The code can not converge

[shuishen112]

It seems that the code can converge, the loss function is always about 133.13.

[congzlwag]

Based on your description of the loss, probably it has converged to the optimum for those the maximum bond dimension is limited to 2. If the maximum bond dimension is indeed set to 2, increasing that is worth trying.

[shuishen112]

Thanks for your feedback. I uncommented the code: m.cutoff = 1e-7 and It works.

It seems the training time will last several days to achieve the loss result in FIgure 4. I run the script overnight and it has only 15 loops. Am I right?