fixing stuff to enable RMSE derivatives (b4871d69) · Commits · KBM / thermo_diff

mod.py

+5 −1

Original line number	Diff line number	Diff line
		@@ -207,7 +207,11 @@ class Mod:
		"""
		for t in self.top:
		self.topname = path
		t.save_mod(self.topname + name + '-' + t.top)
		final_name = self.topname + name + '-' + t.top
		if not os.path.exists(final_name):
		t.save_mod(final_name)
		else:
		print(f'file {final_name} already exists, leaving as it is')

		def goto_mydir(self):
		"""

mod_nbfix.py

+33 −29

Original line number	Diff line number	Diff line
		@@ -23,16 +23,20 @@ class ModNbfix(Mod):
		self.nbe = 'm' if 'm' in changes else ''
		if not (self.nbs or self.nbe):
		raise ValueError("no mode selected appropriate for an ModParam object")
		for t in self.top:
		for t, tnames, tres, tresnames in zip(self.top, self.names, self.res, self.resnames):
		self.LJ_sigmas = []
		self.LJ_epsilons = []
		self.NB_sigmas = None
		self.NB_epsilons = None
		self.types_are_consistent(t)
		self.types_are_consistent(t, tnames, tres, tresnames)
		if 'nonbond_params' not in t.section_headers:
		self.add_nbfix_section(t)
		self.types = (self.get_type(t, self.names[0][0], self.res[0][0], self.resnames[0][0]),
		self.get_type(t, self.names[1][0], self.res[1][0], self.resnames[1][0]))
		try:
		self.types = (self.get_type(t, tnames[0][0], tres[0][0], tresnames[0][0]),
		self.get_type(t, tnames[1][0], tres[1][0], tresnames[1][0]))
		except IndexError:
		pass
		else:
		self.check_lj_params(t)
		if self.types[0] == self.types[1]:
		self.prefixes = ['Y', 'Y']
		@@ -40,11 +44,11 @@ class ModNbfix(Mod):
		self.prefixes = ['Q', 'Y']
		# we mod the first type
		self.clone_type(t, atomtype=self.types[0], prefix=self.prefixes[0])
		self.prefix_type(t, 0)
		self.prefix_type(t, tnames, tres, tresnames, 0)
		# and then second, if different than first
		if self.types[0] != self.types[1]:
		self.clone_type(t, atomtype=self.types[1], prefix=self.prefixes[1])
		self.prefix_type(t, 1)
		self.prefix_type(t, tnames, tres, tresnames, 1)
		self.sort_dihedrals(t)
		self.mod_nb_sigma_eps(t)

		@@ -100,16 +104,16 @@ class ModNbfix(Mod):
		returnlist.append(fstring.format(*mod_list))
		return returnlist

		def prefix_type(self, top, prefix_number):
		def prefix_type(self, top, tnames, tres, tresnames, prefix_number):
		"""
		looks for a line in topology section 'atoms'
		and modifies it as needed
		:return: None
		"""
		for x in range(len(self.names[prefix_number])):
		name = self.names[prefix_number][x]
		resname = self.resnames[prefix_number][x]
		resnum = self.res[prefix_number][x]
		for x in range(len(tnames[prefix_number])):
		name = tnames[prefix_number][x]
		resname = tresnames[prefix_number][x]
		resnum = tres[prefix_number][x]
		for section in top.list_sections('atoms'):
		for line in range(len(top.sections[section])):
		lspl = top.sections[section][line].split()
		@@ -141,18 +145,18 @@ class ModNbfix(Mod):
		raise RuntimeError("You need to provide mass for the atom to be changed")
		return fstring.format(*cont_list[:8], prefix + cont_list[1], cont_list[6], cont_list[7])

		def types_are_consistent(self, top):
		def types_are_consistent(self, top, tnames, tres, tresnames):
		"""
		if many atoms are passed in a single input line, need to make sure
		they are all of the same type; raises ValueError if this is not the case
		:return: None
		"""
		consistent1 = all([self.get_type(top, self.names[0][0], self.res[0][0], self.resnames[0][0])
		== self.get_type(top, self.names[0][n], self.res[0][n], self.resnames[0][n])
		for n in range(len(self.names[0]))])
		consistent2 = all([self.get_type(top, self.names[1][0], self.res[1][0], self.resnames[1][0])
		== self.get_type(top, self.names[1][n], self.res[1][n], self.resnames[1][n])
		for n in range(len(self.names[1]))])
		consistent1 = all([self.get_type(top, tnames[0][0], tres[0][0], tresnames[0][0])
		== self.get_type(top, tnames[0][n], tres[0][n], tresnames[0][n])
		for n in range(len(tnames[0]))])
		consistent2 = all([self.get_type(top, tnames[1][0], tres[1][0], tresnames[1][0])
		== self.get_type(top, tnames[1][n], tres[1][n], tresnames[1][n])
		for n in range(len(tnames[1]))])
		if not consistent1 or not consistent2:
		raise ValueError("atoms within a single nbfix modification need to have consistent types")

processing.py

+20 −15

Original line number	Diff line number	Diff line
		@@ -100,6 +100,9 @@ def run_gromacs(args):
		'-dhdl {fname}/{tn}-dhdl >> rerun.log 2>&1'.format(gmx=gmx, fname=folder_name,
		traj=traj_path, tn=traj_alias,
		tpr=tpr), shell=True)
		else:
		print("in mod {fname} skipping trajectory {tn}, already calculated".format(fname=folder_name,
		tn=traj_alias))


