Improve to_torch/to_numpy converters (#147)

* Enable converting list/tuple back and forth from/to numpy/torch.

* Add fallbacks.

* Fix PEP8

* Update unit tests.

* Type annotation. Robust dtype check.

* List of object are converted individually, as a single tensor otherwise.

* Improve robustness of _to_array_with_correct_type

* Add unit tests.

* Do not catch exception at _to_array_with_correct_type level.

* Use _parse_value

* Fix PEP8

* Fix _parse_value list output type fallback.

* Catch torch exception.

* Do not convert torch tensor during fallback.

* Improve unit tests.

* Add unit tests.

* FIx missing import

* Remove support of numpy arrays of tensors for Batch value parser.

* Forbid numpy arrays of tensors.

* Fix PEP8.

* Fix comment.

* Reduce _parse_value branch number.

* Fix None value.

* Forward error message for debugging purpose.

* Fix _is_scalar.

* More specific try/catch blocks.

* Fix exception chaining.

* Fix PEP8.

* Fix _is_scalar.

* Fix missing corner case.

* Fix PEP8.

* Allow Batch empty key.

* Fix multi-dim array datatype check.

Co-authored-by: Alexis Duburcq <alexis.duburcq@wandercraft.eu>

test/base/test_batch.py

@@ -3,8 +3,9 @@ import copy
 import pickle
 import pytest
 import numpy as np
+from itertools import starmap
 
-from tianshou.data import Batch, to_torch
+from tianshou.data import Batch, to_torch, to_numpy
 
 
 def test_batch():
@@ -28,8 +29,19 @@ def test_batch():
     assert b.a == 3
     with pytest.raises(AssertionError):
         Batch({1: 2})
+    with pytest.raises(TypeError):
+        Batch(a=[np.zeros((2, 3)), np.zeros((3, 3))])
+    with pytest.raises(TypeError):
+        Batch(a=[np.zeros((3, 2)), np.zeros((3, 3))])
+    with pytest.raises(TypeError):
+        Batch(a=[torch.zeros((2, 3)), torch.zeros((3, 3))])
+    with pytest.raises(TypeError):
+        Batch(a=[torch.zeros((3, 3)), np.zeros((3, 3))])
+    with pytest.raises(TypeError):
+        Batch(a=[1, np.zeros((3, 3)), torch.zeros((3, 3))])
     batch = Batch(a=[torch.ones(3), torch.ones(3)])
     assert torch.allclose(batch.a, torch.ones(2, 3))
+    Batch(a=[])
     batch = Batch(obs=[0], np=np.zeros([3, 4]))
     assert batch.obs == batch["obs"]
     batch.obs = [1]
@@ -307,7 +319,7 @@ def test_batch_over_batch_to_torch():
     assert batch.b.d.dtype == torch.float32
 
 
-def test_utils_to_torch():
+def test_utils_to_torch_numpy():
     batch = Batch(
         a=np.float64(1.0),
         b=Batch(
@@ -323,8 +335,37 @@ def test_utils_to_torch():
     assert batch_torch_float.a.dtype == torch.float32
     assert batch_torch_float.b.c.dtype == torch.float32
     assert batch_torch_float.b.d.dtype == torch.float32
-    array_list = [float('nan'), 1.0]
+    data_list = [float('nan'), 1]
-    assert to_torch(array_list).dtype == torch.float64
+    data_list_torch = to_torch(data_list)
+    assert data_list_torch.dtype == torch.float64
+    data_list_2 = [np.random.rand(3, 3), np.random.rand(3, 3)]
+    data_list_2_torch = to_torch(data_list_2)
+    assert data_list_2_torch.shape == (2, 3, 3)
+    assert np.allclose(to_numpy(to_torch(data_list_2)), data_list_2)
+    data_list_3 = [np.zeros((3, 2)), np.zeros((3, 3))]
+    data_list_3_torch = to_torch(data_list_3)
+    assert isinstance(data_list_3_torch, list)
+    assert all(isinstance(e, torch.Tensor) for e in data_list_3_torch)
+    assert all(starmap(np.allclose,
+               zip(to_numpy(to_torch(data_list_3)), data_list_3)))
+    data_list_4 = [np.zeros((2, 3)), np.zeros((3, 3))]
+    data_list_4_torch = to_torch(data_list_4)
+    assert isinstance(data_list_4_torch, list)
+    assert all(isinstance(e, torch.Tensor) for e in data_list_4_torch)
+    assert all(starmap(np.allclose,
+               zip(to_numpy(to_torch(data_list_4)), data_list_4)))
+    data_list_5 = [np.zeros(2), np.zeros((3, 3))]
+    data_list_5_torch = to_torch(data_list_5)
+    assert isinstance(data_list_5_torch, list)
+    assert all(isinstance(e, torch.Tensor) for e in data_list_5_torch)
+    data_array = np.random.rand(3, 2, 2)
+    data_empty_tensor = to_torch(data_array[[]])
+    assert isinstance(data_empty_tensor, torch.Tensor)
+    assert data_empty_tensor.shape == (0, 2, 2)
+    data_empty_array = to_numpy(data_empty_tensor)
+    assert isinstance(data_empty_array, np.ndarray)
+    assert data_empty_array.shape == (0, 2, 2)
+    assert np.allclose(to_numpy(to_torch(data_array)), data_array)
 
