Download Built-in model parameters from ASE_ANI (#481)

* download resources from ASE_ANI

* update gitignore

* get resources with request

* fix flke8

* rm svn checkout for resources

* update

* fix flake8

.gitignore

@@ -30,6 +30,9 @@ dist
 *.swo
 /download
 /download.tar.xz
+/torchani/resources/*.info
+/torchani/resources/ani-1ccx_8x
+/torchani/resources/ani-2x_8x
 *.qdrep
 *.qdstrm
 *.zip

download.sh

 #!/bin/bash
 
+# download the data
 wget https://www.dropbox.com/sh/2c8zdqc1hrqsgwy/AAD6l24ngoiFa6DRapF6HPk5a/ -O download.zip
 unzip download.zip -d download || [[ $? == 2 ]]  # unzip return 2 for dropbox created zip file

torchani/models.py

@@ -24,10 +24,12 @@ directly calculate energies or get an ASE calculator. For example:
     model0.species_to_tensor(['C', 'H', 'H', 'H', 'H'])
 """
 import os
+import io
+import requests
+import zipfile
 import torch
 from torch import Tensor
 from typing import Tuple, Optional
-from pkg_resources import resource_filename
 from . import neurochem
 from .nn import SpeciesConverter, SpeciesEnergies
 from .aev import AEVComputer
@@ -70,11 +72,26 @@ class BuiltinModel(torch.nn.Module):
 
     @staticmethod
     def _parse_neurochem_resources(info_file_path):
-        def get_resource(file_path):
-            package_name = '.'.join(__name__.split('.')[:-1])
-            return resource_filename(package_name, 'resources/' + file_path)
-
-        info_file = get_resource(info_file_path)
+        def get_resource(resource_path, file_path):
+            return os.path.join(resource_path, 'resources/' + file_path)
+
+        resource_path = os.path.dirname(__file__)
+        local_dir = os.path.expanduser('~/.local/torchani')
+
+        if not os.path.isfile(get_resource(resource_path, info_file_path)):
+            if not os.path.isfile(get_resource(local_dir, info_file_path)):
+                print('Downloading ANI model parameters ...')
+                resource_res = requests.get("https://www.dropbox.com/sh/otrzul6yuye8uzs/AABuaihE22vtaB_rdrI0r6TUa?dl=1")
+                resource_zip = zipfile.ZipFile(io.BytesIO(resource_res.content))
+                try:
+                    resource_zip.extractall(resource_path)
+                except PermissionError:
+                    resource_zip.extractall(local_dir)
+                    resource_path = local_dir
+            else:
+                resource_path = local_dir
+
+        info_file = get_resource(resource_path, info_file_path)
 
         with open(info_file) as f:
             # const_file: Path to the file with the builtin constants.
@@ -82,9 +99,9 @@ class BuiltinModel(torch.nn.Module):
             # ensemble_prefix: Prefix of the neurochem resource directories.
             lines = [x.strip() for x in f.readlines()][:4]
             const_file_path, sae_file_path, ensemble_prefix_path, ensemble_size = lines
-            const_file = get_resource(const_file_path)
-            sae_file = get_resource(sae_file_path)
-            ensemble_prefix = get_resource(ensemble_prefix_path)
+            const_file = get_resource(resource_path, const_file_path)
+            sae_file = get_resource(resource_path, sae_file_path)
+            ensemble_prefix = get_resource(resource_path, ensemble_prefix_path)
             ensemble_size = int(ensemble_size)
             consts = neurochem.Constants(const_file)
         return consts, sae_file, ensemble_prefix, ensemble_size

torchani/resources/ani-1ccx_8x.info

-ani-1ccx_8x/rHCNO-5.2R_16-3.5A_a4-8.params
-ani-1ccx_8x/sae_linfit.dat
-ani-1ccx_8x/train
-8
-

torchani/resources/ani-1ccx_8x/rHCNO-5.2R_16-3.5A_a4-8.params

-TM = 1
-Rcr = 5.2000e+00
-Rca = 3.5000e+00
-EtaR = [1.6000000e+01]
-ShfR = [9.0000000e-01,1.1687500e+00,1.4375000e+00,1.7062500e+00,1.9750000e+00,2.2437500e+00,2.5125000e+00,2.7812500e+00,3.0500000e+00,3.3187500e+00,3.5875000e+00,3.8562500e+00,4.1250000e+00,4.3937500e+00,4.6625000e+00,4.9312500e+00]
-Zeta = [3.2000000e+01]
-ShfZ = [1.9634954e-01,5.8904862e-01,9.8174770e-01,1.3744468e+00,1.7671459e+00,2.1598449e+00,2.5525440e+00,2.9452431e+00]
-EtaA = [8.0000000e+00]
-ShfA = [9.0000000e-01,1.5500000e+00,2.2000000e+00,2.8500000e+00]
-Atyp = [H,C,N,O]

