How to resume DFT calculation

[McAlexDudu]

DFT calculation will get unconverged every now and then, and when this happened, the calculation will stop and we have to restart all over. How to resume? any suggestion?

[YuuuXie]

Currently if you set write_model in otf module to 2 or 3, the GP model will be saved automatically, then you can load gp by

gp_model = GaussianProcess.from_file('filename')

Also, you can grab the positions of atoms where your calculation was interrupted, save it as an np.array of shape (n_atoms, 3), denoted as previous_pos.

Then in the otf setup, use

my_otf = OTF(prev_pos_init = previous_pos,
             gp = gp_model,
             <other arguments>)
my_otf.curr_step = <your last step number>
my_otf.run()

Above is a temporary solution. But we are also considering having a more simple interface for restarting otf, there was an old issue #49 here. We will add it recent days

[McAlexDudu]

Currently if you set write_model in otf module to 2 or 3, the GP model will be saved automatically, then you can load gp by

gp_model = GaussianProcess.from_file('filename')

Also, you can grab the positions of atoms where your calculation was interrupted, save it as an np.array of shape (n_atoms, 3), denoted as previous_pos.

Then in the otf setup, use

my_otf = OTF(prev_pos_init = previous_pos,
             gp = gp_model,
             <other arguments>)
my_otf.curr_step = <your last step number>
my_otf.run()

Above is a temporary solution. But we are also considering having a more simple interface for restarting otf, there was an old issue #49 here. We will add it recent days

Thanks a lot, I will try it.