if args.is_interleave == 1: # fixed interleaver.
seed = np.random.randint(0, 1)
rand_gen = mtrand.RandomState(seed)
p_array = rand_gen.permutation(arange(args.block_len))
elif args.is_interleave == 0:
p_array = range(args.block_len) # no interleaver.
else:
seed = np.random.randint(0, args.is_interleave)
rand_gen = mtrand.RandomState(seed)
p_array = rand_gen.permutation(arange(args.block_len))
print('using random interleaver', p_array)
encoder = ENC(args, p_array)
decoder = DEC(args, p_array)
# choose support channels
from channel_ae import Channel_AE
model = Channel_AE(args, encoder, decoder).to(device)
# make the model parallel
if args.is_parallel == 1:
model.enc.set_parallel()
model.dec.set_parallel()
# weight loading
if args.init_nw_weight == 'default':
pass
else:
elif args.is_interleave == 0:
p_array1 = range(args.block_len) # no interleaver.
p_array2 = range(args.block_len) # no interleaver.
else:
seed = np.random.randint(0, args.is_interleave)
rand_gen = mtrand.RandomState(seed)
p_array1 = rand_gen.permutation(arange(args.block_len))
seed = np.random.randint(0, args.is_interleave)
rand_gen = mtrand.RandomState(seed)
p_array2 = rand_gen.permutation(arange(args.block_len))
print('using random interleaver', p_array1, p_array2)
if args.encoder == 'turboae_2int' and args.decoder == 'turboae_2int':
encoder = ENC(args, p_array1, p_array2)
decoder = DEC(args, p_array1, p_array2)
else:
encoder = ENC(args, p_array1)
decoder = DEC(args, p_array1)
# choose support channels
from channel_ae import Channel_AE
model = Channel_AE(args, encoder, decoder).to(device)
# model = Channel_ModAE(args, encoder, decoder).to(device)
# make the model parallel
if args.is_parallel == 1:
model.enc.set_parallel()
model.dec.set_parallel()