refactor(aggregation): Explicitly cast arrays to float64 (#236)

When casting a Tensor to a numpy array, the obtained array is typically typed as float32 while numpy typically uses float64. This makes the cast explicit.

src/torchjd/aggregation/cagrad.py

@@ -76,7 +76,7 @@ def forward(self, matrix: Tensor) -> Tensor:
         U, S, _ = torch.svd(gramian)
 
         reduced_matrix = U @ S.sqrt().diag()
-        reduced_array = reduced_matrix.cpu().detach().numpy()
+        reduced_array = reduced_matrix.cpu().detach().numpy().astype(np.float64)
 
         dimension = matrix.shape[0]
         reduced_g_0 = reduced_array.T @ np.ones(dimension) / dimension

src/torchjd/aggregation/dualproj.py

@@ -107,10 +107,10 @@ def __init__(
 
     def forward(self, matrix: Tensor) -> Tensor:
         weights = self.weighting(matrix)
-        weights_array = weights.cpu().detach().numpy()
+        weights_array = weights.cpu().detach().numpy().astype(np.float64)
 
         gramian = _compute_normalized_gramian(matrix, self.norm_eps)
-        gramian_array = gramian.cpu().detach().numpy()
+        gramian_array = gramian.cpu().detach().numpy().astype(np.float64)
         dimension = gramian.shape[0]
 
         # Because of numerical errors, `gramian_array` might have slightly negative eigenvalue(s),

src/torchjd/aggregation/upgrad.py

@@ -109,7 +109,7 @@ def forward(self, matrix: Tensor) -> Tensor:
 
     def _compute_lagrangian(self, matrix: Tensor, weights: Tensor) -> Tensor:
         gramian = _compute_normalized_gramian(matrix, self.norm_eps)
-        gramian_array = gramian.cpu().detach().numpy()
+        gramian_array = gramian.cpu().detach().numpy().astype(np.float64)
         dimension = gramian.shape[0]
 
         regularization_array = self.reg_eps * np.eye(dimension)