Works on GPU

data_provider/data_factory.py

@@ -1,4 +1,4 @@
-from data_provider.data_loader import Dataset_ETT_hour, Dataset_ETT_minute, Dataset_Custom, Dataset_M4
+from data_provider.data_loader import Dataset_ean, Dataset_ETT_hour, Dataset_ETT_minute, Dataset_Custom, Dataset_M4
 from torch.utils.data import DataLoader
 
 data_dict = {
@@ -10,6 +10,7 @@
     'Traffic': Dataset_Custom,
     'Weather': Dataset_Custom,
     'm4': Dataset_M4,
+    'ean': Dataset_ean,
 }
 
 
@@ -42,6 +43,7 @@ def data_provider(args, flag):
             freq=freq,
             seasonal_patterns=args.seasonal_patterns
         )
+
     else:
         data_set = Data(
             root_path=args.root_path,

data_provider/data_loader.py

@@ -10,6 +10,80 @@
 warnings.filterwarnings('ignore')
 
 
+class Dataset_ean(Dataset):
+    def __init__(self, root_path, flag='train', size=None,
+                 features='S', data_path='data.csv',
+                 target='sold_units', scale=False, timeenc=0, freq='W', percent=100,
+                 seasonal_patterns=None):
+        if size is None:
+            self.seq_len = 13  # Use 0.25 year of data for sequence
+            self.label_len = 4  # Labels from last month
+            self.pred_len = 4   # Predict 1 month ahead
+        else:
+            self.seq_len, self.label_len, self.pred_len = size
+
+        assert flag in ['train', 'test', 'val']
+        self.set_type = {'train': 0, 'val': 1, 'test': 2}[flag]
+
+        self.features = features
+        self.target = target
+        self.scale = scale
+        self.timeenc = timeenc
+        self.freq = freq
+
+        self.root_path = root_path
+        self.data_path = data_path
+        self.__read_data__()
+        self.enc_in = self.data_x.shape[-1]
+    def __read_data__(self):
+        self.scaler = StandardScaler()
+        df_raw = pd.read_csv(os.path.join(self.root_path, self.data_path))
+        df_raw.drop("Unnamed: 0", axis=1, inplace=True)# don't forget to remove this once regulated the data
+        num_weeks = len(df_raw)
+        num_train = int(num_weeks * 0.7)
+        num_val = int(num_weeks * 0.2)
+        num_test = num_weeks - num_train - num_val
+
+        border1s = [0, num_train, num_train + num_val]
+        border2s = [num_train, num_train + num_val, num_weeks]
+
+        border1 = border1s[self.set_type]
+        border2 = border2s[self.set_type]
+
+        if self.scale:
+            train_data = df_raw.iloc[:num_train]
+            self.scaler.fit(train_data[[self.target]].values)
+            data = self.scaler.transform(df_raw[[self.target]].values)
+        else:
+            data = df_raw[[self.target]].values
+
+        df_stamp = pd.to_datetime(df_raw.iloc[:, 0][border1:border2])  # 'end_date' is first column
+
+        time_features = np.vstack((df_stamp.dt.year, df_stamp.dt.month, df_stamp.dt.day, df_stamp.dt.weekday)).T
+
+
+        self.data_x = data[border1:border2]
+        self.data_y = data[border1:border2]
+        self.time_features = time_features
+
+
+
+    def __getitem__(self, index):
+        seq_x = self.data_x[index:index+self.seq_len]
+        seq_y = self.data_y[index+self.seq_len:index+self.seq_len+self.pred_len]
+        seq_x_mark = self.time_features[index:index+self.seq_len]
+        seq_y_mark = self.time_features[index+self.seq_len:index+self.seq_len+self.pred_len]
+        return seq_x, seq_y, seq_x_mark, seq_y_mark
+
+    def __len__(self):
+        return (len(self.data_x) - self.seq_len - self.pred_len + 1) * self.enc_in
+
+    def inverse_transform(self, data):
+        return self.scaler.inverse_transform(data)
+
+
+
+
 class Dataset_ETT_hour(Dataset):
     def __init__(self, root_path, flag='train', size=None,
                  features='S', data_path='ETTh1.csv',

