Fix call

Author: EmilyBourne

pygyro/model/layout.py

@@ -4,7 +4,7 @@
 import operator
 import time
 #from .accelerated_layout import flat_transpose
-from hptt import tensorTransposeAndUpdate
+import hptt
 import torch
 
 from abc import ABC
@@ -13,50 +13,7 @@ def my_transpose(dest, source, axes):
     if axes == list(range(len(axes))):
         dest[:] = source
     else:
-        #print(axes, source.shape, dest.shape, source.size, dest.size)
-        #assert tuple(source.shape[a] for a in axes) == dest.shape
-        #print(source.flags)
-        #print(dest.flags)
-        #assert source.flags['C_CONTIGUOUS'] or source.flags['F_CONTIGUOUS']
-        #assert dest.flags['C_CONTIGUOUS'] or dest.flags['F_CONTIGUOUS']
-        tensorTransposeAndUpdate(tuple(axes), 1.0, source, 1.0, dest)
-    #s = time.time()
-    #dest[:] = source.transpose(axes)
-    #print("NumpyTranspose : ", time.time() - s)
-
-    #s = time.time()
-    #new_shape = [source.shape[0]]
-    #idx = 0
-    #new_axes = [axes[0]]
-    #for i,a in enumerate(axes[1:],1):
-    #    if a == axes[i-1]+1:
-    #        new_shape[-1] *= source.shape[i]
-    #    else:
-    #        new_shape.append(source.shape[i])
-    #        new_axes.append(a)
-    #if len(new_shape) == 1:
-    #    dest[:] = source
-    #else:
-    #    new_axes = np.argsort(np.argsort(new_axes))
-    #    dim1, dim2 = next((i,a) for i,a in enumerate(new_axes) if i!=a)
-    #    print(new_shape, axes, new_axes)
-    #    #tensorTransposeAndUpdate(tuple(axes), 1.0, source, 1.0, dest)
-    #    #flat_transpose(dest.ravel(), source.ravel(), source.shape, dest.shape, tuple(axes))
-    #    #dest[:] = np.transpose(source, axes)
-    #    print(source.shape)
-    #    print(dest.shape)
-    #    dest[:] = torch.transpose(torch.from_numpy(source), int(dim1), int(dim2)).contiguous()
-    #print("Torch : ", time.time() - s)
-    #s = time.time()
-    #dest[:] = np.ascontiguousarray(source.transpose(axes))
-    #print("NumpyTransposeContig : ", time.time() - s)
-    ##print(source.shape, source.strides)
-    ##print(dest.shape, dest.strides)
-    ##print(source.flags)
-    ##print(dest.flags)
-    #s = time.time()
-    #tensorTransposeAndUpdate(tuple(axes), 1.0, source, 1.0, dest)
-    #print("HPTT Transpose : ", time.time() - s)
+        hptt.tensorTransposeAndUpdate(axes, 1.0, source, 0.0, dest)
 
 
 class Layout: