PoliBa-DeepLearning/NanoSocrates
2
0
Fork 0
Code Issues 3 Pull Requests 1 Actions Packages Projects Releases Wiki Activity

updated decoder_input to work without embedder

Browse Source
This commit is contained in:
GassiGiuseppe 2025-10-11 16:53:36 +02:00
parent 1649cd7768
commit 49946727d8
1 changed files with 4 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
Project_Model/Libs/Transformer/Utils/decoder_input.py
View File

@ -1,15 +1,14 @@
import torch import torch
from Project_Model.Libs.Transformer import normalize_sequence from Project_Model.Libs.Transformer import normalize_sequence
from Project_Model.Libs.Embedder import NanoSocratesEmbedder as Embedder # from Project_Model.Libs.Embedder import NanoSocratesEmbedder as Embedder
def get_decoder_input(batch_size, sos_token,pad_token, seq_len, embedder: Embedder): def get_decoder_input(batch_size, sos_token,pad_token, seq_len):
single_decoder_input, _ = normalize_sequence([sos_token],seq_len,pad_token, end_token=0, add_ending=False) single_decoder_input, _ = normalize_sequence([sos_token],seq_len,pad_token, end_token=0, add_ending=False)
tensor_decoder_input = torch.tensor(single_decoder_input[:]) tensor_decoder_input = torch.tensor(single_decoder_input[:])
embedded_decoder_intput = embedder(tensor_decoder_input) # embedded_decoder_intput = embedder(tensor_decoder_input)
batch_decoder_input = embedded_decoder_intput.unsqueeze(0).repeat(batch_size, 1) batch_decoder_input = tensor_decoder_input.unsqueeze(0).repeat(batch_size, 1)
return batch_decoder_input return batch_decoder_input