def __init__(self, vocab_size, d_model, N, heads, dropout):
super().__init__()
self.N = N
self.embed = Embedder(vocab_size, d_model)
self.pe = PositionalEncoder(d_model, dropout=dropout)
self.layers = get_clones(DecoderLayer(d_model, heads, dropout), N)
self.norm = Norm(d_model)
def __init__(self, bert, hidden_size, num_hidden_layers, num_attention_heads, dropout):
super().__init__()
self.N = num_hidden_layers
self.bert = bert
self.pe = PositionalEncoder(hidden_size, dropout=dropout)
self.layers = get_clones(DecoderLayer(hidden_size, num_attention_heads, dropout), num_hidden_layers)
self.norm = Norm(hidden_size)
def __init__(self, vocab_size, d_model, N, heads, dropout, c, n_layers):
super().__init__()
self.N = N
self.embed = Embedder(vocab_size, d_model)
self.pe = PositionalEncoder(d_model, dropout=dropout)
self.dynam = Ucb_bandit(n_layers, c)
self.layer = EncoderLayer(d_model, heads, dropout)
self.layers = get_clones(EncoderLayer(d_model, heads, dropout), N)
self.norm = Norm(d_model)
def __init__(self, vocab_size, d_model, N, heads, dropout, field, word_emb,
opt):
super().__init__()
self.N = N
self.word_emb = word_emb
self.opt = opt # unused, just for querying
self.embed = Embedder(vocab_size, d_model, word_emb, field)
self.pe = PositionalEncoder(d_model, dropout=dropout)
self.layers = get_clones(EncoderLayer(d_model, heads, dropout),
N) # attention
self.norm = Norm(d_model)
def __init__(self, vocab_size, d_model, N, heads, dropout, device):
super().__init__()
self.N = N
# We need to use the embedder
# self.embed = Embedder(vocab_size, d_model)
# self.embed = nn.Linear(vocab_size, d_model)
self.pe = PositionalEncoder(d_model, dropout=dropout, device=device)
self.layers = get_clones(EncoderLayer(d_model, heads, dropout), N)
self.norm = Norm(d_model)