config/FPT/AirQuality.yaml

@@ -0,0 +1,51 @@
+basic:
+  dataset: AirQuality
+  device: cuda:0
+  mode: train
+  model: FPT
+  seed: 2023
+
+data:
+  batch_size: 16
+  column_wise: false
+  days_per_week: 7
+  horizon: 24
+  input_dim: 6
+  lag: 24
+  normalizer: std
+  num_nodes: 35
+  steps_per_day: 24
+  test_ratio: 0.2
+  val_ratio: 0.2
+
+model:
+  d_model: 768
+  gpt_layers: 9
+  gpt_path: ./GPT-2
+  input_dim: 6
+  n_heads: 1
+  num_nodes: 35
+  patch_len: 6
+  pred_len: 24
+  seq_len: 24
+  stride: 7
+
+train:
+  batch_size: 16
+  debug: false
+  early_stop: true
+  early_stop_patience: 15
+  epochs: 100
+  grad_norm: false
+  log_step: 100
+  loss_func: mae
+  lr_decay: true
+  lr_decay_rate: 0.3
+  lr_decay_step: 5,20,40,70
+  lr_init: 0.003
+  mae_thresh: None
+  mape_thresh: 0.001
+  max_grad_norm: 5
+  output_dim: 6
+  plot: false
+  weight_decay: 0

config/FPT/BJTaxi-InFlow.yaml

@@ -0,0 +1,51 @@
+basic:
+  dataset: BJTaxi-InFlow
+  device: cuda:0
+  mode: train
+  model: FPT
+  seed: 2023
+
+data:
+  batch_size: 32
+  column_wise: false
+  days_per_week: 7
+  horizon: 24
+  input_dim: 1
+  lag: 24
+  normalizer: std
+  num_nodes: 1024
+  steps_per_day: 48
+  test_ratio: 0.2
+  val_ratio: 0.2
+
+model:
+  d_model: 768
+  gpt_layers: 9
+  gpt_path: ./GPT-2
+  input_dim: 1
+  n_heads: 1
+  num_nodes: 1024
+  patch_len: 6
+  pred_len: 24
+  seq_len: 24
+  stride: 7
+
+train:
+  batch_size: 32
+  debug: false
+  early_stop: true
+  early_stop_patience: 15
+  epochs: 100
+  grad_norm: false
+  log_step: 100
+  loss_func: mae
+  lr_decay: true
+  lr_decay_rate: 0.3
+  lr_decay_step: 5,20,40,70
+  lr_init: 0.003
+  mae_thresh: None
+  mape_thresh: 0.001
+  max_grad_norm: 5
+  output_dim: 1
+  plot: false
+  weight_decay: 0

config/FPT/BJTaxi-OutFlow.yaml

@@ -0,0 +1,51 @@
+basic:
+  dataset: BJTaxi-OutFlow
+  device: cuda:0
+  mode: train
+  model: FPT
+  seed: 2023
+
+data:
+  batch_size: 32
+  column_wise: false
+  days_per_week: 7
+  horizon: 24
+  input_dim: 1
+  lag: 24
+  normalizer: std
+  num_nodes: 1024
+  steps_per_day: 48
+  test_ratio: 0.2
+  val_ratio: 0.2
+
+model:
+  d_model: 768
+  gpt_layers: 9
+  gpt_path: ./GPT-2
+  input_dim: 1
+  n_heads: 1
+  num_nodes: 1024
+  patch_len: 6
+  pred_len: 24
+  seq_len: 24
+  stride: 7
+
+train:
+  batch_size: 32
+  debug: false
+  early_stop: true
+  early_stop_patience: 15
+  epochs: 100
+  grad_norm: false
+  log_step: 100
+  loss_func: mae
+  lr_decay: true
+  lr_decay_rate: 0.3
+  lr_decay_step: 5,20,40,70
+  lr_init: 0.003
+  mae_thresh: None
+  mape_thresh: 0.001
+  max_grad_norm: 5
+  output_dim: 1
+  plot: false
+  weight_decay: 0

config/FPT/METR-LA.yaml

@@ -0,0 +1,52 @@
+basic:
+  dataset: METR-LA
+  device: cuda:0
+  mode: train
+  model: FPT
+  seed: 2023
+
+data:
+  batch_size: 16
+  column_wise: false
+  days_per_week: 7
+  horizon: 24
+  input_dim: 1
+  lag: 24
+  normalizer: std
+  num_nodes: 207
+  steps_per_day: 288
+  test_ratio: 0.2
+  val_ratio: 0.2
+
+model:
+  d_model: 768
+  gpt_layers: 9
+  gpt_path: ./GPT-2
+  input_dim: 1
+  n_heads: 1
+  num_nodes: 207
+  patch_len: 6
+  pred_len: 24
+  seq_len: 24
+  stride: 7
+
+train:
+  batch_size: 16
+  debug: false
+  early_stop: true
+  early_stop_patience: 15
+  epochs: 100
+  grad_norm: false
+  log_step: 1000
+  loss_func: mae
+  lr_decay: true
+  lr_decay_rate: 0.3
+  lr_decay_step: 5,20,40,70
+  lr_init: 0.003
+  mae_thresh: None
+  mape_thresh: 0.001
+  max_grad_norm: 5
+  output_dim: 1
+  plot: false
+  real_value: true
+  weight_decay: 0

