cp_als giving many errors about data not being F-ordered

[tgkolda]

Here is the code that is giving errors with the current development version:

from scipy.io import loadmat
import requests
from io import BytesIO

# Download the MATLAB file from the URL
url = "https://gitlab.com/tensors/tensor_data_eem/-/raw/master/EEM18.mat?ref_type=heads&inline=false"
response = requests.get(url)
mat_data = loadmat(BytesIO(response.content))

# Convert the MATLAB tensor to a PyTTB tensor
X_mat = mat_data['X']
data = X_mat[0,0][0]
X = ttb.tensor(data, data.shape)

num_runs = 5
Xnrm = X.norm()
ranks = np.arange(1, 6)
errors_matrix = np.zeros((len(ranks), num_runs))
seeds = np.zeros((len(ranks), num_runs), dtype=int)
np.random.seed(0)
for i, r in enumerate(ranks):
    for run in range(num_runs):
        seed = np.random.randint(1, 10000000)
        seeds[i, run] = seed
        np.random.seed(seed)
        
        P, _, info = ttb.cp_als(X, r, printitn=0)
        Pfull = P.full()
        Diff = X - Pfull
        Diffnrm = Diff.norm()
        errors_matrix[i, run] = Diffnrm / Xnrm

print("Errors Matrix (Rank x Run):")
print(errors_matrix)
print("Seeds Matrix (Rank x Run):")
print(seeds)

[tgkolda]

It looks like pyttb is depending on the user to make the data F-ordered before handing it over?

[ntjohnson1]

Duplicate of #368 there were a few internal calls that I didn't update correctly. Everything was functionally fine but hit the warning because we were trying to reduce unnecessary copies.