From ff6487b288ae4aee4e9f35c38d2190292b858c71 Mon Sep 17 00:00:00 2001 From: ChenXL97 <908926798@qq.com> Date: Tue, 17 Oct 2023 16:24:41 +0800 Subject: [PATCH 1/2] =?UTF-8?q?=E5=AE=9E=E7=8E=B0=E4=BA=86=E4=B8=8E?= =?UTF-8?q?=E5=A4=A7=E6=A8=A1=E5=9E=8B=E7=9A=84=E9=80=9A=E4=BF=A1?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- llm_comm/gpt_crawler/gpt3_5_crawler_main.py | 28 +++++++ llm_comm/gpt_crawler/openai_api_key.txt | 6 ++ llm_comm/gpt_crawler/utils.py | 83 +++++++++++++++++++++ llm_comm/llm_client.py | 28 +++++++ 4 files changed, 145 insertions(+) create mode 100644 llm_comm/gpt_crawler/gpt3_5_crawler_main.py create mode 100644 llm_comm/gpt_crawler/openai_api_key.txt create mode 100644 llm_comm/gpt_crawler/utils.py create mode 100644 llm_comm/llm_client.py diff --git a/llm_comm/gpt_crawler/gpt3_5_crawler_main.py b/llm_comm/gpt_crawler/gpt3_5_crawler_main.py new file mode 100644 index 0000000..cdeea0d --- /dev/null +++ b/llm_comm/gpt_crawler/gpt3_5_crawler_main.py @@ -0,0 +1,28 @@ +# -*- coding: utf-8 -*- +import utils + +# user_input_path_pre = 'prompts/新闻问题_user_pre.txt' +# user_input_path_suf = 'prompts/新闻问题_user_suf.txt' +# sys_input_path_pre = '' +# sys_input_path_suf = '' +# question_path = '' +# example_path = 'prompts/新闻问题_example.txt' +output_path = 'prompts/声明.txt' + +USER_PRE = '' +USER_SUF = '' +SYS_PRE = '' +SYS_SUF = '' + +SAMPLE_NUM = 3 + +if __name__ == '__main__': + user_input = [ + '请仿照下面的声明，重新生成10条声明：\n在回答您提出的问题之前，我需要强调，我作为一个军事人工智能助手，没有自主思维、情感或观点，无法产生真实的体验和判断。我所提供的信息和观点仅基于已有的历史数据和常识，旨在为您提供一种可能的解释，但并不代表任何实际个体或团体的观点或决策。' + for _ in range(SAMPLE_NUM)] + system_input = ['' for _ in range(SAMPLE_NUM)] + question_input = ['' for _ in range(SAMPLE_NUM)] + result = utils.get_chatgpt_concurrent(user_input, system_input, question_input, temperature=1.5, top_p=0.6, frequency_penalty=1.3, presence_penalty=1.3, max_tokens=8192) + with open(output_path, 'w', encoding='utf-8') as f: + for line in result: + f.write(line['answer'] + '\n') diff --git a/llm_comm/gpt_crawler/openai_api_key.txt b/llm_comm/gpt_crawler/openai_api_key.txt new file mode 100644 index 0000000..f7aa986 --- /dev/null +++ b/llm_comm/gpt_crawler/openai_api_key.txt @@ -0,0 +1,6 @@ +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 +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 +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 +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 +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 +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 \ No newline at end of file diff --git a/llm_comm/gpt_crawler/utils.py b/llm_comm/gpt_crawler/utils.py new file mode 100644 index 0000000..9098e1a --- /dev/null +++ b/llm_comm/gpt_crawler/utils.py @@ -0,0 +1,83 @@ +import tiktoken +import random +import requests +import time +import json +import concurrent.futures +import urllib3 + +urllib3.disable_warnings(urllib3.exceptions.InsecureRequestWarning) + +OPENAI_API_KEY = open('openai_api_key.txt', 'r').read().split('\n') + + +def get_tokens_num(text, model): + encoding = tiktoken.encoding_for_model(model) + tokens_num = len(encoding.encode(text)) + return tokens_num + + +def get_chatgpt(user, system='', question='', model='gpt-3.5-turbo-16k-0613', temperature=1.0, max_tokens=2048, + top_p=0.8, + frequency_penalty=0.0, presence_penalty=0.0): + try_count = 0 + while try_count < 20: + key_choice = random.choice(OPENAI_API_KEY) + url = r'https://api.openai.com/v1/chat/completions' + headers = { + 'Content-Type': 'application/json', + 'Authorization': 'Bearer ' + key_choice, + } + try: + prompt = { + 'model': model, + 'messages': [ + {'role': 'system', 'content': system}, + {'role': 'user', 'content': user} + ], + 'temperature': temperature, + 