Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

new output class and add configuration output in ".xyz" format #42

Merged
merged 21 commits into from
Sep 13, 2019
Merged

new output class and add configuration output in ".xyz" format #42

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
21 commits
Select commit Hold shift + click to select a range
20374f9
change the file open and close behavior in output file
Sep 12, 2019
8361339
fix bugs of species_labels and reopening files
Sep 12, 2019
005bc81
remove mpirun from the serial QE call
Sep 12, 2019
35fa77f
fix the file close problem
Sep 13, 2019
407df11
change output module into a class
Sep 13, 2019
847049e
change the output interface in otf
Sep 13, 2019
caaa249
change output interface in gp_from_adim
Sep 13, 2019
53fcd87
change output interface in md_run
Sep 13, 2019
fd142cf
add docstring to be filled
Sep 13, 2019
40df9bf
remove a duplicated line
Sep 13, 2019
ba51183
fix a docstring
Sep 13, 2019
0a67a68
add interface to open new log
Sep 13, 2019
83e742c
pep8 formatting
Sep 13, 2019
ccabfbf
update git ignore
Sep 13, 2019
a48290a
remove .out suffix from OTF init arguments in test_OTF; remove genera…
Sep 13, 2019
82ae8ac
add conditions for file opening
Sep 13, 2019
54143ab
redirect the screen printing of hyper-parameter optimization to log s…
Sep 13, 2019
3574b55
fix the interface between gp_algebra and gp.train for the hyper-param…
Sep 13, 2019
27263b2
add .dat to .gitignore
Sep 13, 2019
74c0ba1
Added some typehints to output.py
stevetorr Sep 13, 2019
52b4653
Fixed pep8 error I introduced
stevetorr Sep 13, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 9 additions & 0 deletions .gitignore
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,4 +15,13 @@ pwscf.save/
/tests/ase_otf/kv3/
/tests/ase_otf/grid3*
/tests/test_outputs/*
/tests/*.xyz
/docs/source/_build/html
/tests/*-bak
/tests/*.out
/tests/*dat
/tests/alog
/tests/gp_from_aimd.out
/tests/h2_otf.out-f.xyz-bak
/tests/pwscf.xml

8 changes: 4 additions & 4 deletions flare/gp.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,7 +24,7 @@ def __init__(self, kernel: Callable,
energy_kernel: Callable = None,
opt_algorithm: str = 'L-BFGS-B',
maxiter=10, par=False,
monitor=False):
output=None):
Copy link
Contributor

@stevetorr stevetorr Sep 13, 2019 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Here and in a few other places where output is passed as an argument, we could typehint output as the Output object; in a few locations, we pass in an output location as a string (such as in the TrajectoryTrainer class). We should make sure it's typehinted there as well.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Type hints will be eliminated eventually, as they don't play well with the Sphinx documentation. Might be better to do without them.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

(The type of the input should be made clear in the docstring anyway)

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sad to hear that Sphinx can't handle them. That's a shame. At least docstrings can contain the same info.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Actually it looks like Sphinx only really struggles with np.ndarray type hints. Here's an example from the docs:
image

If we can find a way to get Sphinx to format that better, type hints should be okay.

"""Initialize GP parameters and training data."""

self.kernel = kernel
Expand All @@ -45,7 +45,7 @@ def __init__(self, kernel: Callable,
self.likelihood = None
self.likelihood_gradient = None
self.par = par
self.monitor = monitor
self.output = output

# TODO unit test custom range
def update_db(self, struc: Structure, forces: list,
Expand Down Expand Up @@ -112,7 +112,7 @@ def force_list_to_np(forces: list) -> np.ndarray:

return forces_np

def train(self, monitor=False, custom_bounds=None,
def train(self, output=None, custom_bounds=None,
grad_tol: float = 1e-4,
x_tol: float = 1e-5,
line_steps: int = 20):
Expand All @@ -126,7 +126,7 @@ def train(self, monitor=False, custom_bounds=None,
x_0 = self.hyps

args = (self.training_data, self.training_labels_np,
self.kernel_grad, self.cutoffs, monitor,
self.kernel_grad, self.cutoffs, output,
self.par)

