Merge pull request mravanelli#170 from shuttle1987/master

[MRG] Clean up source code formatting
BrandonLiang · Oct 23, 2019 · f9e7065 · f9e7065
2 parents 5374de9 + e5e759e
commit f9e7065
Show file tree

Hide file tree

Showing 17 changed files with 6,136 additions and 4,868 deletions.
diff --git a/core.py b/core.py
diff --git a/data_io.py b/data_io.py
diff --git a/kaldi_decoding_scripts/utils/filt.py b/kaldi_decoding_scripts/utils/filt.py
@@ -1,14 +1,15 @@
 #!/usr/bin/env python
 
 # Apache 2.0
+from __future__ import print_function
 
 import sys
 
-vocab=set()
+vocab = set()
 with open(sys.argv[1]) as vocabfile:
     for line in vocabfile:
         vocab.add(line.strip())
 
 with open(sys.argv[2]) as textfile:
     for line in textfile:
-        print " ".join(map(lambda word: word if word in vocab else '<UNK>', line.strip().split()))
+        print(" ".join(map(lambda word: word if word in vocab else "<UNK>", line.strip().split())))
diff --git a/kaldi_decoding_scripts/utils/nnet/gen_dct_mat.py b/kaldi_decoding_scripts/utils/nnet/gen_dct_mat.py
@@ -16,9 +16,10 @@
 # limitations under the License.
 
 # ./gen_dct_mat.py
-# script generates matrix with DCT transform, which is sparse 
-# and takes into account that data-layout is along frequency axis, 
+# script generates matrix with DCT transform, which is sparse
+# and takes into account that data-layout is along frequency axis,
 # while DCT is done along temporal axis.
+from __future__ import print_function
 
 from math import *
 import sys
@@ -27,41 +28,40 @@
 from optparse import OptionParser
 
 parser = OptionParser()
-parser.add_option('--fea-dim', dest='dim', help='feature dimension')
-parser.add_option('--splice', dest='splice', help='applied splice value')
-parser.add_option('--dct-basis', dest='dct_basis', help='number of DCT basis')
+parser.add_option("--fea-dim", dest="dim", help="feature dimension")
+parser.add_option("--splice", dest="splice", help="applied splice value")
+parser.add_option("--dct-basis", dest="dct_basis", help="number of DCT basis")
 (options, args) = parser.parse_args()
 
-if(options.dim == None):
+if options.dim == None:
     parser.print_help()
     sys.exit(1)
 
-dim=int(options.dim)
-splice=int(options.splice)
-dct_basis=int(options.dct_basis)
+dim = int(options.dim)
+splice = int(options.splice)
+dct_basis = int(options.dct_basis)
 
-timeContext=2*splice+1
+timeContext = 2 * splice + 1
 
 
-#generate the DCT matrix
+# generate the DCT matrix
 M_PI = 3.1415926535897932384626433832795
 M_SQRT2 = 1.4142135623730950488016887
 
 
-#generate sparse DCT matrix
-print '['
+# generate sparse DCT matrix
+print("[")
 for k in range(dct_basis):
     for m in range(dim):
         for n in range(timeContext):
-          if(n==0): 
-              print m*'0 ',
-          else: 
-              print (dim-1)*'0 ',
-          print str(sqrt(2.0/timeContext)*cos(M_PI/timeContext*k*(n+0.5))),
-          if(n==timeContext-1):
-              print (dim-m-1)*'0 ',
-        print
-    print 
-
-print ']'
-
+            if n == 0:
+                print(m * "0 ", end=" ")
+            else:
+                print((dim - 1) * "0 ", end=" ")
+            print(str(sqrt(2.0 / timeContext) * cos(M_PI / timeContext * k * (n + 0.5))), end=" ")
+            if n == timeContext - 1:
+                print((dim - m - 1) * "0 ", end=" ")
+        print()
+    print()
+
+print("]")
diff --git a/kaldi_decoding_scripts/utils/nnet/gen_hamm_mat.py b/kaldi_decoding_scripts/utils/nnet/gen_hamm_mat.py
@@ -17,6 +17,7 @@
 
 # ./gen_hamm_mat.py
 # script generates diagonal matrix with hamming window values
+from __future__ import print_function
 
 from math import *
 import sys
@@ -25,33 +26,31 @@
 from optparse import OptionParser
 
 parser = OptionParser()
-parser.add_option('--fea-dim', dest='dim', help='feature dimension')
-parser.add_option('--splice', dest='splice', help='applied splice value')
+parser.add_option("--fea-dim", dest="dim", help="feature dimension")
+parser.add_option("--splice", dest="splice", help="applied splice value")
 (options, args) = parser.parse_args()
 
-if(options.dim == None):
+if options.dim == None:
     parser.print_help()
     sys.exit(1)
 
-dim=int(options.dim)
-splice=int(options.splice)
+dim = int(options.dim)
+splice = int(options.splice)
 
 
-#generate the diagonal matrix with hammings
+# generate the diagonal matrix with hammings
 M_2PI = 6.283185307179586476925286766559005
 