dataset/data.csv

@@ -0,0 +1,222 @@
+,end_date,sold_units
+19,2020-01-04,89.0
+87,2020-01-11,89.0
+100,2020-01-18,67.0
+2,2020-01-25,82.0
+104,2020-02-01,96.0
+132,2020-02-08,113.0
+95,2020-02-15,94.0
+18,2020-02-22,69.0
+67,2020-02-29,128.0
+138,2020-03-07,192.0
+79,2020-03-14,113.0
+38,2020-03-21,104.0
+96,2020-03-28,111.0
+131,2020-04-04,74.0
+11,2020-04-11,21.0
+119,2020-04-18,4.0
+108,2020-04-25,0.0
+133,2020-05-02,43.0
+53,2020-05-09,321.0
+117,2020-05-16,237.0
+144,2020-05-23,217.0
+48,2020-05-30,234.0
+101,2020-06-06,356.0
+137,2020-06-13,317.0
+85,2020-06-20,415.0
+156,2020-06-27,972.0
+31,2020-07-04,313.0
+115,2020-07-11,317.0
+33,2020-07-18,306.0
+116,2020-07-25,216.0
+97,2020-08-01,215.0
+149,2020-08-08,211.0
+21,2020-08-15,167.0
+112,2020-08-22,148.0
+37,2020-08-29,191.0
+35,2020-09-05,169.0
+110,2020-09-12,244.0
+56,2020-09-19,146.0
+126,2020-09-26,176.0
+29,2020-10-03,381.0
+118,2020-10-10,127.0
+123,2020-10-17,88.0
+98,2020-10-24,89.0
+127,2020-10-31,200.0
+154,2020-11-07,156.0
+151,2020-11-14,254.0
+45,2020-11-21,260.0
+143,2020-11-28,243.0
+125,2020-12-05,254.0
+140,2020-12-12,191.0
+102,2020-12-19,191.0
+105,2020-12-26,87.0
+41,2021-01-02,334.0
+83,2021-01-09,171.0
+77,2021-01-16,182.0
+46,2021-01-23,163.0
+129,2021-01-30,135.0
+94,2021-02-06,177.0
+71,2021-02-13,159.0
+139,2021-02-20,370.0
+9,2021-02-27,277.0
+20,2021-03-06,332.0
+136,2021-03-13,185.0
+15,2021-03-20,157.0
+36,2021-03-27,196.0
+121,2021-04-03,226.0
+14,2021-04-10,159.0
+43,2021-04-17,381.0
+44,2021-04-24,147.0
+153,2021-05-01,117.0
+141,2021-05-08,130.0
+122,2021-05-15,108.0
+23,2021-05-22,86.0
+32,2021-05-29,64.0
+1,2021-06-05,99.0
+60,2021-06-12,149.0
+73,2021-06-19,158.0
+130,2021-06-26,130.0
+106,2021-07-03,175.0
+72,2021-07-10,128.0
+148,2021-07-17,101.0
+59,2021-07-24,92.0
+91,2021-07-31,129.0
+4,2021-08-07,192.0
+47,2021-08-14,190.0
+27,2021-08-21,221.0
+13,2021-08-28,123.0
+24,2021-09-04,132.0
+128,2021-09-11,98.0
+113,2021-09-18,87.0
+150,2021-09-25,97.0
+25,2021-10-02,112.0
+42,2021-10-09,101.0
+30,2021-10-16,112.0
+86,2021-10-23,85.0
+8,2021-10-30,91.0
+12,2021-11-06,83.0
+78,2021-11-13,128.0
+0,2021-11-20,112.0
+50,2021-11-27,164.0
+84,2021-12-04,331.0
+28,2021-12-11,162.0
+7,2021-12-18,113.0
+92,2021-12-25,74.0
+114,2022-01-01,92.0
+40,2022-01-08,107.0
+76,2022-01-15,100.0
+63,2022-01-22,77.0
+34,2022-01-29,94.0
+80,2022-02-05,83.0
+3,2022-02-12,135.0
+88,2022-02-19,106.0
+107,2022-02-26,108.0
+147,2022-03-05,126.0