if self.algo == 'L-BFGS-B':
Expand Down
24 changes: 11 additions & 13 deletions flare/gp_algebra.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -261,11 +261,11 @@ def get_like_grad_from_mats(ky_mat, hyp_mat, training_labels_np):

def get_neg_likelihood(hyps: np.ndarray, training_data: list,
training_labels_np: np.ndarray,
kernel, cutoffs=None, monitor: bool = False,
kernel, cutoffs=None, output = None,
par=False):

if monitor:
print('hyps: ' + str(hyps))
if output is not None:
output.write_to_log('hyps: ' + str(hyps), name="hyps")

if par:
ky_mat = \
Expand All @@ -278,20 +278,19 @@ def get_neg_likelihood(hyps: np.ndarray, training_data: list,

like = get_like_from_ky_mat(ky_mat, training_labels_np)

if monitor:
print('like: ' + str(like))
print('\n')
if output is not None:
output.write_to_log('like: ' + str(like)+'\n', name="hyps")
Copy link
Contributor

@stevetorr stevetorr Sep 13, 2019 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

A suggestion which is somewhat to personal preference: I personally find formatting strings as
'like: {} \n'.format(like)
to be more readable.
I know that you also just translated the existing line this way, but going forward we could do all of our string formatting this way.


return -like


def get_neg_like_grad(hyps: np.ndarray, training_data: list,
training_labels_np: np.ndarray,
kernel_grad, cutoffs=None,
monitor: bool = False, par=False):
output = None, par=False):

if monitor:
print('hyps: ' + str(hyps))
if output is not None:
output.write_to_log('hyps: ' + str(hyps)+'\n', name="hyps")

if par:
hyp_mat, ky_mat = \
Expand All @@ -305,9 +304,8 @@ def get_neg_like_grad(hyps: np.ndarray, training_data: list,
like, like_grad = \
get_like_grad_from_mats(ky_mat, hyp_mat, training_labels_np)

if monitor:
print('like grad: ' + str(like_grad))
print('like: ' + str(like))
print('\n')
if output is not None:
output.write_to_log('like grad: ' + str(like_grad)+'\n', name="hyps")
output.write_to_log('like: ' + str(like)+'\n', name="hyps")

return -like, -like_grad
53 changes: 25 additions & 28 deletions flare/gp_from_aimd.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,7 +15,7 @@
import numpy as np
from copy import deepcopy
import pickle
import flare.output as output
from flare.output import Output


class TrajectoryTrainer(object):
Expand All @@ -27,7 +27,7 @@ def __init__(self, frames: List[Structure],
parallel: bool = False,
skip: int = 0,
calculate_energy: bool = False,
output_name: str = 'gp_from_aimd.out',
output_name: str = 'gp_from_aimd',
max_atoms_from_frame: int = np.inf, max_trains: int = np.inf,
min_atoms_added: int = 1,
n_cpus: int = 1, shuffle_frames: bool = False,
Expand Down Expand Up @@ -92,7 +92,7 @@ def __init__(self, frames: List[Structure],
else:
self.pred_func = predict_on_structure

self.output_name = output_name
self.output = Output(output_name)

# set number of cpus for parallelization
self.n_cpus = n_cpus
Expand All @@ -119,16 +119,15 @@ def pre_run(self):
training set, then seed with at least one atom from each
"""

output.write_header(self.gp.cutoffs,
self.gp.kernel_name,
self.gp.hyps,
self.gp.algo,
dt=0,
Nsteps=len(self.frames),
structure=self.frames[0],
std_tolerance=(self.rel_std_tolerance,
self.abs_std_tolerance),
output_name=self.output_name)
self.output.write_header(self.gp.cutoffs,
self.gp.kernel_name,
self.gp.hyps,
self.gp.algo,
dt=0,
Nsteps=len(self.frames),
structure=self.frames[0],
std_tolerance=(self.rel_std_tolerance,
self.abs_std_tolerance))

self.start_time = time.time()