-dim_mat=(2*splice+1)*dim
-timeContext=2*splice+1
-print '['
+dim_mat = (2 * splice + 1) * dim
+timeContext = 2 * splice + 1
+print("[")
 for row in range(dim_mat):
     for col in range(dim_mat):
-        if col!=row:
-            print '0',
+        if col != row:
+            print("0", end=" ")
         else:
-            i=int(row/dim)
-            print str(0.54 - 0.46*cos((M_2PI * i) / (timeContext-1))),
-    print
-
-print ']'
-
+            i = int(row / dim)
+            print(str(0.54 - 0.46 * cos((M_2PI * i) / (timeContext - 1))), end=" ")
+    print()
 
+print("]")
diff --git a/kaldi_decoding_scripts/utils/nnet/gen_splice.py b/kaldi_decoding_scripts/utils/nnet/gen_splice.py
@@ -17,6 +17,7 @@
 
 # ./gen_splice.py
 # generates <splice> Component
+from __future__ import print_function
 
 from math import *
 import sys
@@ -25,27 +26,31 @@
 from optparse import OptionParser
 
 parser = OptionParser()
-parser.add_option('--fea-dim', dest='dim_in', help='feature dimension')
-parser.add_option('--splice', dest='splice', help='number of frames to concatenate with the central frame')
-parser.add_option('--splice-step', dest='splice_step', help='splicing step (frames dont need to be consecutive, --splice 3 --splice-step 2 will select offsets: -6 -4 -2 0 2 4 6)', default='1' )
+parser.add_option("--fea-dim", dest="dim_in", help="feature dimension")
+parser.add_option("--splice", dest="splice", help="number of frames to concatenate with the central frame")
+parser.add_option(
+    "--splice-step",
+    dest="splice_step",
+    help="splicing step (frames dont need to be consecutive, --splice 3 --splice-step 2 will select offsets: -6 -4 -2 0 2 4 6)",
+    default="1",
+)
 (options, args) = parser.parse_args()
 
-if(options.dim_in == None):
+if options.dim_in == None:
     parser.print_help()
     sys.exit(1)
 
-dim_in=int(options.dim_in)
-splice=int(options.splice)
-splice_step=int(options.splice_step)
+dim_in = int(options.dim_in)
+splice = int(options.splice)
+splice_step = int(options.splice_step)
 
-dim_out=(2*splice+1)*dim_in
+dim_out = (2 * splice + 1) * dim_in
 
-print '<splice>', dim_out, dim_in
-print '[',
+print("<splice>", dim_out, dim_in)
+print("[", end=" ")
 
-splice_vec = range(-splice*splice_step, splice*splice_step+1, splice_step)
+splice_vec = range(-splice * splice_step, splice * splice_step + 1, splice_step)
 for idx in range(len(splice_vec)):
-    print splice_vec[idx],
-
-print ']'
+    print(splice_vec[idx], end=" ")
 
+print("]")
diff --git a/kaldi_decoding_scripts/utils/nnet/make_blstm_proto.py b/kaldi_decoding_scripts/utils/nnet/make_blstm_proto.py
@@ -16,61 +16,90 @@
 # limitations under the License.
 
 # Generated Nnet prototype, to be initialized by 'nnet-initialize'.
+from __future__ import print_function
 
 import sys
 
 ###
 ### Parse options
 ###
 from optparse import OptionParser
-usage="%prog [options] <feat-dim> <num-leaves> >nnet-proto-file"
+
+usage = "%prog [options] <feat-dim> <num-leaves> >nnet-proto-file"
 parser = OptionParser(usage)
 #
-parser.add_option('--num-cells', dest='num_cells', type='int', default=800, 
-                   help='Number of LSTM cells [default: %default]');
-parser.add_option('--num-recurrent', dest='num_recurrent', type='int', default=512, 
-                   help='Number of LSTM recurrent units [default: %default]');
-parser.add_option('--num-layers', dest='num_layers', type='int', default=2, 
-                   help='Number of LSTM layers [default: %default]');
-parser.add_option('--lstm-stddev-factor', dest='lstm_stddev_factor', type='float', default=0.01, 
-                   help='Standard deviation of initialization [default: %default]');
-parser.add_option('--param-stddev-factor', dest='param_stddev_factor', type='float', default=0.04, 
-                   help='Standard deviation in output layer [default: %default]');
-parser.add_option('--clip-gradient', dest='clip_gradient', type='float', default=5.0, 
-                   help='Clipping constant applied to gradients [default: %default]');
+parser.add_option(
+    "--num-cells", dest="num_cells", type="int", default=800, help="Number of LSTM cells [default: %default]"
+)
+parser.add_option(
+    "--num-recurrent",
+    dest="num_recurrent",
+    type="int",
+    default=512,
+    help="Number of LSTM recurrent units [default: %default]",
+)
+parser.add_option(
+    "--num-layers", dest="num_layers", type="int", default=2, help="Number of LSTM layers [default: %default]"
+)
+parser.add_option(
+    "--lstm-stddev-factor",
+    dest="lstm_stddev_factor",
+    type="float",
+    default=0.01,
+    help="Standard deviation of initialization [default: %default]",
+)
+parser.add_option(
+    "--param-stddev-factor",
+    dest="param_stddev_factor",
+    type="float",
+    default=0.04,
+    help="Standard deviation in output layer [default: %default]",
+)
+parser.add_option(
+    "--clip-gradient",
+    dest="clip_gradient",
+    type="float",
+    default=5.0,
+    help="Clipping constant applied to gradients [default: %default]",
+)
 #
-(o,args) = parser.parse_args()
-if len(args) != 2 : 
-  parser.print_help()
-  sys.exit(1)
+(o, args) = parser.parse_args()
+if len(args) != 2:
+    parser.print_help()
+    sys.exit(1)
 
