performance mprovement (microsoft#921)

* performance mprovement

* memory refine

qlib/contrib/model/pytorch_nn.py

@@ -7,6 +7,7 @@
 from collections import defaultdict
 
 import os
+import gc
 import numpy as np
 import pandas as pd
 from typing import Callable, Optional, Text, Union
@@ -32,7 +33,6 @@
 from ...workflow import R
 from qlib.contrib.meta.data_selection.utils import ICLoss
 from torch.nn import DataParallel
-from torch.utils.data import DataLoader, SequentialSampler
 
 
 class DNNModelPytorch(Model):
@@ -201,7 +201,7 @@ def fit(
                     seg, col_set=["feature", "label"], data_key=self.valid_key if seg == "valid" else DataHandlerLP.DK_L
                 )
                 all_df["x"][seg] = df["feature"]
-                all_df["y"][seg] = df["label"]
+                all_df["y"][seg] = df["label"].copy()  # We have to use copy to remove the reference to release mem
                 if reweighter is None:
                     all_df["w"][seg] = pd.DataFrame(np.ones_like(all_df["y"][seg].values), index=df.index)
                 elif isinstance(reweighter, Reweighter):
@@ -216,6 +216,10 @@ def fit(
                     all_t[v][seg] = all_t[v][seg].to(self.device)  # This will consume a lot of memory !!!!
 
                 evals_result[seg] = []
+                # free memory
+                del df
+                del all_df["x"]
+                gc.collect()
 
         save_path = get_or_create_path(save_path)
         stop_steps = 0
@@ -266,7 +270,7 @@ def fit(
                         loss_val = cur_loss_val.item()
                         metric_val = (
                             self.get_metric(
-                                preds.reshape(-1), all_t["y"]["valid"].reshape(-1), all_df["x"]["valid"].index
+                                preds.reshape(-1), all_t["y"]["valid"].reshape(-1), all_df["y"]["valid"].index
                             )
                             .detach()
                             .cpu()
@@ -281,7 +285,7 @@ def fit(
                                 self.get_metric(
                                     self._nn_predict(all_t["x"]["train"], return_cpu=False),
                                     all_t["y"]["train"].reshape(-1),
-                                    all_df["x"]["train"].index,
+                                    all_df["y"]["train"].index,
                                 )
                                 .detach()
                                 .cpu()
@@ -351,31 +355,17 @@ def _nn_predict(self, data, return_cpu=True):
         1) test inference (data may come from CPU and expect the output data is on CPU)
         2) evaluation on training (data may come from GPU)
         """
-        if isinstance(data, torch.Tensor) and data.device.type != "cpu":
-            # GPU data
-            # CUDA data don't support pin_memory and multi-processing workers
-            num_workers = 0
-            pin_memory = False
-        else:
-            # CPU data
-            if not isinstance(data, torch.Tensor):
-                if isinstance(data, pd.DataFrame):
-                    data = data.values
-            # else: CPU Tensor
-            num_workers = 8
-            pin_memory = True
-        data_loader = DataLoader(
-            data,
-            sampler=SequentialSampler(data),
-            batch_size=self.batch_size,
-            drop_last=False,
-            num_workers=num_workers,
-            pin_memory=pin_memory,
-        )
+        if not isinstance(data, torch.Tensor):
+            if isinstance(data, pd.DataFrame):
+                data = data.values
+            data = torch.Tensor(data)
+        data = data.to(self.device)
         preds = []
         self.dnn_model.eval()
         with torch.no_grad():
-            for x in data_loader:
+            batch_size = 8096
+            for i in range(0, len(data), batch_size):
+                x = data[i : i + batch_size]
                 preds.append(self.dnn_model(x.to(self.device)).detach().reshape(-1))
         if return_cpu:
             preds = np.concatenate([pr.cpu().numpy() for pr in preds])

qlib/contrib/report/data/base.py

@@ -8,14 +8,14 @@
 
 """
 import pandas as pd
-from blocks.utils.log import logt
+from qlib.log import TimeInspector
 from qlib.contrib.report.utils import sub_fig_generator
 
 
 class FeaAnalyser:
     def __init__(self, dataset: pd.DataFrame):
         self._dataset = dataset
-        with logt("calc_stat_values"):
+        with TimeInspector.logt("calc_stat_values"):
             self.calc_stat_values()
 
     def calc_stat_values(self):