 
 def test_batch_pickle():
@@ -432,7 +473,7 @@ if __name__ == '__main__':
     test_batch()
     test_batch_over_batch()
     test_batch_over_batch_to_torch()
-    test_utils_to_torch()
+    test_utils_to_torch_numpy()
     test_batch_pickle()
     test_batch_from_to_numpy_without_copy()
     test_batch_standard_compatibility()

tianshou/data/batch.py

@@ -4,6 +4,7 @@ import warnings
 import numpy as np
 from copy import deepcopy
 from numbers import Number
+from collections.abc import Collection
 from typing import Any, List, Tuple, Union, Iterator, Optional
 
 # Disable pickle warning related to torch, since it has been removed
@@ -36,8 +37,11 @@ def _is_scalar(value: Any) -> bool:
     # 3. python object rather than dict / Batch / tensor
     # the check of dict / Batch is omitted because this only checks a value.
     # a dict / Batch will eventually check their values
-    value = np.asanyarray(value)
+    if isinstance(value, torch.Tensor):
-    return value.size == 1 and not value.shape
+        return value.numel() == 1 and not value.shape
+    else:
+        value = np.asanyarray(value)
+        return value.size == 1 and not value.shape
 
 
 def _is_number(value: Any) -> bool:
@@ -53,16 +57,21 @@ def _is_number(value: Any) -> bool:
 def _to_array_with_correct_type(v: Any) -> np.ndarray:
     # convert the value to np.ndarray
     # convert to np.object data type if neither bool nor number
+    # raises an exception if array's elements are tensors themself
     v = np.asanyarray(v)
     if not issubclass(v.dtype.type, (np.bool_, np.number)):
         v = v.astype(np.object)
-    if v.dtype == np.object and not v.shape:
+    if v.dtype == np.object:
         # scalar ndarray with np.object data type is very annoying
         # a=np.array([np.array({}, dtype=object), np.array({}, dtype=object)])
         # a is not array([{}, {}], dtype=object), and a[0]={} results in
         # something very strange:
         # array([{}, array({}, dtype=object)], dtype=object)
-        v = v.item(0)
+        if not v.shape:
+            v = v.item(0)
+        elif any(isinstance(e, (np.ndarray, torch.Tensor))
+                 for e in v.reshape(-1)):
+            raise ValueError("Numpy arrays of tensors are not supported yet.")
     return v
 