-(feat_dim, num_leaves) = map(int,args);
+(feat_dim, num_leaves) = list(map(int, args))
 
 # Original prototype from Jiayu,
-#<NnetProto>
-#<Transmit> <InputDim> 40 <OutputDim> 40
-#<LstmProjectedStreams> <InputDim> 40 <OutputDim> 512 <CellDim> 800 <ParamScale> 0.01 <NumStream> 4
-#<AffineTransform> <InputDim> 512 <OutputDim> 8000 <BiasMean> 0.000000 <BiasRange> 0.000000 <ParamStddev> 0.04
-#<Softmax> <InputDim> 8000 <OutputDim> 8000
-#</NnetProto>
+# <NnetProto>
+# <Transmit> <InputDim> 40 <OutputDim> 40
+# <LstmProjectedStreams> <InputDim> 40 <OutputDim> 512 <CellDim> 800 <ParamScale> 0.01 <NumStream> 4
+# <AffineTransform> <InputDim> 512 <OutputDim> 8000 <BiasMean> 0.000000 <BiasRange> 0.000000 <ParamStddev> 0.04
+# <Softmax> <InputDim> 8000 <OutputDim> 8000
+# </NnetProto>
 
-print "<NnetProto>"
+print("<NnetProto>")
 # normally we won't use more than 2 layers of LSTM
 if o.num_layers == 1:
-    print "<BLstmProjectedStreams> <InputDim> %d <OutputDim> %d <CellDim> %s <ParamScale> %f <ClipGradient> %f" % \
-        (feat_dim, 2*o.num_recurrent, o.num_cells, o.lstm_stddev_factor, o.clip_gradient)
+    print(
+        "<BLstmProjectedStreams> <InputDim> %d <OutputDim> %d <CellDim> %s <ParamScale> %f <ClipGradient> %f"
+        % (feat_dim, 2 * o.num_recurrent, o.num_cells, o.lstm_stddev_factor, o.clip_gradient)
+    )
 elif o.num_layers == 2:
-    print "<BLstmProjectedStreams> <InputDim> %d <OutputDim> %d <CellDim> %s <ParamScale> %f <ClipGradient> %f" % \
-        (feat_dim, 2*o.num_recurrent, o.num_cells, o.lstm_stddev_factor, o.clip_gradient)
-    print "<BLstmProjectedStreams> <InputDim> %d <OutputDim> %d <CellDim> %s <ParamScale> %f <ClipGradient> %f" % \
-        (2*o.num_recurrent, 2*o.num_recurrent, o.num_cells, o.lstm_stddev_factor, o.clip_gradient)
+    print(
+        "<BLstmProjectedStreams> <InputDim> %d <OutputDim> %d <CellDim> %s <ParamScale> %f <ClipGradient> %f"
+        % (feat_dim, 2 * o.num_recurrent, o.num_cells, o.lstm_stddev_factor, o.clip_gradient)
+    )
+    print(
+        "<BLstmProjectedStreams> <InputDim> %d <OutputDim> %d <CellDim> %s <ParamScale> %f <ClipGradient> %f"
+        % (2 * o.num_recurrent, 2 * o.num_recurrent, o.num_cells, o.lstm_stddev_factor, o.clip_gradient)
+    )
 else:
     sys.stderr.write("make_lstm_proto.py ERROR: more than 2 layers of LSTM, not supported yet.\n")
     sys.exit(1)
-print "<AffineTransform> <InputDim> %d <OutputDim> %d <BiasMean> 0.0 <BiasRange> 0.0 <ParamStddev> %f" % \
-    (2*o.num_recurrent, num_leaves, o.param_stddev_factor)
-print "<Softmax> <InputDim> %d <OutputDim> %d" % \
-    (num_leaves, num_leaves)
-print "</NnetProto>"
-
-
+print(
+    "<AffineTransform> <InputDim> %d <OutputDim> %d <BiasMean> 0.0 <BiasRange> 0.0 <ParamStddev> %f"
+    % (2 * o.num_recurrent, num_leaves, o.param_stddev_factor)
+)
+print("<Softmax> <InputDim> %d <OutputDim> %d" % (num_leaves, num_leaves))
+print("</NnetProto>")