Skip to content

Bug fixes and new sampling method #4

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

Open
vwcruzeiro wants to merge 12 commits into
base: master
Choose a base branch
from
Open

Bug fixes and new sampling method #4

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
12 commits
Select commit Hold shift + click to select a range
59058c6
Adds a script that can generate protonated and deprotonated forms of …
Oct 29, 2018
5ac9422
Merge remote-tracking branch 'upstream/master'
Dec 19, 2018
1041eb0
Adds MD sampling to al_ions_cycle.py
Dec 19, 2018
ddf4953
Input file modifications in al_ions_cycle.py
Dec 19, 2018
647084c
New sampling technique: structural sampling
Jun 13, 2019
a2fbf87
Remove activelearning/al_1b-nn.py
Oct 20, 2019
03e33f8
Add structural sampling to al_conf_cycle.py and al_ions_cycle.py
Oct 20, 2019
74be704
Merge remote-tracking branch 'upstream/master'
Oct 20, 2019
e7e494a
Fix a mistake in the expression for the distortions of each normal mo…
Oct 20, 2019
7fa346d
Fix __getCharge__ inside lib/nmstools.py
Oct 20, 2019
ac1a797
Modify the algorithm for the random number generator inside the NMS c…
Oct 20, 2019
bc44c65
Remove notebooks/GDB_charged_molecules.ipynb
Oct 20, 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
17 changes: 11 additions & 6 deletions activelearning/al_conf_cycle.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,6 +22,8 @@

h5stor = root_dir + 'h5files/'# h5store location

strucsfolder = root_dir + 'strucs/'# strucs location

optlfile = root_dir + 'optimized_input_files.dat'

#Comet
Expand Down Expand Up @@ -73,6 +75,10 @@
'sig' : M,
}

strucsparams = {'N': 40, # number of maximum structures to select before QBC from each XYZ file
'sig': M,
}

mdsparams = {'N': 2, # trajectories to run
'T1': 300,
'T2': 1000,
Expand All @@ -91,7 +97,7 @@
'tsfiles': ['/home/jsmith48/scratch/auto_rxn_al/rxns/'],
'nmfile':None, #path to gaussian log file containing the data
'nm':0, #id of normal mode
'perc':0, #Move the molecules initial coordiantes along the mode by this amount. Negative numbers are ok.
'perc':0, #Move the molecules initial coordiantes along the mode by this amount. Negative numbers are ok.
}

dhparams = { 'Nmol': 100,
Expand All @@ -106,9 +112,9 @@
dmrparams = {#'mdselect' : [(400,0),(60,2),(40,3),(5,4)],
'mdselect' : [(10,0), (1,11)],
'N' : 20,
'T' : 400.0, # running temp
'T' : 400.0, # running temp
'L' : 25.0, # box length
'V' : 0.04, # Random init velocities
'V' : 0.04, # Random init velocities
'dt' : 0.5, # MD time step
'Nm' : 140, # Molecules to embed
#'Nm' : 160, # Molecules to embed
Expand Down Expand Up @@ -186,10 +192,11 @@
os.mkdir(root_dir + datdir + str(i+1).zfill(2))

## Run active learning sampling ##
acs = alt.alconformationalsampler(ldtdir, datdir + str(i+1).zfill(2), optlfile, fpatoms, netdict)
acs = alt.alconformationalsampler(ldtdir, datdir + str(i+1).zfill(2), optlfile, strucsfolder, fpatoms, netdict)
#acs.run_sampling_cluster(gcmddict, GPU)
#acs.run_sampling_dimer(dmrparams, GPU)
#acs.run_sampling_nms(nmsparams, GPU)
#acs.run_sampling_strucs(strucsparams, GPU)
#acs.run_sampling_md(mdsparams, perc=0.5, gpus=GPU)
#acs.run_sampling_TS(tsparams, gpus=[2])
acs.run_sampling_dhl(dhparams, gpus=GPU+GPU)
Expand All @@ -199,5 +206,3 @@

## Submit jobs, return and pack data
ast.generateQMdata(hostname, username, swkdir, ldtdir, datdir + str(i+1).zfill(2), h5stor, mae, jtime)


