import traceback
import sampling_functions
from __run_tools import get_logger
logger = get_logger(__name__)
'''
This file provides an interface for using sampling functions.
The sampling scheme should be programmed in such a function by
the user and saved in sampling_functions.py. These are called
through sampling_function defined in get_one, which handles
the arguments and provides the routines that should be replicated
for all sampling functions, currently this is simply converting
them to a runnable form, ie trajectory objects.
'''
def get_one(name_func, **sfkwargs):
name_backup_func = 'random_sampling_trajectories'
_sampling_function = getattr(sampling_functions, name_func)
logger.info("Retrieved sampling function: {}".format(_sampling_function))
backup_sampling_function = getattr(sampling_functions, name_backup_func)
logger.info(
"Backup sampling function: {}".format(backup_sampling_function))
assert callable(_sampling_function)
assert callable(backup_sampling_function)
# Use Sampled Frames to make New Trajectories
def sampling_function(project, engine, length, number, *args, **skwargs):
- use something like the np.choice shown in
xplor_microstates. here the weights were given
by a vector with probability for each state i.
5. Returns these frames
- these frames are converted to trajectories for execution
by the sampling interface component, no need to do
'''
from __future__ import print_function
import numpy as np
from __run_tools import get_logger
LOGGER = get_logger(__name__)
oty = lambda d: d['input']['modeller'].outtype
stride = lambda d: d['input']['modeller'].engine.types[oty(d)].stride
dfti = lambda d, fi: fi/stride(d)
def long_trajectories(project, number=1, trajectories=None, uselast=True, **kwargs):
'''
Continually grow the same trajectories to longer total
length.
'''
#Note -- This currently does not make TrajectoryExtensionTasks
# which would be the expected output...
#TODO -- have this function return or indicate to interface