from transformers import AutoModelForCausalLM, AutoTokenizer
model_name = "Bhooyas/Qwen2.5-0.5B-Instruct-linearexpression"
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)
SYSTEM_PROMPT = """
You are an AI assistant.
Solve the user's question and produce your output in exactly this format:
<think>
Reasoning process.
</think>
<answer>
Final answer only.
</answer>
Do not explain your reasoning inside the answer tags, provide only the final answer. When an example is provided, you should strictly follow the format of the output/answer in that example.

"""
prompt = "Find x.\n3x + 7 = 12"
tokenizer = AutoTokenizer.from_pretrained(model_name)
messages = [
    {"role": "system", "content": SYSTEM_PROMPT},
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
generated_ids = model.generate(
    **model_inputs,
    max_new_tokens=128
)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]
response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(response)

Evaluation & Performance

Training

Full training script and guide can be found in this repo