def __init__(self,
input_dim,
emb_dim,
enc_hid_dim,
dec_hid_dim,
attn_dim,
num_layers,
dropout,
attention,
dec_method,
emb_freeze,
pad_idx,
embeddings=None
):
super().__init__()
self.emb_dim = emb_dim
self.enc_hid_dim = enc_hid_dim
self.dec_hid_dim = dec_hid_dim
dec_input_dim = self.calc_dec_input_dim()
self.output_dim = input_dim
self.num_layers = num_layers
self.dec_method = dec_method
self.emb = utils.embedding(input_dim, emb_dim, embeddings, emb_freeze, pad_idx)
self.decoder = nn.GRU(dec_input_dim, dec_hid_dim,
num_layers=num_layers, bidirectional=False)
self.out = nn.Linear(enc_hid_dim + dec_hid_dim + emb_dim, self.output_dim)
self.bias_out = nn.Linear(enc_hid_dim + dec_hid_dim + emb_dim, self.output_dim)
self.Vo = nn.Parameter(torch.Tensor(dec_hid_dim, 1))
self.attend = attention
self.dropout = nn.Dropout(dropout)
nn.init.kaiming_uniform_(self.Vo, a=math.sqrt(5))
def __init__(
self,
input_dim,
emb_dim,
enc_hid_dim,
n_profile,
attention,
dropout,
emb_freeze,
pad_idx,
embeddings=None
):
"""use better classify:
https://github.com/brightmart/text_classification
https://github.com/kk7nc/Text_Classification
https://github.com/nadbordrozd/text-top-model
https://github.com/prakashpandey9/Text-Classification-Pytorch
https://github.com/dennybritz/cnn-text-classification-tf
"""
super().__init__()
self.emb_dim = emb_dim
self.emb = utils.embedding(input_dim, emb_dim, embeddings, emb_freeze, pad_idx)
self.attention = attention
self.dropout = nn.Dropout(dropout)
def __init__(self,
input_dim,
emb_dim,
enc_hid_dim,
dec_hid_dim,
attn_dim,
num_layers,
dropout,
attention,
emb_freeze,
pad_idx,
embeddings=None
):
super().__init__()
self.emb_dim = emb_dim
self.enc_hid_dim = enc_hid_dim
self.dec_hid_dim = dec_hid_dim
dec_input_dim = self.calc_dec_input_dim()
self.output_dim = input_dim
self.num_layers = num_layers
self.emb = utils.embedding(input_dim, emb_dim, embeddings, emb_freeze, pad_idx)
self.decoder = nn.GRU(dec_input_dim, dec_hid_dim,
num_layers=num_layers, bidirectional=False)
self.out = nn.Linear(enc_hid_dim + dec_hid_dim + emb_dim, self.output_dim)
self.attend = attention
self.dropout = nn.Dropout(dropout)
def __init__(
self,
input_dim,
emb_dim,
dropout,
emb_freeze,
pad_idx,
embeddings=None
):
super().__init__()
self.emb = utils.embedding(input_dim, emb_dim, embeddings, emb_freeze, pad_idx)
self.dropout = nn.Dropout(dropout)
def __init__(self,
input_dim,
emb_dim,
enc_hid_dim,
dec_hid_dim,
num_layers,
dropout,
enc_bidi,
emb_freeze,
pad_idx,
embeddings=None
):
super().__init__()
self.enc_bidi = enc_bidi
self.num_layers = num_layers
self.num_directions = _num_dir(enc_bidi)
self.enc_hid_dim = enc_hid_dim
self.emb = utils.embedding(input_dim, emb_dim, embeddings, emb_freeze, pad_idx)
self.encoder = nn.GRU(emb_dim, enc_hid_dim,
num_layers=num_layers, bidirectional=enc_bidi)
self.out = nn.Linear(enc_hid_dim * self.num_directions, dec_hid_dim)
self.dropout = nn.Dropout(dropout)