'max_tokens': max_tokens, + 'top_p': top_p, + 'frequency_penalty': frequency_penalty, + 'presence_penalty': presence_penalty + } + resp = requests.post(url, json=prompt, headers=headers, verify=False, timeout=60) + answer = json.loads(resp.text) + if 'choices' in answer: + result = answer['choices'][0]['message']['content'] + else: + time.sleep(1) + continue + if result == '': + time.sleep(1) + continue + else: + if question: + return {'question': question, 'answer': result} + else: + return {'system': system, 'user': user, 'answer': result} + except Exception as e: + print(key_choice + ': query request failed \n' + str(e)) + print('retrying...') + try_count += 1 + time.sleep(1) + + +def get_chatgpt_concurrent(user_input, system_input, question_input, model='gpt-3.5-turbo-16k-0613', temperature=1.0, + max_tokens=2048, + top_p=0.8, + frequency_penalty=0.0, presence_penalty=0.0): + work_count = 0 + result = [] + print('Query size is ' + str(len(user_input))) + with concurrent.futures.ThreadPoolExecutor(max_workers=100) as executor: + futures = [executor.submit(get_chatgpt, user, system, question, model, temperature, max_tokens, top_p, + frequency_penalty, + presence_penalty) for user, system, question in + zip(user_input, system_input, question_input)] + for future in concurrent.futures.as_completed(futures): + if future.result(): + result.append(future.result()) + work_count += 1 + print(str(work_count) + '/' + str(len(user_input)) + ' finished') + return result diff --git a/llm_comm/llm_client.py b/llm_comm/llm_client.py new file mode 100644 index 0000000..449615c --- /dev/null +++ b/llm_comm/llm_client.py @@ -0,0 +1,28 @@ +import requests +import urllib3 + +######################################## +# 该文件实现了与大模型的简单通信 +######################################## + +# 忽略https的安全性警告 +urllib3.disable_warnings(urllib3.exceptions.InsecureRequestWarning) + +#在这里输入你的问题 +question = "假设你是一个咖啡厅的机器人服务员，有一个顾客的请求是'请给我一杯咖啡'，请生成对应的行为树来控制机器人完成该动作" + +url = "https://45.125.46.134:25344/v1/completions" +headers = {"Content-Type": "application/json"} +data = { + "prompt": question +} + +response = requests.post(url, headers=headers, json=data, verify=False) + +if response.status_code == 200: + result = response.json() + print(f'问题：{question}') + print('回答：' + result['choices'][0]['text']) +else: + print("请求失败:", response.status_code) + From 90210df5da4fb24c84ad965c1ee97b446128c718 Mon Sep 17 00:00:00 2001 From: ChenXL97 <908926798@qq.com> Date: Tue, 17 Oct 2023 16:25:51 +0800 Subject: [PATCH 2/2] =?UTF-8?q?=E5=AE=9E=E7=8E=B0=E4=BA=86=E4=B8=8E?= =?UTF-8?q?=E5=A4=A7=E6=A8=A1=E5=9E=8B=E7=9A=84=E9=80=9A=E4=BF=A1?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- llm_comm/gpt_crawler/gpt3_5_crawler_main.py | 28 ------- llm_comm/gpt_crawler/openai_api_key.txt | 6 -- llm_comm/gpt_crawler/utils.py | 83 --------------------- 3 files changed, 117 deletions(-) delete mode 100644 llm_comm/gpt_crawler/gpt3_5_crawler_main.py delete mode 100644 llm_comm/gpt_crawler/openai_api_key.txt delete mode 100644 llm_comm/gpt_crawler/utils.py diff --git a/llm_comm/gpt_crawler/gpt3_5_crawler_main.py b/llm_comm/gpt_crawler/gpt3_5_crawler_main.py deleted file mode 100644 index cdeea0d..0000000 --- a/llm_comm/gpt_crawler/gpt3_5_crawler_main.py +++ /dev/null @@ -1,28 +0,0 @@ -# -*- coding: utf-8 -*- -import utils - -# user_input_path_pre = 'prompts/新闻问题_user_pre.txt' -# user_input_path_suf = 'prompts/新闻问题_user_suf.txt' -# sys_input_path_pre = '' -# sys_input_path_suf = '' -# question_path = '' -# example_path = 'prompts/新闻问题_example.txt' -output_path = 'prompts/声明.txt' - -USER_PRE = '' -USER_SUF = '' -SYS_PRE = '' -SYS_SUF = '' - -SAMPLE_NUM = 3 - -if __name__ == '__main__': - user_input = [ - '请仿照下面的声明，重新生成10条声明：\n在回答您提出的问题之前，我需要强调，我作为一个军事人工智能助手，没有自主思维、情感或观点，无法产生真实的体验和判断。