Expand Down Expand Up @@ -166,7 +165,7 @@ def pre_run(self):
if (self.gp.l_mat is None) \
or (self.seed_frames is not None
or self.seed_envs is not None):
self.gp.train(monitor=self.verbose)
self.gp.train(output=self.output if self.verbose > 0 else None)

def run(self):
"""
Expand All @@ -190,9 +189,9 @@ def run(self):
mae = np.mean(np.abs(cur_frame.forces - dft_forces)) * 1000
mac = np.mean(np.abs(dft_forces)) * 1000

output.write_gp_dft_comparison(
self.output.write_gp_dft_comparison(
curr_step=i, frame=cur_frame,
start_time=time.time(), output_name=self.output_name,
start_time=time.time(),
dft_forces=dft_forces,
mae=mae, mac=mac, local_energies=None)

Expand All @@ -207,7 +206,7 @@ def run(self):
if self.train_count < self.max_trains:
self.train_gp()

output.conclude_run(self.output_name)
self.output.conclude_run()

if self.pickle_name:
with open(self.pickle_name, 'wb') as f:
Expand All @@ -225,13 +224,11 @@ def update_gp_and_print(self, frame: Structure, train_atoms: List[int],
:return:
"""

output.write_to_output('\nAdding atom(s) {} to the '
'training set.\n'
.format(train_atoms, ),
self.output_name)
output.write_to_output('Uncertainties: {}.\n'
.format(frame.stds[train_atoms]),
self.output_name)
self.output.write_to_log('\nAdding atom(s) {} to the '
'training set.\n'
.format(train_atoms, ))
self.output.write_to_log('Uncertainties: {}.\n'
.format(frame.stds[train_atoms]))

# update gp model
self.gp.update_db(frame, frame.forces, custom_range=train_atoms)
Expand All @@ -244,11 +241,11 @@ def train_gp(self):
"""
Train the Gaussian process and write the results to the output file.
"""
self.gp.train(monitor=True if self.verbose >= 2 else False)
self.gp.train(output=self.output if self.verbose >= 2 else None)
Copy link
Contributor

@stevetorr stevetorr Sep 13, 2019 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I like the way you implemented this (passing output as the output object or None based on verbosity).


output.write_hyps(self.gp.hyp_labels, self.gp.hyps,
self.start_time, self.output_name,
self.gp.like, self.gp.like_grad)
self.output.write_hyps(self.gp.hyp_labels, self.gp.hyps,
self.start_time,
self.gp.like, self.gp.like_grad)
self.train_count += 1

def is_std_in_bound(self, frame: Structure)->(bool, List[int]):
Expand Down
19 changes: 10 additions & 9 deletions flare/md_run.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,7 @@
import datetime
import concurrent.futures
from flare import md
from flare import output
from flare.output import Output


class MD:
Expand Down Expand Up @@ -42,12 +42,11 @@ def __init__(self, dt: float, number_of_steps: int, gp: GaussianProcess,
self.pes = []
self.kes = []

self.output_name = output_name
self.output = Output(output_name)

def run(self):
output.write_header(self.gp.cutoffs, self.gp.kernel_name, self.gp.hyps,
self.gp.algo, self.dt, self.Nsteps, self.structure,
self.output_name)
self.output.write_header(self.gp.cutoffs, self.gp.kernel_name, self.gp.hyps,
self.gp.algo, self.dt, self.Nsteps, self.structure)
self.start_time = time.time()

while self.curr_step < self.Nsteps:
Expand All @@ -59,7 +58,7 @@ def run(self):
self.update_positions(new_pos)
self.curr_step += 1

output.conclude_run(self.output_name)
self.output.conclude_run()

def predict_on_structure_en(self):
for n in range(self.structure.nat):
Expand Down Expand Up @@ -110,6 +109,8 @@ def update_temperature(self, new_pos):
def record_state(self):
self.pes.append(np.sum(self.local_energies))
self.kes.append(self.KE)
output.write_md_config(self.dt, self.curr_step, self.structure,
self.temperature, self.KE, self.local_energies,
self.start_time, self.output_name)
self.output.write_md_config(self.dt, self.curr_step, self.structure,
self.temperature, self.KE, self.local_energies,
self.start_time)
self.output.write_xyz_config(self.curr_step, self.structure,
self.dft_step)
Loading