torchani/resources/ani-1ccx_8x/sae_linfit.dat

-H,0=-0.5991501324919538
-C,1=-38.03750806057356
-N,2=-54.67448347695333
-O,3=-75.16043537275567

torchani/resources/ani-1ccx_8x/train0/inputtrain.ipt

-!InputFile for Force Prediction Network
-sflparamsfile=rHCNO-5.2R_16-3.5A_a4-8.params
-ntwkStoreDir=networks/
-atomEnergyFile=sae_linfit.dat
-nmax=0! Maximum number of iterations (0 = inf)
-tolr=200! Tolerance - early stopping
-emult=0.5!Multiplier by eta after tol switch
-eta=0.001! Eta -- Learning rate
-tcrit=1.0E-6! Eta termination criterion
-tmax=0! Maximum time (0 = inf)
-tbtchsz=2560
-vbtchsz=2560
-gpuid=0
-ntwshr=0
-nkde=2
-energy=1
-freezelayer=0,2
-force=0
-fmult=0.0
-pbc=0
-cmult =0.001
-runtype=ANNP_CREATE_HDNN_AND_TRAIN!Create and train a HDN network
-network_setup {
-  inputsize=384;
-  atom_net H $
-    layer [
-          nodes=160;
-          activation=9;
-          type=0;
-    ]
-    layer [
-          nodes=128;
-          activation=9;
-          type=0;
-          l2norm=1;
-          l2valu=0.00001;
-    ]
-    layer [
-          nodes=96;
-          activation=9;
-          type=0;
-          l2norm=1;
-          l2valu=0.000001;
-    ]
-    layer [
-          nodes=1;
-          activation=6;
-          type=0;
-    ]
-    $
-  atom_net C $
-    layer [
-          nodes=144;
-          activation=9;
-          type=0;
-    ]
-    layer [
-          nodes=112;
-          activation=9;
-          type=0;
-          l2norm=1;
-          l2valu=0.00001;
-    ]
-    layer [
-          nodes=96;
-          activation=9;
-          type=0;
-          l2norm=1;
-          l2valu=0.000001;
-    ]
-    layer [
-          nodes=1;
-          activation=6;
-          type=0;
-    ]
-    $
-  atom_net N $
-    layer [
-          nodes=128;
-          activation=9;
-          type=0;
-    ]
-    layer [
-          nodes=112;
-          activation=9;
-          type=0;
-          l2norm=1;
-          l2valu=0.00001;
-    ]
-    layer [
-          nodes=96;
-          activation=9;
-          type=0;
-          l2norm=1;
-          l2valu=0.000001;
-    ]
-    layer [
-          nodes=1;
-          activation=6;
-          type=0;
-    ]
-    $
-  atom_net O $
-    layer [
-          nodes=128;
-          activation=9;
-          type=0;
-    ]
-    layer [
-          nodes=112;
-          activation=9;
-          type=0;
-          l2norm=1;
-          l2valu=0.00001;
-    ]
-    layer [
-          nodes=96;
-          activation=9;
-          type=0;
-          l2norm=1;
-          l2valu=0.000001;
-    ]
-    layer [
-          nodes=1;
-          activation=6;
-          type=0;
-    ]
-    $
-
-}
-adptlrn=OFF ! Adaptive learning (OFF,RMSPROP)
-decrate=0.9 !Decay rate of RMSPROP
-moment=ADAM! Turn on momentum or nesterov momentum (OFF,CNSTTEMP,TMANNEAL,REGULAR,NESTEROV)
-mu=0.99 ! Mu factor for momentum
-

torchani/resources/ani-1ccx_8x/train0/networks/ANN-C.nnf-l4.bparam

-iU
\ No newline at end of file

torchani/resources/ani-1ccx_8x/train0/networks/ANN-C.nnf-l4.wparam

-9l=a9vw<<޽e;&=˼Th
<eKRІ=-=AܽjT==b><
-=8<KΧ;VCyd
-p=xn'<n%X > *<<u$(켭S=oX;n=˄۾<5Kz5_LJW:=P9ܼ!)3ֳ=ƽ>fʼŐ=H?AX=72M	w+p(=;=7#%=<y۽to@qT=WbEq c@7Q۴捋=4.==|EZ=>?f>.f=dk<Mk<y<Ba=-B썽=5^=r=
\ No newline at end of file