def load_abs_net(abs_dir):
abs_meta = json.load(open(join(abs_dir, 'meta.json')))
assert abs_meta['net'] == 'base_abstractor'
abs_args = abs_meta['net_args']
abs_ckpt = load_best_ckpt(abs_dir)
word2id = pkl.load(open(join(abs_dir, 'vocab.pkl'), 'rb'))
abstractor = CopySumm(**abs_args)
abstractor.load_state_dict(abs_ckpt)
return abstractor, word2id
def configure_net(abs_dir):
abs_meta = json.load(open(join(abs_dir, 'meta.json')))
assert abs_meta['net'] == 'base_abstractor'
abs_meta = json.load(open(join(abs_dir, 'meta.json')))
assert abs_meta['net'] == 'base_abstractor'
net_args = abs_meta['net_args']
abs_ckpt = load_best_ckpt(abs_dir)
net = CopySumm(**net_args)
net.load_state_dict(abs_ckpt)
return net, net_args
def __init__(self, abs_dir, max_len=30, cuda=True):
abs_meta = json.load(open(join(abs_dir, 'meta.json')))
assert abs_meta['net'] == 'base_abstractor'
abs_args = abs_meta['net_args']
abs_ckpt = load_best_ckpt(abs_dir)
word2id = pkl.load(open(join(abs_dir, 'vocab.pkl'), 'rb'))
abstractor = CopySumm(**abs_args)
abstractor.load_state_dict(abs_ckpt)
self._device = torch.device('cuda' if cuda else 'cpu')
self._net = abstractor.to(self._device)
self._word2id = word2id
self._id2word = {i: w for w, i in word2id.items()}
self._max_len = max_len
def __init__(self, abs_dir, max_len=30, cuda=True):
abs_meta = json.load(open(join(abs_dir, 'meta.json')))
assert abs_meta['net'] == 'base_abstractor'
abs_args = abs_meta['net_args']
abs_ckpt = load_best_ckpt(abs_dir)
word2id = pkl.load(open(join(abs_dir, 'vocab.pkl'), 'rb'))
abstractor = CopySumm(**abs_args)
abstractor.load_state_dict(abs_ckpt)
self._device = torch.device('cuda' if cuda else 'cpu')
self._net = abstractor.to(self._device)
self._word2id = word2id
self._id2word = {i: w for w, i in word2id.items()}
self._max_len = max_len
def configure_net(vocab_size, emb_dim, n_hidden, bidirectional, n_layer):
net_args = {}
net_args['vocab_size'] = vocab_size
net_args['emb_dim'] = emb_dim
net_args['n_hidden'] = n_hidden
net_args['bidirectional'] = bidirectional
net_args['n_layer'] = n_layer
net = CopySumm(**net_args)
return net, net_args
def configure_net(vocab_size,
emb_dim,
n_hidden,
bidirectional,
n_layer,
load_from=None):
net_args = {}
net_args['vocab_size'] = vocab_size
net_args['emb_dim'] = emb_dim
net_args['n_hidden'] = n_hidden
net_args['bidirectional'] = bidirectional
net_args['n_layer'] = n_layer
net = CopySumm(**net_args)
if load_from is not None:
abs_ckpt = load_best_ckpt(load_from)
net.load_state_dict(abs_ckpt)
return net, net_args
def configure_net(vocab_size, emb_dim, n_hidden, bidirectional, n_layer):
net_args = {}
net_args['vocab_size'] = vocab_size
net_args['emb_dim'] = emb_dim
net_args['n_hidden'] = n_hidden
net_args['bidirectional'] = bidirectional
net_args['n_layer'] = n_layer
net_args[
'dropoute'] = 0.2 # dropout to remove words from embedding layer (0 = no dropout)
net_args[
'dropouti'] = 0.2 # dropout for input embedding layers (0 = no dropout)
net_args[
'dropout'] = 0.2 # dropout applied to other layers (0 = no dropout)
net_args[
'wdrop'] = 0.2 # amount of weight dropout to apply to the RNN hidden to hidden matrix
net_args['dropouth'] = 0.2 # dropout for rnn layers (0 = no dropout)
net = CopySumm(**net_args)
return net, net_args
