from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("edith81/Qwen2.5-1.5B-Vietnamese-MCQ-AllSubjects-merged")
tokenizer = AutoTokenizer.from_pretrained("edith81/Qwen2.5-1.5B-Vietnamese-MCQ-AllSubjects-merged")

# Use the model
messages = [
    {"role": "system", "content": "Bạn là trợ lý trả lời trắc nghiệm.\n\nTrả lời ĐÚNG định dạng JSON:\n{\"answer\":\"A\"}\n\nTrong đó A/B/C/D là đáp án đúng.\nKHÔNG giải thích, KHÔNG suy luận, CHỈ trả lời JSON."},
    {"role": "user", "content": "Your question here\nA. Option A\nB. Option B\nC. Option C\nD. Option D"}
]

inputs = tokenizer.apply_chat_template(messages, return_tensors="pt", add_generation_prompt=True)
outputs = model.generate(inputs, max_new_tokens=64, temperature=0.1, do_sample=False)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)

print(response)  # {"answer":"A"}

Model Details

Base Model: Qwen/Qwen2.5-1.5B-Instruct
Type: Merged model (LoRA + base)
Format: Standard Transformers model
Output: JSON format {"answer":"A"} (or B/C/D)

Competition Submission

Use this repo name when submitting to the competition judging system:

edith81/Qwen2.5-1.5B-Vietnamese-MCQ-AllSubjects-merged

Merged Date

Merged on: 2025-12-19 10:04:26