config/FPT/NYCBike-InFlow.yaml

@@ -0,0 +1,51 @@
+basic:
+  dataset: NYCBike-InFlow
+  device: cuda:0
+  mode: train
+  model: FPT
+  seed: 2023
+
+data:
+  batch_size: 32
+  column_wise: false
+  days_per_week: 7
+  horizon: 24
+  input_dim: 1
+  lag: 24
+  normalizer: std
+  num_nodes: 128
+  steps_per_day: 48
+  test_ratio: 0.2
+  val_ratio: 0.2
+
+model:
+  d_model: 768
+  gpt_layers: 9
+  gpt_path: ./GPT-2
+  input_dim: 1
+  n_heads: 1
+  num_nodes: 128
+  patch_len: 6
+  pred_len: 24
+  seq_len: 24
+  stride: 7
+
+train:
+  batch_size: 32
+  debug: false
+  early_stop: true
+  early_stop_patience: 15
+  epochs: 100
+  grad_norm: false
+  log_step: 100
+  loss_func: mae
+  lr_decay: true
+  lr_decay_rate: 0.3
+  lr_decay_step: 5,20,40,70
+  lr_init: 0.003
+  mae_thresh: None
+  mape_thresh: 0.001
+  max_grad_norm: 5
+  output_dim: 1
+  plot: false
+  weight_decay: 0

config/FPT/NYCBike-OutFlow.yaml

@@ -0,0 +1,51 @@
+basic:
+  dataset: NYCBike-OutFlow
+  device: cuda:0
+  mode: train
+  model: FPT
+  seed: 2023
+
+data:
+  batch_size: 32
+  column_wise: false
+  days_per_week: 7
+  horizon: 24
+  input_dim: 1
+  lag: 24
+  normalizer: std
+  num_nodes: 128
+  steps_per_day: 48
+  test_ratio: 0.2
+  val_ratio: 0.2
+
+model:
+  d_model: 768
+  gpt_layers: 9
+  gpt_path: ./GPT-2
+  input_dim: 1
+  n_heads: 1
+  num_nodes: 128
+  patch_len: 6
+  pred_len: 24
+  seq_len: 24
+  stride: 7
+
+train:
+  batch_size: 32
+  debug: false
+  early_stop: true
+  early_stop_patience: 15
+  epochs: 100
+  grad_norm: false
+  log_step: 100
+  loss_func: mae
+  lr_decay: true
+  lr_decay_rate: 0.3
+  lr_decay_step: 5,20,40,70
+  lr_init: 0.003
+  mae_thresh: None
+  mape_thresh: 0.001
+  max_grad_norm: 5
+  output_dim: 1
+  plot: false
+  weight_decay: 0

config/FPT/PEMS-BAY.yaml

@@ -0,0 +1,51 @@
+basic:
+  dataset: PEMS-BAY
+  device: cuda:0
+  mode: train
+  model: FPT
+  seed: 2023
+
+data:
+  batch_size: 16
+  column_wise: false
+  days_per_week: 7
+  horizon: 24
+  input_dim: 1
+  lag: 24
+  normalizer: std
+  num_nodes: 325
+  steps_per_day: 288
+  test_ratio: 0.2
+  val_ratio: 0.2
+
+model:
+  d_model: 768
+  gpt_layers: 9
+  gpt_path: ./GPT-2
+  input_dim: 1
+  n_heads: 1
+  num_nodes: 325
+  patch_len: 6
+  pred_len: 24
+  seq_len: 24
+  stride: 7
+
+train:
+  batch_size: 16
+  debug: false
+  early_stop: true
+  early_stop_patience: 15
+  epochs: 100
+  grad_norm: false
+  log_step: 100
+  loss_func: mae
+  lr_decay: true
+  lr_decay_rate: 0.3
+  lr_decay_step: 5,20,40,70
+  lr_init: 0.003
+  mae_thresh: None
+  mape_thresh: 0.001
+  max_grad_norm: 5
+  output_dim: 1
+  plot: false
+  weight_decay: 0

