main.py

import torch

torch.manual_seed(32)

"""
# ! 进行数据清洗，提炼出符合条件的数据充足的用户
from data_featuring_scripts import TravelDataFilter
data_filter = TravelDataFilter(
    need_filter=False, 
    dataset_path="datasets/example_sets/travel_behavior/dataset_TIST2015_Checkins_withPOIs_example.parquet")
data_filter.filter_user_and_checkins()
# """

"""
# ! 进行自然语言化处理
from data_featuring_scripts import LanguageConverter
language_converter = LanguageConverter()
language_converter.form_diary()
# """

"""
# ! 进行自然语言化处理
from data_featuring_scripts import LanguageConverter
language_converter = LanguageConverter(dataset_path="datasets/example_sets/travel_behavior/dataset_TIST2015_Checkins_withPOIs_example_valid.parquet")
language_converter.form_diary()
# """

"""
# ! 数据模型的 dataset 测试
from data_featuring_scripts import TravelBCPDatasetLanguage
from utils import ProgressBar
from pprint import pprint
from torch.utils.data import Subset, DataLoader
bcp_dataset_language = TravelBCPDatasetLanguage(
    context_window_size=7,
    dataset_path="datasets/example_sets/travel_behavior/dataset_TIST2015_Checkins_withPOIs_example_valid.parquet",
    diary_path="datasets/example_sets/travel_behavior/dataset_TIST2015_Checkins_withPOIs_example_valid_diaries.parquet"
)
train_indices = bcp_dataset_language.train_indices
val_indices = bcp_dataset_language.val_indices
test_indices = bcp_dataset_language.test_indices
train_set = Subset(bcp_dataset_language, train_indices)
val_set = Subset(bcp_dataset_language, val_indices)
test_set = Subset(bcp_dataset_language, test_indices)
print(len(train_set), len(val_set), len(test_set))
train_loader = DataLoader(train_set, batch_size=64, shuffle=True, num_workers=8)
val_loader = DataLoader(val_set, batch_size=64, shuffle=False, num_workers=8)
test_loader = DataLoader(test_set, batch_size=64, shuffle=False, num_workers=8)
print(len(train_loader), len(val_loader), len(test_loader))
for batch in ProgressBar(test_loader, description="test travel bcp dataset discrete train loader"):
    batch_shape = {key: len(value) for key, value in batch.items()}
    # print(batch['context Day'])
    # input()
# """


"""
# ! 对于 api 的测试
from apis import YunwuAPI
import asyncio

yunwu_api = YunwuAPI()
message_list = [
    [{"role": 'user', "content": "Hello, who are you?"}],
    [{"role": 'user', "content": "What can you do?"}],
    [{"role": 'user', "content": "Tell me a joke."}]
]

result = asyncio.run(yunwu_api.chat_completion_async_batch(message_list))
print(result)
# """

"""
# llmbaseline 在media的example数据集上的尝试
# import pandas as pd
# from apis import YunwuAPI
# import asyncio
# import json
# from data_featuring_scripts import MediaBCPDatasetLanguage, MediaLanguageConverter

# media_dataset=MediaBCPDatasetLanguage()
# length = len(media_dataset)
# system_prompt =(
#     "you are a human media behavior analyst. Please generate the predicted rating and comment based on the user's current state, the context of the business, and the user's historical behavior."
#     "comment should be written as if you are the user."
#     "the output should be a JSON object with keys 'rating' and 'comment'."
    
# )
# message_list = []
# user_id_list = []
# for i in range(length):
#     info_dict = media_dataset[i]
#     llm_text = (
#         info_dict.get('context current', '') +
#         info_dict.get('context business', '') +
#         info_dict.get('context user', '') +
#         info_dict.get('context diaries', '')
#     )
#     message = [
#         {"role": "system", "content": system_prompt},
#         {"role": "user", "content": llm_text}
#     ]
#     # print(llm_text)
#     # print()
#     # input("press the ENTER")
#     message_list.append(message)
#     user_id_list.append(info_dict.get('user_id', f'idx_{i}'))

# # 实例化 YunwuAPI 并批量调用
# yunwu_api = YunwuAPI()
# result = asyncio.run(yunwu_api.chat_completion_async_batch(message_list))

# # 结果
# ratings = []
# comments = []
# for res in result:
#     try:
#         data = json.loads(res)
#         ratings.append(data.get('rating', None))
#         comments.append(data.get('comment', None))
#     except Exception as e:
#         ratings.append(None)
#         comments.append(None)

# # 保存为DataFrame
# result_df = pd.DataFrame({
#     'user_id': user_id_list,
#     'rating': ratings,
#     'comment': comments
# })
# print(result_df.head())
# result_df.to_csv('yunwuapi_results.csv', index=False)

from baseline_tester import toy_baseline
toy_baseline()
# """


"""
from baseline_tester import TravelBehaviorDiscreteTrainer, TravelBehaviorDiscreteTester

travel_behavior_discrete_trainer = TravelBehaviorDiscreteTrainer()
travel_behavior_discrete_tester = TravelBehaviorDiscreteTester()
travel_behavior_discrete_trainer.train()
travel_behavior_discrete_tester.test()
# """

"""
from baseline_tester import TravelBehaviorLanguageTrainer, TravelBehaviorLanguageTester

travel_behavior_language_tester = TravelBehaviorLanguageTester()
travel_behavior_language_tester.test()
# """