Open
Description
self.calculate_num = args.enc_calculate_num
self.rk_type = args.rk_type
if self.calculate_num == 2 and self.rk_type == "learnable":
# RK2-learnable
self.gate_linear = Linear(2 * embed_dim, 1)
elif self.calculate_num == 2 and self.rk_type == "initialization"
self.alpha = torch.nn.Parameter(torch.Tensor(self.calculate_num))
self.alpha.data.fill_(1)
else:
raise ValueError("invalid rk_type!")
self.use_word_dropout = args.use_word_dropout
self.word_dropout = args.word_dropout
I'm currently learning from your ODE transformer implementation, and I believe there might be a missing colon on line 330 of fairseq/models/ode_transformer.py. This co//uld potentially lead to an error during execution.
Metadata
Metadata
Assignees
Labels
No labels