		def get_pot_code(err_msg):
		@@ -300,33 +303,35 @@ def parse_input_line(line, top, td):
		delim = ';'
		for mod in mods:
		if c == 0:
		t1 = line_list[2]
		t2 = line_list[3]
		ll1 = top.find_type(t1)
		ll2 = top.find_type(t2)
		n = [tuple(x.split('-')[2] for x in ll1), tuple(x.split('-')[2] for x in ll2)]
		rn = [tuple(int(x.split('-')[1]) for x in ll1), tuple(int(x.split('-')[1]) for x in ll2)]
		r = [tuple(x.split('-')[0] for x in ll1), tuple(x.split('-')[0] for x in ll2)]
		t1 = line_list[1]
		t2 = line_list[2]
		n, rn, r = [], [], []
		for t in top:
		ll1 = t.find_type(t1)
		ll2 = t.find_type(t2)
		n.append([tuple(x.split('-')[2] for x in ll1), tuple(x.split('-')[2] for x in ll2)])
		rn.append([tuple(int(x.split('-')[1]) for x in ll1), tuple(int(x.split('-')[1]) for x in ll2)])
		r.append([tuple(x.split('-')[0] for x in ll1), tuple(x.split('-')[0] for x in ll2)])
		elif c == 1:
		assert line.count(delim) == 1
		assert line.count(delim) == 1 and len(top) == 1 # TODO not clear how to implement for more tops, think
		line_first, line_second = line.split(delim)
		line_first, line_second = line_first.split(), line_second.split()
		ll1 = line_first[2:]
		ll2 = line_second[:]
		num = int(line_list[1].split('-')[1])
		res = line_list[1].split('-')[0]
		n = [tuple(ll1), tuple(ll2)]
		rn = [tuple(num for _ in ll1), tuple(num for _ in ll2)]
		r = [tuple(res for _ in ll1), tuple(res for _ in ll2)]
		n = [[tuple(ll1), tuple(ll2)]]
		rn = [[tuple(num for _ in ll1), tuple(num for _ in ll2)]]
		r = [[tuple(res for _ in ll1), tuple(res for _ in ll2)]]
		elif c == 2:
		assert line.count(delim) == 1
		assert line.count(delim) == 1 and len(top) == 1
		line_first, line_second = line.split(delim)
		line_first, line_second = line_first.split(), line_second.split()
		ll1 = line_first[1:]
		ll2 = line_second[:]
		n = [tuple(x.split('-')[2] for x in ll1), tuple(x.split('-')[2] for x in ll2)]
		rn = [tuple(int(x.split('-')[1]) for x in ll1), tuple(x.split('-')[1] for x in ll2)]
		r = [tuple(x.split('-')[0] for x in ll1), tuple(x.split('-')[0] for x in ll2)]
		n = [[tuple(x.split('-')[2] for x in ll1), tuple(x.split('-')[2] for x in ll2)]]
		rn = [[tuple(int(x.split('-')[1]) for x in ll1), tuple(x.split('-')[1] for x in ll2)]]
		r = [[tuple(x.split('-')[0] for x in ll1), tuple(x.split('-')[0] for x in ll2)]]
		else:
		raise ValueError("input line {} cannot be processed".format(' '.join(line_list)))
		new_mods.append(ModNbfix(top=deepcopy(top), master=td, resnr=rn, names=n, resnames=r, changes=mod))

quantity.py

+8 −3

Original line number	Diff line number	Diff line
		@@ -158,9 +158,14 @@ class Quantity:
		self.discrete_free_energy_derivative = np.array(self.multi_discrete_free_derivative())

		def calc_discrete_derivs(self):
		mean_der = [np.sum(np.concatenate([traj.derivatives * (traj.rc == state) for traj in self.trajs])) /
		np.sum(np.concatenate([traj.rc == state for traj in self.trajs])) for state in self.discrete_states]
		mean_der = [np.sum(np.concatenate([traj.derivatives * traj.weights * (traj.rc == state)
		for traj in self.trajs])) /
		np.sum(np.concatenate([traj.weights * (traj.rc == state)
		for traj in self.trajs])) for state in self.discrete_states]
		if len(self.discrete_states) > 1:
		rel_derivatives = [mean_der[0] - der for der in mean_der[1:]]
		else:
		rel_derivatives = mean_der[0]
		return np.array([[deriv, 0] for deriv in rel_derivatives]).ravel() # TODO implement stderr

		def calc_ens_avg_profile(self, attr_name, second_attr=None):

td.py

+1 −1

Original line number	Diff line number	Diff line
		@@ -23,7 +23,7 @@ class ThermoDiff:
		"""
		def __init__(self, options):
		self.wdir = os.getcwd()
		self.topol = [self.abs_path(x) for x in options.topol.split()]
		self.topol = [self.abs_path(x) for x in options.topol.split()] # TODO should crash if topols have identical names
		self.struct = [self.abs_path(x) for x in options.struct.split()]
		self.list_mods = self.abs_path(options.list_mods)
		# need a bit of workaround in case trajs_list file is in a different directory than the actual trajectories