parser.py

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "Qwen/Qwen3-0.6B"

# load the tokenizer and the model
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)
   
# parsing = [
#     {"name":"education", "type":"List[str]","description":"attended school, university, and other education programs"},
#     {"name":"experience", "type":"float", "description":"years of experience"},
#     {"name":"skills", "type":"List[str]", "description":"list of skills"},
#     {"name":"name", "type":"str", "description":"name of the person"},
#     {"name":"location", "type":"str", "description":"location of the person"},
#     {"name":"email", "type":"str", "description":"email of the person"},
#     {"name":"websites", "type":"List[str]", "description":"urls related of the person"},
#     {"name":"certifications", "type":"List[str]", "description":"list of certifications"},
#     {"name":"languages", "type":"List[str]", "description":"list of languages"},
#     {"name":"projects", "type":"List[str]", "description":"list of projects"},
#     {"name":"note", "type":"str", "description":"additional note which highlight the best or uniqueness of the person"}
# ]


def parse_resume(resume, parsing):
    format_parsing = [f"{x['name']} : {x['type']} = {x['description']}\n" for x in parsing]
    
    prompt = f"""Based on the below resume, tell me the summary details of skills, name, experience years, education, etc in short
    The Output must be the JSON object with the following format:
    {format_parsing}
    
    RESUME:\n""" + resume
    
    messages = [
        {"role": "user", "content": prompt}
    ]
    
    text = tokenizer.apply_chat_template(
        messages,
        tokenize=False,
        add_generation_prompt=True,
        enable_thinking=True # Switches between thinking and non-thinking modes. Default is True.
    )
    model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
    
    # conduct text completion
    generated_ids = model.generate(
        **model_inputs,
        max_new_tokens=32768
    )
    output_ids = generated_ids[0][len(model_inputs.input_ids[0]):].tolist() 
    
    # parsing thinking content
    try:
        # rindex finding 151668 (</think>)
        index = len(output_ids) - output_ids[::-1].index(151668)
    except ValueError:
        index = 0
    
    thinking_content = tokenizer.decode(output_ids[:index], skip_special_tokens=True).strip("\n")
    content = tokenizer.decode(output_ids[index:], skip_special_tokens=True).strip("\n")
    return thinking_content, content