main.py

# -*-coding:utf-8 -*-
"""
    Run SELF
"""

import numpy as np
from self.generate import SELF

threshold = 0.7 # 当最近几个指令相似度都很高的时候停止


def main(seed_file, output_file, openai_key, n_human, n_machine, n_instruct, max_iter, max_gen):
    instance = SELF(seed_file, openai_key, n_human, n_machine, n_instruct, None)

    n_iter = 0
    while n_iter < max_iter and instance.n_keep < max_gen:
        instance.step()
        n_iter +=1
        print(f'已生成{instance.n_gen} 可用{instance.n_keep}')
        if n_iter >3 and np.average([i['avg_similarity_score'] for i in instance.machine_instruction_data[-5:]] )> threshold:
            break

    # dump file
    instance.dump_file(output_file)



if __name__ == '__main__':
    seed_file = './ape/data/seed_task.json'
    openai_key ='a'
    n_human=2
    n_machine=1
    n_instruct=4
    instance = SELF(seed_file, openai_key, n_human, n_machine, n_instruct, None)

    scorer = rouge_scorer.RougeScorer(['rougeL'], use_stemmer=False, tokenizer=ChineseTokenizer())
    inst_tokens = scorer._tokenizer.tokenize('诊断患者')
    with Pool(os.cpu_count()) as p:
        rouge_scores = p.map(partial(rouge_scorer._score_lcs,inst_tokens), instance.all_instruction_tokens)
    rouge_l = [score.fmeasure for score in rouge_scores]