config/FPT/SolarEnergy.yaml

@@ -0,0 +1,51 @@
+basic:
+  dataset: SolarEnergy
+  device: cuda:0
+  mode: train
+  model: FPT
+  seed: 2023
+
+data:
+  batch_size: 16
+  column_wise: false
+  days_per_week: 7
+  horizon: 24
+  input_dim: 1
+  lag: 24
+  normalizer: std
+  num_nodes: 137
+  steps_per_day: 24
+  test_ratio: 0.2
+  val_ratio: 0.2
+
+model:
+  d_model: 768
+  gpt_layers: 9
+  gpt_path: ./GPT-2
+  input_dim: 1
+  n_heads: 1
+  num_nodes: 137
+  patch_len: 6
+  pred_len: 24
+  seq_len: 24
+  stride: 7
+
+train:
+  batch_size: 16
+  debug: false
+  early_stop: true
+  early_stop_patience: 15
+  epochs: 100
+  grad_norm: false
+  log_step: 100
+  loss_func: mae
+  lr_decay: true
+  lr_decay_rate: 0.3
+  lr_decay_step: 5,20,40,70
+  lr_init: 0.003
+  mae_thresh: None
+  mape_thresh: 0.001
+  max_grad_norm: 5
+  output_dim: 1
+  plot: false
+  weight_decay: 0

model/ASTRA/astra.py

@@ -206,7 +206,6 @@ class ASTRA(nn.Module):
         enc_out, n_vars = self.patch_embedding(x_enc) # (B, N, d_model * input_dim)
         # 应用图增强编码器（自动生成图结构）
         graph_enhanced = self.graph_encoder(enc_out) # (B, N, K * hidden_dim)
-        # 特征融合 - 现在两个张量都是三维的 [B, N, d_model]
         enc_out = torch.cat([enc_out, graph_enhanced], dim=-1)
         enc_out = self.feature_fusion(enc_out)
         

model/FPT/fpt.py

@@ -0,0 +1,45 @@
+import torch.nn as nn
+from transformers.models.gpt2.modeling_gpt2 import GPT2Model
+from einops import rearrange
+
+class fpt(nn.Module):
+    def __init__(self, configs):
+        super(fpt, self).__init__()
+        self.patch_len = configs['patch_len']
+        self.stride = configs['stride']
+        self.input_dim = configs['input_dim']
+        self.seq_len = configs['seq_len']
+        self.pred_len = configs['pred_len']
+        self.gpt_layers = configs['gpt_layers']  # 使用的GPT2层数
+        self.d_model = configs['d_model']
+        self.gpt_path = configs['gpt_path']
+
+        self.patch_num = int((self.seq_len - self.patch_len) / self.stride + 2)  # 补丁数量
+        self.padding_patch_layer = nn.ReplicationPad1d((0, self.stride))
+
+        self.gpts = GPT2Model.from_pretrained(self.gpt_path, output_attentions=True, output_hidden_states=True)
+        self.gpts.h = self.gpts.h[:self.gpt_layers]
+        for i, (name, param) in enumerate(self.gpts.named_parameters()):
+                if 'wpe' in name:
+                    param.requires_grad = True
+                else:
+                    param.requires_grad = False
+        
+        self.in_layer = nn.Linear(self.patch_len, self.d_model)
+        self.out_layer = nn.Linear(self.d_model * self.patch_num, self.pred_len)
+
+    def forward(self, x):
+        B, L, M = x.shape
+        x = x[..., :self.input_dim]
+        x = rearrange(x, 'b l m -> b m l')
+
+        x = self.padding_patch_layer(x)
+        x = x.unfold(dimension = -1, size = self.patch_len, step = self.stride)
+        x = rearrange(x, 'b m n p -> (b m) n p')
+
+        outputs = self.in_layer(x)
+        outputs = self.gpts(inputs_embeds=outputs).last_hidden_state
+        outputs = self.out_layer(outputs.reshape(B*M, -1))
+        outputs = rearrange(outputs, '(b m) l -> b l m', b = B)
+        return outputs
+

model/FPT/model_config.json

@@ -0,0 +1,7 @@
+[
+  {
+    "name": "FPT",
+    "module": "model.FPT.fpt",
+    "entry": "fpt"
+  }
+]

train.py

@@ -90,9 +90,9 @@ def main(model, data, debug=False):
 if __name__ == "__main__":
     # 调试用
     # model_list = ["iTransformer", "PatchTST", "HI"]
-    # model_list = ["ASTRA_v3"]
-    model_list = ["PatchTST"]
-    dataset_list = ["AirQuality", "SolarEnergy", "PEMS-BAY", "METR-LA", "BJTaxi-InFlow", "BJTaxi-OutFlow", "NYCBike-InFlow", "NYCBike-OutFlow"]
+    model_list = ["FPT"]
+    # model_list = ["PatchTST"]
+    dataset_list = ["METR-LA", "BJTaxi-InFlow", "BJTaxi-OutFlow", "NYCBike-InFlow", "NYCBike-OutFlow"]
     # dataset_list = ["AirQuality"]
     # dataset_list = ["METR-LA"]
     main(model_list, dataset_list, debug = False)

trainer/trainer_selector.py

@@ -20,7 +20,7 @@ def select_trainer(
         scaler, args, lr_scheduler
     )
 
-    if model_name in {"HI", "PatchTST", "iTransformer"}:
+    if model_name in {"HI", "PatchTST", "iTransformer", "FPT"}:
         return TSTrainer(*base_args)
 
     trainer_map = {