Support PyTorch 2.3.0

.github/workflows/ci.yaml

@@ -19,15 +19,13 @@ jobs:
         mpi: [ 'openmpi' ]
         install-options: [ '.', '.[hdf5,netcdf]' ]
         pytorch-version:
-          - 'torch==1.11.0+cpu torchvision==0.12.0+cpu torchaudio==0.11.0'
           - 'torch==1.12.1+cpu torchvision==0.13.1+cpu torchaudio==0.12.1'
           - 'torch==1.13.1+cpu torchvision==0.14.1+cpu torchaudio==0.13.1'
           - 'torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2'
           - 'torch==2.1.2 torchvision==0.16.2 torchaudio==2.1.2'
-          - 'torch torchvision torchaudio'
+          - 'torch==2.2.2 torchvision==0.17.2 torchaudio==2.2.2'
+          - 'torch==2.3.0 torchvision==0.18.0 torchaudio==2.3.0'
         exclude:
-          - py-version: '3.11'
-            pytorch-version: 'torch==1.11.0+cpu torchvision==0.12.0+cpu torchaudio==0.11.0'
           - py-version: '3.11'
             pytorch-version: 'torch==1.12.1+cpu torchvision==0.13.1+cpu torchaudio==0.12.1'
           - py-version: '3.11'

heat/core/arithmetics.py

@@ -1954,14 +1954,7 @@ def left_shift(
         elif dtypes[dt] == types.bool:
             arrs[dt] = types.int(arrs[dt])
 
-    try:
-        result = _operations.__binary_op(torch.bitwise_left_shift, t1, t2, out, where)
-    except AttributeError:  # pragma: no cover
-        result = _operations.__binary_op(
-            torch.Tensor.__lshift__, t1, t2, out, where
-        )  # pytorch < 1.10
-
-    return result
+    return _operations.__binary_op(torch.bitwise_left_shift, t1, t2, out, where)
 
 
 def _lshift(self, other):
@@ -2875,14 +2868,7 @@ def right_shift(
         elif dtypes[dt] == types.bool:
             arrs[dt] = types.int(arrs[dt])
 
-    try:
-        result = _operations.__binary_op(torch.bitwise_right_shift, t1, t2, out, where)
-    except AttributeError:  # pragma: no cover
-        result = _operations.__binary_op(
-            torch.Tensor.__rshift__, t1, t2, out, where
-        )  # pytorch < 1.10
-
-    return result
+    return _operations.__binary_op(torch.bitwise_right_shift, t1, t2, out, where)
 
 
 def _rshift(self, other):

heat/core/io.py

@@ -134,7 +134,7 @@ def load_hdf5(
             data = handle[dataset]
             gshape = data.shape
             if split is not None:
-                gshape = np.array(gshape)
+                gshape = list(gshape)
                 gshape[split] = int(gshape[split] * load_fraction)
                 gshape = tuple(gshape)
             dims = len(gshape)

heat/core/stride_tricks.py

@@ -44,19 +44,6 @@ def broadcast_shape(shape_a: Tuple[int, ...], shape_b: Tuple[int, ...]) -> Tuple
     """
     try:
         resulting_shape = torch.broadcast_shapes(shape_a, shape_b)
-    except AttributeError:  # torch < 1.8
-        it = itertools.zip_longest(shape_a[::-1], shape_b[::-1], fillvalue=1)
-        resulting_shape = max(len(shape_a), len(shape_b)) * [None]
-        for i, (a, b) in enumerate(it):
-            if a == 0 and b == 1 or b == 0 and a == 1:
-                resulting_shape[i] = 0
-            elif a == 1 or b == 1 or a == b:
-                resulting_shape[i] = max(a, b)
-            else:
-                raise ValueError(
-                    f"operands could not be broadcast, input shapes {shape_a} {shape_b}"
-                )
-        return tuple(resulting_shape[::-1])
     except TypeError:
         raise TypeError(f"operand 1 must be tuple of ints, not {type(shape_a)}")
     except NameError:

heat/fft/fft.py

@@ -858,11 +858,6 @@ def ihfft2(
     -----
     This function requires MPI communication if the input array is distributed and the split axis is transformed.
     """
-    torch_has_ihfftn = hasattr(torch.fft, "ihfftn")
-    if not torch_has_ihfftn:  # pragma: no cover
-        raise NotImplementedError(
-            f"n-dim inverse Hermitian FFTs not implemented for torch < 1.11.0. Your environment runs torch {torch.__version__}. Please upgrade torch."
-        )
     return __real_fftn_op(x, torch.fft.ihfft2, s=s, axes=axes, norm=norm)
 
 
@@ -896,11 +891,6 @@ def ihfftn(
     -----
     This function requires MPI communication if the input array is distributed and the split axis is transformed.
     """
-    torch_has_ihfftn = hasattr(torch.fft, "ihfftn")
-    if not torch_has_ihfftn:  # pragma: no cover
-        raise NotImplementedError(
-            f"n-dim inverse Hermitian FFTs not implemented for torch < 1.11.0. Your environment runs torch {torch.__version__}. Please upgrade torch."
-        )
     return __real_fftn_op(x, torch.fft.ihfftn, s=s, axes=axes, norm=norm)
 
 

setup.py

@@ -35,7 +35,7 @@
     install_requires=[
         "mpi4py>=3.0.0",
         "numpy>=1.22.0",
-        "torch>=1.11.0, <2.2.3",
+        "torch>=1.12.0, <2.3.1",
         "scipy>=1.10.0",
         "pillow>=6.0.0",
         "torchvision>=0.12.0",