80 changes: 67 additions & 13 deletions activelearning/al_ions_cycle.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,6 +20,8 @@

h5stor = root_dir + 'h5files/'# h5store location

strucsfolder = root_dir + 'strucs/'# strucs locations

optlfile = root_dir + 'optimized_input_files.dat'

#Comet
Expand Down Expand Up @@ -58,10 +60,59 @@
root = '/home/jsmith48/scratch/auto_ion_al'

wkdir = root+'/modelions/ANI-AL-smallions/'
iptfile = root+'/modelions/inputtrain.ipt'
#iptfile = root+'/modelions/inputtrain.ipt'
saefile = root+'/modelions/sae_linfit.dat'
cstfile = root+'/modelions/rHCNOSFCl-5.2R_16-3.8A_a4-8.params'

ipt = alt.anitrainerinputdesigner()

ipt.set_parameter('atomEnergyFile','sae_linfit.dat')
ipt.set_parameter('sflparamsfile','rHCNOSFCl-5.2R_16-3.8A_a4-8.params')
ipt.set_parameter('eta','0.001')
ipt.set_parameter('energy','1')
ipt.set_parameter('force','0')
ipt.set_parameter('fmult','1.0')
ipt.set_parameter('feps','0.001')
ipt.set_parameter('dipole','0')
ipt.set_parameter('charge','0')
ipt.set_parameter('cdweight','2.0')
ipt.set_parameter('tolr','100')
ipt.set_parameter('tbtchsz','2560')
ipt.set_parameter('vbtchsz','2560')
ipt.set_parameter('nkde','2')


# Set network layers
ipt.add_layer('H',{"nodes":320,"activation":9,"type":0,"l2norm":1,"l2valu":5.000e-3})
ipt.add_layer('H',{"nodes":64 ,"activation":9,"type":0,"l2norm":1,"l2valu":1.000e-6})
ipt.add_layer('H',{"nodes":128,"activation":9,"type":0,"l2norm":1,"l2valu":1.000e-6})

ipt.add_layer('C',{"nodes":288,"activation":9,"type":0,"l2norm":1,"l2valu":5.000e-3})
ipt.add_layer('C',{"nodes":96 ,"activation":9,"type":0,"l2norm":1,"l2valu":1.000e-6})
ipt.add_layer('C',{"nodes":128,"activation":9,"type":0,"l2norm":1,"l2valu":1.000e-6})

ipt.add_layer('N',{"nodes":256,"activation":9,"type":0,"l2norm":1,"l2valu":5.000e-3})
ipt.add_layer('N',{"nodes":48 ,"activation":9,"type":0,"l2norm":1,"l2valu":1.000e-6})
ipt.add_layer('N',{"nodes":96 ,"activation":9,"type":0,"l2norm":1,"l2valu":1.000e-6})

ipt.add_layer('O',{"nodes":256,"activation":9,"type":0,"l2norm":1,"l2valu":5.000e-3})
ipt.add_layer('O',{"nodes":48 ,"activation":9,"type":0,"l2norm":1,"l2valu":1.000e-6})
ipt.add_layer('O',{"nodes":48 ,"activation":9,"type":0,"l2norm":1,"l2valu":1.000e-6})

ipt.add_layer('S',{"nodes":192,"activation":9,"type":0,"l2norm":1,"l2valu":5.000e-3})
ipt.add_layer('S',{"nodes":48 ,"activation":9,"type":0,"l2norm":1,"l2valu":1.000e-6})
ipt.add_layer('S',{"nodes":80 ,"activation":9,"type":0,"l2norm":1,"l2valu":1.000e-6})

ipt.add_layer('F',{"nodes":192,"activation":9,"type":0,"l2norm":1,"l2valu":5.000e-3})
ipt.add_layer('F',{"nodes":48 ,"activation":9,"type":0,"l2norm":1,"l2valu":1.000e-6})
ipt.add_layer('F',{"nodes":80 ,"activation":9,"type":0,"l2norm":1,"l2valu":1.000e-6})

ipt.add_layer('Cl',{"nodes":192,"activation":9,"type":0,"l2norm":1,"l2valu":5.000e-3})
ipt.add_layer('Cl',{"nodes":48 ,"activation":9,"type":0,"l2norm":1,"l2valu":1.000e-6})
ipt.add_layer('Cl',{"nodes":80 ,"activation":9,"type":0,"l2norm":1,"l2valu":1.000e-6})

ipt.print_layer_parameters()

