the size of pre-train model 's input data
willow-yll opened this issue · 1 comments
willow-yll commented
Hi,
I want to use pretrain-model and add some layers to train a downstream task,
what kind of data should I input?
Should I process the amino acid sequence into one-hot encoding?
Or if there is a file of tape I can use to deal with the amino acid sequence?
I don't know the data size to input....
my model like this:
pretrained = pretrained = UniRepModel.from_pretrained('babbler-1900', force_download=False)
class UniRep_bilstm(nn.Module):
def __init__(self,emb_dim,hidden_dim,num_layers,output_dim,max_len):
super(UniRep_bilstm,self).__init__()
self.unirep = pretrained
self.bilstm = bilstm(emb_dim,hidden_dim,num_layers,output_dim,max_len)
def forward(self,input_ids):
unirep_outputs = self.unirep(input_ids)
outputs = self.bilstm(unirep_outputs)
return outputs
willow-yll commented
Could you please tell me the size and type of input data , and the way to deal with amino acid sequence? thanks :)