Qwen3.5-9B-Danish-Instruct

Danish instruction-tuned version of unsloth/Qwen3.5-9B, fine-tuned on the kobprof/skolegpt-instruct dataset.

Usage

from transformers import AutoModelForCausalLM, AutoProcessor

model = AutoModelForCausalLM.from_pretrained("AndersMK/Qwen3.5-9B-Danish-Instruct")
processor = AutoProcessor.from_pretrained("AndersMK/Qwen3.5-9B-Danish-Instruct")

messages = [
    {"role": "system", "content": [{"type": "text", "text": "Du er en hjælpsom assistent."}]},
    {"role": "user", "content": [{"type": "text", "text": "Hvad er koldskål?"}]}
]

inputs = processor.apply_chat_template(
    messages, tokenize=True, add_generation_prompt=True,
    enable_thinking=False, return_tensors="pt", return_dict=True
)
outputs = model.generate(**inputs, max_new_tokens=256)
input_len = inputs["input_ids"].shape[1]
print(processor.decode(outputs[0][input_len:], skip_special_tokens=True))

Training

• Base model: unsloth/Qwen3.5-9B
• Dataset: kobprof/skolegpt-instruct
• Method: LoRA fine-tuning with Unsloth

Acknowledgments

• Base model authors
• Dataset authors