analyze_cmd.add_argument(
"--dataset",
type=str,
default='swbd_verification',
choices=['swbd_recognition', 'swbd_verification', 'swbd_short'],
help="Which dataset to use")
rng = numpy.random.RandomState(1)
args = ap.parse_args()
t_start = time.time()
if args.cmd == 'analyze':
batchsize = args.batchsize
filename = args.load_from
print "load from %s" % filename
# model = pickle.load(open(filename,'rb'))
model = NN_parallel({}, njobs=1)
model.load(filename)
npz_tst = numpy.load(
"/data2/tawara/work/ttic/data/icassp15.0/swbd.test.npz")
npz_trn = numpy.load(
"/data2/tawara/work/ttic/data/icassp15.0/swbd.train.npz")
utt_ids_trn = sorted(npz_trn.keys())
utt_ids_tst = sorted(npz_tst.keys())
x_train = numpy.array([npz_trn[i]
for i in utt_ids_trn]).astype(numpy.float32)
x_test = numpy.array([npz_tst[i]
for i in utt_ids_tst]).astype(numpy.float32)
tmp = []
for l in x_test:
tmp.append([l])
x_test = numpy.array(tmp)
ali_to_pdf=KaldiCommand('bin/ali-to-pdf', option='/data2/tawara/work/ttic/MyPython/src/kaldi/timit/exp/tri3_ali/final.mdl')
# feats=KaldiCommand('featbin/apply-cmvn', '--norm-means=true --norm-vars=true --utt2spk=ark:/data2/tawara/work/ttic/MyPython/src/kaldi/timit/data/fbank/train_tr90/utt2spk scp:/data2/tawara/work/ttic/MyPython/src/kaldi/timit/data/fbank/train_tr90/cmvn.scp')
x_train, y_train = load_timit_labelled_kaldi(\
# feats('/data2/tawara/work/ttic/MyPython/src/kaldi/timit/data/fbank/train_tr90/feats.scp'),
KaldiArk('/data2/tawara/work/ttic/MyPython/src/kaldi/timit/feats_cmvn.ark'),
ali_to_pdf('\"gunzip -c /data2/tawara/work/ttic/MyPython/src/kaldi/timit/exp/tri3_ali/ali.*.gz |\"'), \
nnet_transf = '/data2/tawara/work/ttic/MyPython/src/kaldi/timit/final.feature_transform')
feats=KaldiCommand('featbin/apply-cmvn', '--norm-means=true --norm-vars=true --utt2spk=ark:/data2/tawara/work/ttic/MyPython/src/kaldi/timit/data/fbank/train_cv10/utt2spk scp:/data2/tawara/work/ttic/MyPython/src/kaldi/timit/data/fbank/train_cv10/cmvn.scp')
x_valid, y_valid = load_timit_labelled_kaldi( \
feats('/data2/tawara/work/ttic/MyPython/src/kaldi/timit/data/fbank/train_cv10/feats.scp'),
ali_to_pdf('\"gunzip -c /data2/tawara/work/ttic/MyPython/src/kaldi/timit/exp/tri3_ali/ali.*.gz |\"'), \
nnet_transf = '/data2/tawara/work/ttic/MyPython/src/kaldi/timit/final.feature_transform')
model = NN_parallel(option_dict, njobs=1)
nnet_file='/data2/tawara/work/ttic/MyPython/src/kaldi/timit/exp/fbank/dnn4_nn5_1024_cmvn_splice10_pretrain-dbn_dnn/nnet_dbn_dnn.init'
if nnet_file is not None:
model.load_parameters_from_nnet(nnet_file)
model.initialize()
N_train_all = x_train.shape[0]
indexes_l = range(N_train_all)
indexes_ul = list(set(range(0,N_train_all))-set(indexes_l))
N_train_l = len(indexes_l)
N_train_ul = len(indexes_ul)
assert N_train_all == N_train_l + N_train_ul
N_valid = x_valid.shape[0]
indexes_valid = range(0, N_valid)
for epoch in xrange(n_epoch):
analyze_cmd.add_argument("--data_type", type=str, default='test', help="Data type to evaluate on")
analyze_cmd.add_argument("--save_to", type=str, default='noname', help="Destination to save the state and results")
analyze_cmd.add_argument("--layer", type=int, default=0, help="Extract layer")
analyze_cmd.add_argument("--batchsize", type=int, default=default_options_dict['batch_size'], help="batchsize")
analyze_cmd.add_argument("--dataset", type=str, default='swbd_verification',
choices=['swbd_recognition', 'swbd_verification', 'swbd_short'], help="Which dataset to use")
rng = numpy.random.RandomState(1)
args = ap.parse_args()
t_start = time.time()
if args.cmd == 'analyze':
batchsize = args.batchsize
filename = args.load_from
print "load from %s" % filename
# model = pickle.load(open(filename,'rb'))
model = NN_parallel({}, njobs=1)
model.load(filename)
npz_tst = numpy.load("/data2/tawara/work/ttic/data/icassp15.0/swbd.test.npz")
npz_trn = numpy.load("/data2/tawara/work/ttic/data/icassp15.0/swbd.train.npz")
utt_ids_trn = sorted(npz_trn.keys())
utt_ids_tst = sorted(npz_tst.keys())
x_train = numpy.array([npz_trn[i] for i in utt_ids_trn]).astype(numpy.float32)
x_test = numpy.array([npz_tst[i] for i in utt_ids_tst]).astype(numpy.float32)
tmp=[]
for l in x_test:
tmp.append([l])
x_test=numpy.array(tmp)
tmp=[]
for l in x_train:
tmp.append([l])
x_train=numpy.array(tmp)
