You can deploy this model and start earning money today!

whotookmycookie/
LLSec

Views

No views yet

LLSec

LLSec is a prototype tool for generating business rule specifications in the securities domain based on large language models. It is the official implementation for paper . In this paper, we propose an automatic specification method for business rules in the securities domain based on large language models, which utilizes the powerful natural language processing ability of the large language models to assist in classification and extraction of software requirement related business rules. In addition, we also use domain knowledge to assist in refining business rules and identifying rule relationships, ultimately forming a requirement specification in the form of a data stream.

This repository presents the trained models used in our tool. The code of model deployment, fine-tuning, and reasoning is available at https://github.com/LingelLi/FinSage

Model List

mengzi_rule_filtering. This is the fine-tuned mengzi model for rule filtering task. Running example:

python

        from transformers import AutoModelForSequenceClassification, AutoTokenizer
        import torch
        model_path = "./mengzi_rule_filtering"

        model = AutoModelForSequenceClassification.from_pretrained(model_path, num_labels=3)
        tokenizer = AutoTokenizer.from_pretrained(model_path)
        model.eval()
        inputs = "采用匹配成交方式的，每个交易日的9:15至9:25为开盘集合匹配时间，9:30至11:30、13:00至15:30为连续匹配时间。"  # input sentence
        input_ids = tokenizer([inputs], max_length=512, padding="max_length", truncation=True, return_tensors="pt").input_ids
        logits = model(input_ids=input_ids).logits
        _, outputs = torch.max(logits, dim=1)
        outputs = outputs.numpy()[0]
        print(outputs)  # classification output

mengzi_rule_extraction. This is the fine-tuned mengzi model for rule extraction task. Running example:

python

        from transformers import AutoModelForTokenClassification, AutoTokenizer
        import torch
        model_path = "./mengzi_rule_extraction"

        model = AutoModelForTokenClassification.from_pretrained(model_path, num_labels=37)
        tokenizer = AutoTokenizer.from_pretrained(model_path)
        model.eval()
        inputs = "采用匹配成交方式的，每个交易日的9:15至9:25为开盘集合匹配时间，9:30至11:30、13:00至15:30为连续匹配时间。"  # input sentence
        input_ids = tokenizer([inputs], max_length=512, padding="max_length", truncation=True, return_tensors="pt").input_ids
        logits = model(input_ids=input_ids).logits
        _, outputs = torch.max(logits, dim=2)
        outputs = outputs.numpy()[0][:len(inputs)+2][1:-1]
        dic = open("tc_data.dict", "r", encoding="utf-8").read()
        dic = {int(num):val for t in dic.split("\n") for num, val in [t.split("\t")]}
        print([dic[o] for o in outputs])  # classification output

llama2_rule_filtering_fine_tune. This is the fine-tuned llama2 model for rule filtering task. Running example:

python

        from transformers import AutoModelForCausalLM, AutoTokenizer
        import torch
        model_path = "./llama2_rule_filtering_fine_tune"

        model = AutoModelForCausalLM.from_pretrained(model_path, return_dict=True, torch_dtype=torch.float16, device_map="cuda:0" if torch.cuda.is_available() else 'auto', trust_remote_code=True, use_flash_attention_2=True)
        tokenizer = AutoTokenizer.from_pretrained(model_path)
        model.eval()
        inputs = "<s>Human: 给出一条规则，请你将它分类为软件需求相关规则、软件需求无关规则和领域知识之一。\n规则: 采用匹配成交方式的，每个交易日的9:15至9:25为开盘集合匹配时间，9:30至11:30、13:00至15:30为连续匹配时间。\n</s><s>Assistant: "  # 输入句子
        input_ids = tokenizer([inputs], return_tensors="pt", add_special_tokens=False).input_ids
        if torch.cuda.is_available():
        input_ids = input_ids.to('cuda:0')
        generate_ids = model.generate(input_ids=input_ids)
        outputs = tokenizer.decode(generate_ids[0])
        print(outputs)  # 输出类别

llama2_rule_extraction_fine_tune. This is the fine-tuned llama2 model for rule extraction task. Running example:

python

        from transformers import AutoModelForCausalLM, AutoTokenizer
        import torch
        model_path = "./llama2_rule_extraction_fine_tune"

