Add SFM config

README.md

@@ -198,6 +198,7 @@ Here is a list of models built on `Qlib`.
 - [LSTM based on pytorcn](qlib/contrib/model/pytorch_lstm.py)
 - [GATs based on pytorch](qlib/contrib/model/pytorch_gats.py)
 - [TabNet based on pytorch](qlib/contrib/model/tabnet.py)
+- [SFM based on pytorch](qlib/contrib/model/pytorch_sfm.py)
 <!-- - [TFT based on tensorflow](examples/benchmarks/TFT/tft.py) -->
 
 Your PR of new Quant models is highly welcomed.

examples/benchmarks/SFM/requirements.txt

@@ -0,0 +1,4 @@
+pandas==1.1.2
+numpy==1.17.4
+scikit_learn==0.23.2
+torch==1.7.0

examples/benchmarks/SFM/workflow_config_sfm.yaml

@@ -0,0 +1,73 @@
+provider_uri: "~/.qlib/qlib_data/cn_data"
+region: cn
+market: &market csi300
+benchmark: &benchmark SH000300
+data_handler_config: &data_handler_config
+    start_time: 2008-01-01
+    end_time: 2020-08-01
+    fit_start_time: 2008-01-01
+    fit_end_time: 2014-12-31
+    instruments: *market
+port_analysis_config: &port_analysis_config
+    strategy:
+        class: TopkDropoutStrategy
+        module_path: qlib.contrib.strategy.strategy
+        kwargs:
+            topk: 50
+            n_drop: 5
+    backtest:
+        verbose: False
+        limit_threshold: 0.095
+        account: 100000000
+        benchmark: *benchmark
+        deal_price: close
+        open_cost: 0.0005
+        close_cost: 0.0015
+        min_cost: 5
+task:
+    model:
+        class: SFM
+        module_path: qlib.contrib.model.pytorch_sfm
+        kwargs:
+            d_feat: 6
+            hidden_size: 64
+            output_dim: 1
+            freq_dim: 15
+            dropout_W: 0.5
+            dropout_U: 0.5
+            n_epochs: 10
+            lr: 1e-3
+            batch_size: 800
+            early_stop: 20
+            eval_steps: 5
+            loss: mse
+            lr_decay: 0.96
+            lr_decay_steps: 100
+            optimizer: gd
+            GPU: 1
+            seed: 0
+    dataset:
+        class: DatasetH
+        module_path: qlib.data.dataset
+        kwargs:
+            handler:
+                class: ALPHA360_Denoise
+                module_path: qlib.contrib.data.handler
+                kwargs: *data_handler_config
+            segments:
+                train: [2008-01-01, 2014-12-31]
+                valid: [2015-01-01, 2016-12-31]
+                test: [2017-01-01, 2020-08-01]
+    record: 
+        - class: SignalRecord
+          module_path: qlib.workflow.record_temp
+          kwargs: {}
+        - class: SigAnaRecord
+          module_path: qlib.workflow.record_temp
+          kwargs: 
+            ana_long_short: False
+            ann_scaler: 252
+        - class: PortAnaRecord
+          module_path: qlib.workflow.record_temp
+          kwargs: 
+            config: *port_analysis_config

examples/workflow_by_code_sfm.py

@@ -62,8 +62,8 @@ if __name__ == "__main__":
             "kwargs": {
                 "d_feat": 6,
                 "hidden_size": 64,
-                "output_dim" : 1,
+                "output_dim": 1,
-                "freq_dim" : 15,
+                "freq_dim": 15,
                 "dropout_W": 0.5,
                 "dropout_U": 0.5,
                 "n_epochs": 10,
@@ -72,9 +72,9 @@ if __name__ == "__main__":
                 "early_stop": 20,
                 "eval_steps": 5,
                 "loss": "mse",
-                "lr_decay" : 0.96,
+                "lr_decay": 0.96,
-                "lr_decay_steps" : 100,
+                "lr_decay_steps": 100,
-                "optimizer" : "gd",
+                "optimizer": "gd",
                 "GPU": 1,
                 "seed": 0,
             },

qlib/contrib/model/pytorch_sfm.py

@@ -21,11 +21,12 @@ from ...model.base import Model
 from ...data.dataset import DatasetH
 from ...data.dataset.handler import DataHandlerLP
 
+
 class SFM_Model(nn.Module):
-    def __init__(self, d_feat=6, output_dim = 1, freq_dim = 10, hidden_size = 64, dropout_W = 0.0, dropout_U = 0.0, device = "cpu"):
+    def __init__(self, d_feat=6, output_dim=1, freq_dim=10, hidden_size=64, dropout_W=0.0, dropout_U=0.0, device="cpu"):
         super().__init__()
 
-        self.input_dim  = d_feat
+        self.input_dim = d_feat
         self.output_dim = output_dim
         self.freq_dim = freq_dim
         self.hidden_dim = hidden_size
@@ -65,13 +66,13 @@ class SFM_Model(nn.Module):
         self.states = []
 
     def forward(self, input):
-        input = input.reshape(len(input), self.input_dim, -1) # [N, F, T]
+        input = input.reshape(len(input), self.input_dim, -1)  # [N, F, T]
-        input = input.permute(0, 2, 1) # [N, T, F]
+        input = input.permute(0, 2, 1)  # [N, T, F]
         time_step = input.shape[1]
 
         for ts in range(time_step):
-            x = input[:, ts,:]
+            x = input[:, ts, :]
-            if(len(self.states)==0): #hasn't initialized yet
+            if len(self.states) == 0:  # hasn't initialized yet
                 self.init_states(x)
             self.get_constants(x)
             p_tm1 = self.states[0]
@@ -89,8 +90,9 @@ class SFM_Model(nn.Module):
             x_c = torch.matmul(x * B_W[0], self.W_c) + self.b_c
             x_o = torch.matmul(x * B_W[0], self.W_o) + self.b_o
 