+145,2022-03-12,139.0
+146,2022-03-19,104.0
+5,2022-03-26,112.0
+90,2022-04-02,115.0
+6,2022-04-09,121.0
+109,2022-04-16,102.0
+152,2022-04-23,180.0
+135,2022-04-30,286.0
+54,2022-05-07,187.0
+61,2022-05-14,171.0
+89,2022-05-21,141.0
+134,2022-05-28,157.0
+66,2022-06-04,113.0
+10,2022-06-11,138.0
+81,2022-06-18,149.0
+75,2022-06-25,154.0
+55,2022-07-02,208.0
+111,2022-07-09,259.0
+93,2022-07-16,224.0
+26,2022-07-23,198.0
+52,2022-07-30,230.0
+58,2022-08-06,102.0
+120,2022-08-13,152.0
+22,2022-08-20,184.0
+70,2022-08-27,189.0
+49,2022-09-03,153.0
+69,2022-09-10,130.0
+103,2022-09-17,135.0
+155,2022-09-24,176.0
+39,2022-10-01,180.0
+68,2022-10-08,246.0
+16,2022-10-15,348.0
+57,2022-10-22,392.0
+99,2022-10-29,435.0
+51,2022-11-05,276.0
+124,2022-11-12,188.0
+17,2022-11-19,169.0
+82,2022-11-26,360.0
+74,2022-12-03,420.0
+65,2022-12-10,219.0
+64,2022-12-17,207.0
+62,2022-12-24,150.0
+142,2022-12-31,131.0
+179,2023-01-07,184.0
+211,2023-01-14,139.0
+160,2023-01-21,126.0
+159,2023-01-28,146.0
+196,2023-02-04,168.0
+212,2023-02-11,140.0
+169,2023-02-18,136.0
+180,2023-02-25,96.0
+168,2023-03-04,124.0
+220,2023-03-11,153.0
+224,2023-03-18,160.0
+206,2023-03-25,145.0
+185,2023-04-01,184.0
+190,2023-04-08,164.0
+188,2023-04-15,143.0
+186,2023-04-22,166.0
+213,2023-04-29,144.0
+230,2023-05-06,131.0
+227,2023-05-13,161.0
+222,2023-05-20,219.0
+187,2023-05-27,239.0
+229,2023-06-03,307.0
+162,2023-06-10,291.0
+171,2023-06-17,251.0
+191,2023-06-24,187.0
+183,2023-07-01,237.0
+199,2023-07-08,204.0
+209,2023-07-15,192.0
+172,2023-07-22,180.0
+167,2023-07-29,152.0
+192,2023-08-05,177.0
+203,2023-08-12,135.0
+181,2023-08-19,170.0
+184,2023-08-26,116.0
+197,2023-09-02,162.0
+202,2023-09-09,193.0
+231,2023-09-16,152.0
+182,2023-09-23,164.0
+174,2023-09-30,196.0
+223,2023-10-07,200.0
+175,2023-10-14,196.0
+201,2023-10-21,143.0
+232,2023-10-28,250.0
+207,2023-11-04,278.0
+177,2023-11-11,508.0
+210,2023-11-18,148.0
+195,2023-11-25,244.0
+173,2023-12-02,488.0
+204,2023-12-09,209.0
+208,2023-12-16,201.0
+163,2023-12-23,164.0
+164,2023-12-30,154.0
+165,2024-01-06,200.0
+215,2024-01-13,181.0
+161,2024-01-20,143.0
+218,2024-01-27,238.0
+226,2024-02-03,190.0
+233,2024-02-10,145.0
+216,2024-02-17,126.0
+194,2024-02-24,198.0
+214,2024-03-02,149.0
+219,2024-03-09,148.0
+157,2024-03-16,146.0
+221,2024-03-23,100.0

dataset/prompt_bank/ean.txt

@@ -0,0 +1 @@
+EAN is a dataset contains only sold_units of one ean product of loreal cosmetic products.