        model = AutoModelForCausalLM.from_pretrained(model_path, return_dict=True, torch_dtype=torch.float16, device_map="cuda:0" if torch.cuda.is_available() else 'auto', trust_remote_code=True, use_flash_attention_2=True)
        tokenizer = AutoTokenizer.from_pretrained(model_path)
        model.eval()
        inputs = "<s>Human: 给出一条规则，请你将它分类为软件需求相关规则、软件需求无关规则和领域知识之一。\n规则: 采用匹配成交方式的，每个交易日的9:15至9:25为开盘集合匹配时间，9:30至11:30、13:00至15:30为连续匹配时间。\n</s><s>Assistant: "  # 输入句子
        input_ids = tokenizer([inputs], return_tensors="pt", add_special_tokens=False).input_ids
        if torch.cuda.is_available():
        input_ids = input_ids.to('cuda:0')
        generate_ids = model.generate(input_ids=input_ids)
        outputs = tokenizer.decode(generate_ids[0])
        print(outputs)  # 输出抽取的规则

llama2_rule_filtering_lora. This is the lora trained llama2 model for rule filtering task. Running example:

python

        # download the pre-trained model first from https://huggingface.co/FlagAlpha/Atom-7B
        from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
        import torch
        from peft import PeftConfig, PeftModel
        model_path = "./model/rule_filtering/llama2_rule_filtering_lora"

        bnb_config = BitsAndBytesConfig(load_in_4bit=True, bnb_4bit_use_double_quant=True, bnb_4bit_quant_type="nf4", bnb_4bit_compute_dtype=torch.bfloat16)
        peft_config = PeftConfig.from_pretrained(model_path)
        model = AutoModelForCausalLM.from_pretrained(peft_config.base_model_name_or_path, torch_dtype=torch.float16, device_map="cuda:0" if torch.cuda.is_available() else 'auto', trust_remote_code=True, use_flash_attention_2=True, quantization_config=bnb_config)
        model = PeftModel.from_pretrained(model, model_path, device_map="cuda:0" if torch.cuda.is_available() else "auto")
        tokenizer = AutoTokenizer.from_pretrained(peft_config.base_model_name_or_path)
        model.eval()
        inputs = "<s>Human: 给出一条规则，请你将它分类为软件需求相关规则、软件需求无关规则和领域知识之一。\n规则: 采用匹配成交方式的，每个交易日的9:15至9:25为开盘集合匹配时间，9:30至11:30、13:00至15:30为连续匹配时间。\n</s><s>Assistant: "  # 输入句子
        input_ids = tokenizer([inputs], return_tensors="pt", add_special_tokens=False).input_ids
        if torch.cuda.is_available():
        input_ids = input_ids.to('cuda:0')
        generate_ids = model.generate(input_ids=input_ids)
        outputs = tokenizer.decode(generate_ids[0])
        print(outputs)  # 输出类别

llama2_rule_extraction_lora. This is the lora trained llama2 model for rule extraction task. Running example:

python

        # download the pre-trained model first from https://huggingface.co/FlagAlpha/Atom-7B
        from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
        import torch
        from peft import PeftConfig, PeftModel
        model_path = "./llama2_rule_extraction_lora"

        bnb_config = BitsAndBytesConfig(load_in_4bit=True, bnb_4bit_use_double_quant=True, bnb_4bit_quant_type="nf4", bnb_4bit_compute_dtype=torch.bfloat16)
        peft_config = PeftConfig.from_pretrained(model_path)
        model = AutoModelForCausalLM.from_pretrained(peft_config.base_model_name_or_path, torch_dtype=torch.float16, device_map="cuda:0" if torch.cuda.is_available() else 'auto', trust_remote_code=True, use_flash_attention_2=True, quantization_config=bnb_config)
        model = PeftModel.from_pretrained(model, model_path, device_map="cuda:0" if torch.cuda.is_available() else "auto")
        tokenizer = AutoTokenizer.from_pretrained(peft_config.base_model_name_or_path)
        model.eval()
        inputs = "<s>Human: 给出一条规则，请你将它分类为软件需求相关规则、软件需求无关规则和领域知识之一。\n规则: 采用匹配成交方式的，每个交易日的9:15至9:25为开盘集合匹配时间，9:30至11:30、13:00至15:30为连续匹配时间。\n</s><s>Assistant: "  # 输入句子
        input_ids = tokenizer([inputs], return_tensors="pt", add_special_tokens=False).input_ids
        if torch.cuda.is_available():
        input_ids = input_ids.to('cuda:0')
        generate_ids = model.generate(input_ids=input_ids)
        outputs = tokenizer.decode(generate_ids[0])
        print(outputs)  # 输出抽取的规则

whotookmycookie/LLSec

LLSec

Model List

whotookmycookie/
LLSec