Error when using pypots library

[nguyen-brat]

i install pypots library on my environment successfully but when running the sample code and it give me the error ImportError: cannot import name 'load_specific_dataset' from 'tsdb' (D:\ML\envs\tf_2\lib\site-packages\tsdb_init_.py). I also try it on google colab but it when i call pip install pypots it running forever, how to fix it thank you !!!!

import numpy as np
from sklearn.preprocessing import StandardScaler
from pypots.data import load_specific_dataset, mcar, masked_fill
from pypots.imputation import SAITS
from pypots.utils.metrics import cal_mae
data = load_specific_dataset('physionet_2012') # For datasets in PyPOTS database, PyPOTS will automatically download and extract it.
X = data['X']
num_samples = len(X['RecordID'].unique())
X = X.drop('RecordID', axis = 1)
X = StandardScaler().fit_transform(X.to_numpy())
X = X.reshape(num_samples, 48, -1)
X_intact, X, missing_mask, indicating_mask = mcar(X, 0.1) # hold out 10% observed values as ground truth
X = masked_fill(X, 1 - missing_mask, np.nan)
saits = SAITS(n_steps=48, n_features=37, n_layers=2, d_model=256, d_inner=128, n_head=4, d_k=64, d_v=64, dropout=0.1, epochs=10)
saits.fit(X) # train the model. Here I use the whole dataset as the training set, because ground truth is not visible to the model.
imputation = saits.impute(X) # impute the originally-missing values and artificially-missing values
mae = cal_mae(imputation, X_intact, indicating_mask) # calculate mean absolute error on the ground truth (artificially-missing values)

it get the error like this:
ImportError Traceback (most recent call last)
Input In [1], in
1 import numpy as np
2 from sklearn.preprocessing import StandardScaler
----> 3 from pypots.data import load_specific_dataset, mcar, masked_fill

File D:\ML\envs\tf_2\lib\site-packages\pypots_init_.py:9, in
1 """
2 PyPOTS package.
3 """
5 # Created by Wenjie Du wenjay.du@gmail.com
6 # License: GPL-v3
----> 9 from . import data
10 from . import imputation
11 from . import utils

File D:\ML\envs\tf_2\lib\site-packages\pypots\data_init_.py:14, in
8 from .base import (
9 BaseDataset,
10 DatasetForBRITS,
11 DatasetForMIT,
12 )
13 from .generating import generate_random_walk, generate_random_walk_for_classification
---> 14 from .integration import (
15 fill_nan_with_mask,
16 mcar,
17 load_specific_dataset,
18 list_available_datasets,
19 )

File D:\ML\envs\tf_2\lib\site-packages\pypots\data\integration.py:10, in
5 # Created by Wenjie Du wenjay.du@gmail.com
6 # License: GLP-v3
9 import pycorruptor as corruptor
---> 10 from tsdb import (
11 load_specific_dataset as _load_specific_dataset,
12 CACHED_DATASET_DIR as _CACHED_DATASET_DIR,
13 list_available_datasets as _list_available_datasets,
14 list_database as _list_database,
15 list_cached_data as _list_cached_data,
16 delete_cached_data as _delete_cached_data,
17 pickle_load as _pickle_load,
18 pickle_dump as _pickle_dump,
19 )
21 CACHED_DATASET_DIR = _CACHED_DATASET_DIR
22 list_database = _list_database

ImportError: cannot import name 'load_specific_dataset' from 'tsdb' (D:\ML\envs\tf_2\lib\site-packages\tsdb_init_.py)

[WenjieDu]

Hi, which version of TSDB you are using? And PyPOTS.

[WenjieDu]

If you use PyPOTS and SAITS and think they are helpful, please star⭐️ them to let more people know about them. This could be your contribution to PyPOTS and SAITS. Thank you in advance.

[nguyen-brat]

Sure, my team is using your library and my leader thinks it really helpfull, i will call them star your git

[WenjieDu]

Cool, that's really nice of you! Thank you very much for your support and recognition of my work.
Is your team currently working on partially-observed time series? About imputation?

[nguyen-brat]

Yeah! my team is woking with a dataset similar to bejjing PM2.5 dataset, some simple imputation method (XGBoost, LightGBM,..) work not really well and we are trying to using another advance and powerfull method !

[WenjieDu]

If your dataset is big enough, I believe deep-learning methods can indeed help.

BTW, is partially-observed time series (POTS) common to see in your projects? You know, my research work lies in this field, so I am keen to know about the situation in the industry. This can make me aware of the requirements of the industry, then I can develop models/algorithms accordingly. If you know pain points related to POTS in the industry, please let me know😉
I'm also working on some new ideas, which will be released in the near future. You can follow me on GitHub, so that you will receive the latest news about my work.

[nguyen-brat]

Actually this project is just a competition I joined with my club so I am not in the industry of this field and I don't know the demand or pain related to POTS but maybe in the future, if I know I will mail you later !!!!! Vào Th 5, 21 thg 7, 2022 vào lúc 13:21 Wenjie Du ***@***.***> đã viết:

…

If your dataset is big enough, I believe deep-learning methods can indeed help. BTW, is partially-observed time series (POTS) common to see in your projects? You know, my research work lies in this field, so I am keen to know about the situation in the industry. This can make me aware of the requirements of the industry, then I can develop models/algorithms accordingly. If you know pain points related to POTS in the industry, please let me know😉 I'm also working on some new ideas, which will be released in the near future. You can follow me on GitHub, so that you will receive the latest news about my work. — Reply to this email directly, view it on GitHub <#6 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AWXAZ6CJH34MN6QDZGT26GDVVDT5TANCNFSM54GDJJ4Q> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[WenjieDu]

Cool. Many thanks. Your feedback is always welcome ;-)

[RafaelSee]

thank you so much mr. wenjie