convert numpy to tensor

fastNLP/core/batch.py

@@ -19,6 +19,9 @@ from collections import defaultdict
 
 from .dataset import DataSet
 from .sampler import SequentialSampler
+from .field import _get_ele_type_and_dim
+from ._logger import logger
+
 
 _python_is_exit = False
 
@@ -31,6 +34,21 @@ def _set_python_is_exit():
 atexit.register(_set_python_is_exit)
 
 
+def may_to_tensor(data, as_numpy, fn):
+    if not as_numpy:
+        dtype, dim = _get_ele_type_and_dim(data)
+        try:
+            data, flag = _to_tensor(data, dtype)
+        except TypeError as e:
+            logger.error(f"Field {fn} cannot be converted to torch.tensor.")
+            raise e
+    return data
+
+
+def convert_tensor(batch_dict, as_numpy):
+    for n, v in batch_dict.items():
+        batch_dict[n] = may_to_tensor(v, as_numpy, n)
+
 class DataSetGetter:
     """
     传递给torch.utils.data.DataLoader获取数据，DataLoder会传入int的idx获取数据(调用这里的__getitem__()函数)。
@@ -80,6 +98,8 @@ class DataSetGetter:
 
         sin_x = pad(sin_x)
         sin_y = pad(sin_y)
+        convert_tensor(sin_x, self.as_numpy)
+        convert_tensor(sin_y, self.as_numpy)
 
         if not self.dataset.collector.is_empty():
             bx, by = self.dataset._collect_batch(ins_list)

fastNLP/core/collect_fn.py

@@ -95,7 +95,7 @@ class Collector:
     def copy_from(self, col):
         assert isinstance(col, Collector)
         new_col = Collector()
-        new_col.collect_fns = deepcopy(col)
+        new_col.collect_fns = deepcopy(col.collect_fns)
         return new_col