 
@@ -113,25 +122,29 @@ def _assert_type_keys(keys):
 
 
 def _parse_value(v: Any):
-    if isinstance(v, (list, tuple, np.ndarray)):
+    if isinstance(v, dict):
-        if not isinstance(v, np.ndarray) and \
-                all(isinstance(e, torch.Tensor) for e in v):
-            v = torch.stack(v)
-            return v
-        v_ = _to_array_with_correct_type(v)
-        if v_.dtype == np.object and _is_batch_set(v):
-            v = Batch(v)  # list of dict / Batch
-        else:
-            # normal data list (main case)
-            # or actually a data list with objects
-            v = v_
-    elif isinstance(v, dict):
         v = Batch(v)
     elif isinstance(v, (Batch, torch.Tensor)):
         pass
     else:
-        # scalar case, convert to ndarray
+        if not isinstance(v, np.ndarray) and isinstance(v, Collection) and \
-        v = _to_array_with_correct_type(v)
+                len(v) > 0 and all(isinstance(e, torch.Tensor) for e in v):
+            try:
+                return torch.stack(v)
+            except RuntimeError as e:
+                raise TypeError("Batch does not support non-stackable iterable"
+                                " of torch.Tensor as unique value yet.") from e
+        try:
+            v_ = _to_array_with_correct_type(v)
+        except ValueError as e:
+            raise TypeError("Batch does not support heterogeneous list/tuple"
+                            " of tensors as unique value yet.") from e
+        if _is_batch_set(v):
+            v = Batch(v)  # list of dict / Batch
+        else:
+            # None, scalar, normal data list (main case)
+            # or an actual list of objects
+            v = v_
     return v
 
 

tianshou/data/utils.py

@@ -3,12 +3,12 @@ import numpy as np
 from numbers import Number
 from typing import Union, Optional
 
-from tianshou.data import Batch
+from tianshou.data.batch import _parse_value, Batch
 
 
 def to_numpy(x: Union[
-    torch.Tensor, dict, Batch, np.ndarray]) -> Union[
+    Batch, dict, list, tuple, np.ndarray, torch.Tensor]) -> Union[
-        dict, Batch, np.ndarray]:
+        Batch, dict, list, tuple, np.ndarray, torch.Tensor]:
     """Return an object without torch.Tensor."""
     if isinstance(x, torch.Tensor):
         x = x.detach().cpu().numpy()
@@ -17,13 +17,20 @@ def to_numpy(x: Union[
             x[k] = to_numpy(v)
     elif isinstance(x, Batch):
         x.to_numpy()
+    elif isinstance(x, (list, tuple)):
+        try:
+            x = to_numpy(_parse_value(x))
+        except TypeError:
+            x = [to_numpy(e) for e in x]
+    else:  # fallback
+        x = np.asanyarray(x)
     return x
 
 
-def to_torch(x: Union[torch.Tensor, dict, Batch, np.ndarray],
+def to_torch(x: Union[Batch, dict, list, tuple, np.ndarray, torch.Tensor],
              dtype: Optional[torch.dtype] = None,
              device: Union[str, int, torch.device] = 'cpu'
-             ) -> Union[dict, Batch, torch.Tensor]:
+             ) -> Union[Batch, dict, list, tuple, np.ndarray, torch.Tensor]:
     """Return an object without np.ndarray."""
     if isinstance(x, torch.Tensor):
         if dtype is not None:
@@ -36,14 +43,19 @@ def to_torch(x: Union[torch.Tensor, dict, Batch, np.ndarray],
         x.to_torch(dtype, device)
     elif isinstance(x, (np.number, np.bool_, Number)):
         x = to_torch(np.asanyarray(x), dtype, device)
-    elif isinstance(x, list) and len(x) > 0 and \
+    elif isinstance(x, (list, tuple)):
-            all(isinstance(e, (np.number, np.bool_, Number)) for e in x):
+        try:
-        x = to_torch(np.asanyarray(x), dtype, device)
+            x = to_torch(_parse_value(x), dtype, device)
-    elif isinstance(x, np.ndarray) and \
+        except TypeError:
-            isinstance(x.item(0), (np.number, np.bool_, Number)):
+            x = [to_torch(e, dtype, device) for e in x]
-        x = torch.from_numpy(x).to(device)
+    else:  # fallback
-        if dtype is not None:
+        x = np.asanyarray(x)
-            x = x.type(dtype)
+        if issubclass(x.dtype.type, (np.bool_, np.number)):
+            x = torch.from_numpy(x).to(device)
+            if dtype is not None:
+                x = x.type(dtype)
+        else:
+            raise TypeError(f"object {x} cannot be converted to torch.")
     return x