-            i = self.inner_activation(x_i + torch.matmul(h_tm1 * B_U[0], self.U_i)) # not sure whether I am doing in the right unsquuze
+            i = self.inner_activation(
-            
+                x_i + torch.matmul(h_tm1 * B_U[0], self.U_i)
+            )  # not sure whether I am doing in the right unsquuze
 
             ste = self.inner_activation(x_ste + torch.matmul(h_tm1 * B_U[0], self.U_ste))
             fre = self.inner_activation(x_fre + torch.matmul(h_tm1 * B_U[0], self.U_fre))
@@ -152,13 +154,14 @@ class SFM_Model(nn.Module):
 
     def get_constants(self, x):
         constants = []
-        constants.append([torch.tensor(1.).to(self.device) for _ in range(6)])
+        constants.append([torch.tensor(1.0).to(self.device) for _ in range(6)])
-        constants.append([torch.tensor(1.).to(self.device) for _ in range(7)])
+        constants.append([torch.tensor(1.0).to(self.device) for _ in range(7)])
-        array = np.array([float(ii)/self.freq_dim for ii in range(self.freq_dim)])
+        array = np.array([float(ii) / self.freq_dim for ii in range(self.freq_dim)])
         constants.append(torch.tensor(array).to(self.device))
 
         self.states[5:] = constants
 
+
 class SFM(Model):
     """SFM Model
 
@@ -185,7 +188,7 @@ class SFM(Model):
         d_feat=6,
         hidden_size=64,
         output_dim=1,
-        freq_dim = 10,
+        freq_dim=10,
         dropout_W=0.0,
         dropout_U=0.0,
         n_epochs=200,
@@ -221,7 +224,7 @@ class SFM(Model):
         self.lr_decay_steps = lr_decay_steps
         self.optimizer = optimizer.lower()
         self.loss_type = loss
-        self.device = 'cuda:%d'%(GPU) if torch.cuda.is_available() else 'cpu'
+        self.device = "cuda:%d" % (GPU) if torch.cuda.is_available() else "cpu"
         self.use_gpu = torch.cuda.is_available()
         self.seed = seed
 
@@ -270,13 +273,13 @@ class SFM(Model):
 
         self.sfm_model = SFM_Model(
             d_feat=self.d_feat,
-            output_dim = self.output_dim,
+            output_dim=self.output_dim,
-            hidden_size = self.hidden_size, 
+            hidden_size=self.hidden_size,
-            freq_dim = self.freq_dim, 
+            freq_dim=self.freq_dim,
             dropout_W=self.dropout_W,
-            dropout_U = self.dropout_U, 
+            dropout_U=self.dropout_U,
-            device = self.device
+            device=self.device,
-            )
+        )
         if optimizer.lower() == "adam":
             self.train_optimizer = optim.Adam(self.sfm_model.parameters(), lr=self.lr)
         elif optimizer.lower() == "gd":
@@ -301,14 +304,7 @@ class SFM(Model):
         self._fitted = False
         self.sfm_model.to(self.device)
 
-    def fit(
+    def fit(self, dataset: DatasetH, evals_result=dict(), verbose=True, save_path=None, **kwargs):
-        self,
-        dataset: DatasetH,
-        evals_result=dict(),
-        verbose=True,
-        save_path=None,
-        **kwargs
-    ):
 
         df_train, df_valid = dataset.prepare(
             ["train", "valid"], col_set=["feature", "label"], data_key=DataHandlerLP.DK_L
@@ -398,12 +394,12 @@ class SFM(Model):
                 # update learning rate
                 self.scheduler.step(cur_loss_val)
 
-        if self.device != 'cpu':
+        if self.device != "cpu":
             torch.cuda.empty_cache()
 
     def get_loss(self, pred, target, loss_type):
         if loss_type == "mse":
-            sqr_loss = (pred - target)**2
+            sqr_loss = (pred - target) ** 2
             loss = sqr_loss.mean()
             return loss
         elif loss_type == "binary":
@@ -424,7 +420,7 @@ class SFM(Model):
         self.sfm_model.eval()
 
         with torch.no_grad():
-            if self.device != 'cpu':
+            if self.device != "cpu":
                 preds = self.sfm_model(x_test).detach().cpu().numpy()
             else:
                 preds = self.sfm_model(x_test).detach().numpy()
@@ -447,8 +443,10 @@ class SFM(Model):
             self.sfm_model.load_state_dict(torch.load(_model_path))
         self._fitted = True
 
+
 class AverageMeter(object):
     """Computes and stores the average and current value"""
+
     def __init__(self):
         self.reset()
 

requirements.txt

@@ -23,3 +23,4 @@ torch==1.6.0
 tqdm==4.49.0
 yahooquery==2.2.7
 mlflow==1.12.1
+pytorch-tabnet==2.0.1