AutoTokenizer.from_pretrained("skrrt-sh/raif-qwen3-4b-lora")
model = PeftModel.from_pretrained(base, "skrrt-sh/raif-qwen3-4b-lora")

This model emits RAIF, not JSON — decode it at the output boundary with the official codec (pure-stdlib, no bun, nothing to clone):

sh
Copy code
pip install raif-format        # or: uv add raif-format

python
Copy code
from raif import decode        # installs as `raif-format`, imports as `raif`

result = decode(model_output)  # {"ok", "value", "repairs"}
data = result["value"] if result["ok"] else None   # ordinary JSON, ready downstream

decode_lenient() recovers the intact leaves of a truncated stream. The codec is the same one used to score this model, kept byte-identical across Python and TypeScript by a shared conformance corpus.

Links

• Format spec & reference codec: https://github.com/skrrt-sh/raif-standard
• Decoder: raif-format on PyPI · raif-format on npm
• Training recipe, eval & the other RAIF models: https://github.com/skrrt-sh/raif-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.