#-----------0---------

# Training varibles
Expand All @@ -74,6 +125,10 @@
'sig' : M,
}

strucsparams = {'N': 40, # number of maximum structures to select before QBC from each XYZ file
'sig': M,
}

mdsparams = {'N': 1, # trajectories to run
'T1': 300,
'T2': 1000,
Expand All @@ -92,7 +147,7 @@
'tsfiles': ['/home/jsmith48/scratch/auto_rxn_al/rxns/'],
'nmfile':None, #path to gaussian log file containing the data
'nm':0, #id of normal mode
'perc':0, #Move the molecules initial coordiantes along the mode by this amount. Negative numbers are ok.
'perc':0, #Move the molecules initial coordiantes along the mode by this amount. Negative numbers are ok.
}

dhparams = { 'Nmol': 250,
Expand All @@ -109,9 +164,9 @@
#'mdselect' : [(10,0), (1,11)],
#'N' : 20,
'maxNa' : 15, # Largest molecule to consider (for dimers max size is 2x maxNa)
'T' : 400.0, # running temp
'T' : 400.0, # running temp
'L' : 25.0, # box length
'V' : 0.04, # Random init velocities
'V' : 0.04, # Random init velocities
'dt' : 0.25, # MD time step
'Nm' : 100, # Molecules to embed
#'Nm' : 160, # Molecules to embed
Expand Down Expand Up @@ -144,7 +199,7 @@

### BEGIN CONFORMATIONAL REFINEMENT LOOP HERE ###
#N = [20]
N = [1]
N = [2]

for i in N:
#netdir = wkdir+'ANI-1x-RXN-0000.00'+str(i).zfill(2)+'/'
Expand All @@ -156,33 +211,32 @@

nnfprefix = netdir + 'train'

netdict = {'iptfile' : iptfile,
'cnstfile' : cstfile,
netdict = {'cnstfile' : cstfile,
'saefile': saefile,
'iptsize': 1008,
'nnfprefix': netdir+'train',
'aevsize': aevsize,
'num_nets': Nnets,
'atomtyp' : ['H','C','N','O','S','F','Cl']
}

## Train the ensemble ##
aln = att.alaniensembletrainer(netdir, netdict, h5stor, Nnets)
aln.build_strided_training_cache(Nblock,Nbvald,Nbtest,False)
aln = att.alaniensembletrainer(netdir, netdict, ipt, h5stor, Nnets)
aln.build_strided_training_cache(Nblock,Nbvald,Nbtest,Ekey='energy',forces=False,dipole=False,rmhighe=True)
aln.train_ensemble(GPU)

if not os.path.exists(root_dir + datdir + str(i+1).zfill(2)):
os.mkdir(root_dir + datdir + str(i+1).zfill(2))

## Run active learning sampling ##
acs = alt.alconformationalsampler(ldtdir, datdir + str(i+1).zfill(2), optlfile, fpatoms+['H'], netdict)
acs = alt.alconformationalsampler(ldtdir, datdir + str(i+1).zfill(2), optlfile, strucsfolder, fpatoms+['H'], netdict)
#acs.run_sampling_cluster(gcmddict, GPU)
#acs.run_sampling_dimer(dmrparams, GPU)
acs.run_sampling_nms(nmsparams, GPU)
#acs.run_sampling_md(mdsparams, perc=0.25, gpus=GPU+GPU+GPU)
#acs.run_sampling_strucs(strucsparams, GPU)
acs.run_sampling_md(mdsparams, perc=0.25, gpus=GPU+GPU+GPU)
#acs.run_sampling_dhl(dhparams, gpus=GPU+GPU)
#acs.run_sampling_TS(tsparams, gpus=GPU)

## Submit jobs, return and pack data
ast.generateQMdata(hostname, username, swkdir, ldtdir, datdir + str(i+1).zfill(2), h5stor, mae, jtime,max_jobs=50)


Loading