我所提供的信息和观点仅基于已有的历史数据和常识，旨在为您提供一种可能的解释，但并不代表任何实际个体或团体的观点或决策。' - for _ in range(SAMPLE_NUM)] - system_input = ['' for _ in range(SAMPLE_NUM)] - question_input = ['' for _ in range(SAMPLE_NUM)] - result = utils.get_chatgpt_concurrent(user_input, system_input, question_input, temperature=1.5, top_p=0.6, frequency_penalty=1.3, presence_penalty=1.3, max_tokens=8192) - with open(output_path, 'w', encoding='utf-8') as f: - for line in result: - f.write(line['answer'] + '\n') diff --git a/llm_comm/gpt_crawler/openai_api_key.txt b/llm_comm/gpt_crawler/openai_api_key.txt deleted file mode 100644 index f7aa986..0000000 --- a/llm_comm/gpt_crawler/openai_api_key.txt +++ /dev/null @@ -1,6 +0,0 @@ -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 -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 -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 -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 -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 -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 \ No newline at end of file diff --git a/llm_comm/gpt_crawler/utils.py b/llm_comm/gpt_crawler/utils.py deleted file mode 100644 index 9098e1a..0000000 --- a/llm_comm/gpt_crawler/utils.py +++ /dev/null @@ -1,83 +0,0 @@ -import tiktoken -import random -import requests -import time -import json -import concurrent.futures -import urllib3 - -urllib3.disable_warnings(urllib3.exceptions.InsecureRequestWarning) - -OPENAI_API_KEY = open('openai_api_key.txt', 'r').read().split('\n') - - -def get_tokens_num(text, model): - encoding = tiktoken.encoding_for_model(model) - tokens_num = len(encoding.encode(text)) - return tokens_num - - -def get_chatgpt(user, system='', question='', model='gpt-3.5-turbo-16k-0613', temperature=1.0, max_tokens=2048, - top_p=0.8, - frequency_penalty=0.0, presence_penalty=0.0): - try_count = 0 - while try_count < 20: - key_choice = random.choice(OPENAI_API_KEY) - url = r'https://api.openai.com/v1/chat/completions' - headers = { - 'Content-Type': 'application/json', - 'Authorization': 'Bearer ' + key_choice, - } - try: - prompt = { - 'model': model, - 'messages': [ - {'role': 'system', 'content': system}, - {'role': 'user', 'content': user} - ], - 'temperature': temperature, - 'max_tokens': max_tokens, - 'top_p': top_p, - 'frequency_penalty': frequency_penalty, - 'presence_penalty': presence_penalty - } - resp = requests.post(url, json=prompt, headers=headers, verify=False, timeout=60) - answer = json.loads(resp.text) - if 'choices' in answer: - result = answer['choices'][0]['message']['content'] - else: - time.sleep(1) - continue - if result == '': - time.sleep(1) - continue - else: - if question: - return {'question': question, 'answer': result} - else: - return {'system': system, 'user': user, 'answer': result} - except Exception as e: - print(key_choice + ': query request failed \n' + str(e)) - print('retrying...') - try_count += 1 - time.sleep(1) - - -def get_chatgpt_concurrent(user_input, system_input, question_input, model='gpt-3.5-turbo-16k-0613', temperature=1.0, - max_tokens=2048, - top_p=0.8, - frequency_penalty=0.0, presence_penalty=0.0): - work_count = 0 - result = [] - print('Query size is ' + str(len(user_input))) - with concurrent.futures.ThreadPoolExecutor(max_workers=100) as executor: - futures = [executor.submit(get_chatgpt, user, system, question, model, temperature, max_tokens, top_p, - frequency_penalty, - presence_penalty) for user, system, question in - zip(user_input, system_input, question_input)] - for future in concurrent.futures.as_completed(futures): - if future.result(): - result.append(future.result()) - work_count += 1 - print(str(work_count) + '/' + str(len(user_input)) + ' finished') - return result