You can deploy this model and start earning money today!

Hodfa71/
gpt-sw3-356m-da-saga-delta-dpo

Name: gpt-sw3-356m-da-saga-delta-dpo
Availability: InStock

AI-Sweden-Models/gpt-sw3-356m

adapter

other

Views

No views yet

GPT-SW3 356M — Danish Grammar-Aligned (SAGA Δ-DPO)

Fine-tuned with SAGA (Syntax-Aware Grammar Alignment), a two-stage pipeline that trains language models to generate grammatically correct Danish text using reinforcement learning from a symbolic parser oracle (SpaCy (Danish dependency parser)).

This is a LoRA adapter. Load it on top of AI-Sweden-Models/gpt-sw3-356m.

SFT skipped (base PS 85.0% ≥ 80% threshold). Δ-DPO applied directly to base.

Results (independent Stanza evaluation)

Metric	Base	+ Δ-DPO
Parse success	85.0%	97.5%
Parse score	0.429	0.689
PPL-Wiki	19.5	21.7

Usage

python

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base = AutoModelForCausalLM.from_pretrained("AI-Sweden-Models/gpt-sw3-356m", torch_dtype="auto")
model = PeftModel.from_pretrained(base, "Hodfa71/gpt-sw3-356m-da-saga-delta-dpo")

tokenizer = AutoTokenizer.from_pretrained("AI-Sweden-Models/gpt-sw3-356m")

prompt = "Dansk er"
inputs = tokenizer(prompt, return_tensors="pt")
output = model.generate(**inputs, max_new_tokens=60, temperature=0.8, do_sample=True)
print(tokenizer.decode(output[0], skip_special_tokens=True))

Training details

Data: 10 000 Danish Wikipedia sentences (filtered for quality)
Method: Δ-DPO — generate N=8 candidates per prompt, keep pairs with parser score gap Δ ≥ 0.25, train with standard DPO loss (β=0.1)
Parser oracle: SpaCy (Danish dependency parser)
LoRA: rank 16, α=32, all linear layers, bfloat16
Auto-SFT rule: SFT is applied first only if base parse success < 80%

Citation

bibtex

@article{fakhar2025saga,
  title={SAGA: Syntax-Aware Grammar Alignment for Low-Resource Nordic Languages},
  author={Fakhar, Hoda and others},
  year={2025},
  note={Under review}
}

License

Inherits the base model license (AI Sweden LLM License / LumiOpen).