Github action: auto-update.

Author: github-actions[bot]

dev/_downloads/07fcc19ba03226cd3d83d4e40ec44385/auto_examples_python.zip

@@ -96,14 +96,14 @@
 
 # %%
 # Initialize the convolution and set the weights to something resembling an edge filter/finit differences
-conv = DiscreteContinuousConv2d(1, 1, grid_in=grid_in, grid_out=grid_out, quad_weights=q_in, kernel_shape=[2,4], radius_cutoff=5/nyo, periodic=False).float()
+conv = DiscreteContinuousConv2d(1, 1, grid_in=grid_in, grid_out=grid_out, quadrature_weights=q_in, kernel_shape=[2,4], radius_cutoff=5/nyo, periodic=False).float()
 
 # initialize a kernel resembling an edge filter
 w = torch.zeros_like(conv.weight)
 w[0,0,1] = 1.0
 w[0,0,3] = -1.0
 conv.weight = nn.Parameter(w)
-psi = conv.get_psi()
+psi = conv.get_local_filter_matrix()
 
 # %% apply the DISCO convolution to the data and plot it
 # in order to compute the convolved image, we need to first bring it into the right shape with `batch_size x n_channels x n_grid_points`
@@ -145,7 +145,7 @@
 # %%
 
 plt.figure(figsize=(4,6), )
-plt.imshow(conv_equi.get_psi()[0].detach(), cmap=cmap)
+plt.imshow(conv_equi.get_local_filter_matrix()[0].detach(), cmap=cmap)
 plt.colorbar()
 
 # # %%
@@ -157,7 +157,7 @@
 # plt.show()
 
 # %% test the transpose convolution
-convt = DiscreteContinuousConvTranspose2d(1, 1, grid_in=grid_out, grid_out=grid_in, quad_weights=q_out, kernel_shape=[2,4], radius_cutoff=3/nyo, periodic=False).float()
+convt = DiscreteContinuousConvTranspose2d(1, 1, grid_in=grid_out, grid_out=grid_in, quadrature_weights=q_out, kernel_shape=[2,4], radius_cutoff=3/nyo, periodic=False).float()
 
 # initialize a flat
 w = torch.zeros_like(conv.weight)

dev/_downloads/0d78e075dd52a34e158d7f5f710dfe89/plot_incremental_FNO_darcy.zip

@@ -130,7 +130,7 @@
       },
       "outputs": [],
       "source": [
-        "conv = DiscreteContinuousConv2d(1, 1, grid_in=grid_in, grid_out=grid_out, quad_weights=q_in, kernel_shape=[2,4], radius_cutoff=5/nyo, periodic=False).float()\n\n# initialize a kernel resembling an edge filter\nw = torch.zeros_like(conv.weight)\nw[0,0,1] = 1.0\nw[0,0,3] = -1.0\nconv.weight = nn.Parameter(w)\npsi = conv.get_psi()"
+        "conv = DiscreteContinuousConv2d(1, 1, grid_in=grid_in, grid_out=grid_out, quadrature_weights=q_in, kernel_shape=[2,4], radius_cutoff=5/nyo, periodic=False).float()\n\n# initialize a kernel resembling an edge filter\nw = torch.zeros_like(conv.weight)\nw[0,0,1] = 1.0\nw[0,0,3] = -1.0\nconv.weight = nn.Parameter(w)\npsi = conv.get_local_filter_matrix()"
       ]
     },
     {
@@ -170,7 +170,7 @@
       },
       "outputs": [],
       "source": [
-        "plt.figure(figsize=(4,6), )\nplt.imshow(conv_equi.get_psi()[0].detach(), cmap=cmap)\nplt.colorbar()\n\n# # %%\n\n# print(\"plt the error:\")\n# plt.figure(figsize=(4,6), )\n# plt.imshow(out1 - out2, cmap=cmap)\n# plt.colorbar()\n# plt.show()"
+        "plt.figure(figsize=(4,6), )\nplt.imshow(conv_equi.get_local_filter_matrix()[0].detach(), cmap=cmap)\nplt.colorbar()\n\n# # %%\n\n# print(\"plt the error:\")\n# plt.figure(figsize=(4,6), )\n# plt.imshow(out1 - out2, cmap=cmap)\n# plt.colorbar()\n# plt.show()"
       ]
     },
     {
@@ -181,7 +181,7 @@
       },
       "outputs": [],
       "source": [
-        "convt = DiscreteContinuousConvTranspose2d(1, 1, grid_in=grid_out, grid_out=grid_in, quad_weights=q_out, kernel_shape=[2,4], radius_cutoff=3/nyo, periodic=False).float()\n\n# initialize a flat\nw = torch.zeros_like(conv.weight)\nw[0,0,0] = 1.0\nw[0,0,1] = 1.0\nw[0,0,2] = 1.0\nw[0,0,3] = 1.0\nconvt.weight = nn.Parameter(w)\n\ndata = nn.functional.interpolate(torch.from_numpy(img).unsqueeze(0).unsqueeze(0), size=(ny,nx)).squeeze().float().permute(1,0).flip(1).reshape(-1)\nout = convt(data.reshape(1, 1, -1))\n\nprint(out.shape)\n\nplt.figure(figsize=(4,6), )\nplt.imshow(torch.flip(out.squeeze().detach().reshape(nx, ny).transpose(0,1), dims=(-2, )), cmap=cmap)\nplt.colorbar()\nplt.show()"
+        "convt = DiscreteContinuousConvTranspose2d(1, 1, grid_in=grid_out, grid_out=grid_in, quadrature_weights=q_out, kernel_shape=[2,4], radius_cutoff=3/nyo, periodic=False).float()\n\n# initialize a flat\nw = torch.zeros_like(conv.weight)\nw[0,0,0] = 1.0\nw[0,0,1] = 1.0\nw[0,0,2] = 1.0\nw[0,0,3] = 1.0\nconvt.weight = nn.Parameter(w)\n\ndata = nn.functional.interpolate(torch.from_numpy(img).unsqueeze(0).unsqueeze(0), size=(ny,nx)).squeeze().float().permute(1,0).flip(1).reshape(-1)\nout = convt(data.reshape(1, 1, -1))\n\nprint(out.shape)\n\nplt.figure(figsize=(4,6), )\nplt.imshow(torch.flip(out.squeeze().detach().reshape(nx, ny).transpose(0,1), dims=(-2, )), cmap=cmap)\nplt.colorbar()\nplt.show()"
       ]
     },
     {