Model Card for Qwen-AzE

Model Details

Model Description

Qwen-AzE is a 1.7B parameter language model based on the Qwen-3 architecture, fine-tuned on a high-quality 51K Azerbaijani Alpaca-style dataset. This model has been optimized for understanding and generating text in Azerbaijani, supporting a variety of natural language processing tasks such as question answering, instruction following, conversational AI, and creative text generation.

Uses

Direct Use

• Instruction-following applications in Azerbaijani
• Conversational agents and chatbots
• Educational content generation in Azerbaijani
• Text completion, summarization, and translation tasks
• Research and experimentation with Azerbaijani NLP

Bias, Risks, and Limitations

• The model is trained primarily on Azerbaijani text and may perform poorly with other languages.
• Outputs may contain inaccuracies, biases, or culturally sensitive content; users should validate critical outputs before deployment.
• Not intended for high-stakes decision-making without human oversight.

How to Get Started with the Model

Use the code below to get started with the model.

from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("khazarai/Qwen-AzE")
model = AutoModelForCausalLM.from_pretrained(
    "khazarai/Qwen-AzE",
    device_map={"": 0}
)

question = """Bu cümləni ifadə edin: "Doğru olanı etmək çətindir"
"""

messages = [
    {"role" : "user", "content" : question}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize = False,
    add_generation_prompt = True, 
    enable_thinking = False, 
)

from transformers import TextStreamer
_ = model.generate(
    **tokenizer(text, return_tensors = "pt").to("cuda"),
    max_new_tokens = 250,
    temperature = 0.7, 
    top_p = 0.8, 
    top_k = 20,
    streamer = TextStreamer(tokenizer, skip_prompt = True),
)

Training Data

51K Azerbaijani Alpaca-style instruction-response pairs, designed to cover diverse topics and conversational contexts.

• Dataset: saillab/alpaca-azerbaijani-cleaned