I’m training an encoder-decoder model (BART) on my task. My task is spell checking, I framed it as a seq2seq task, so for example given the sentence I liike thus framewok much., the model is trained to predict the corrected sentence : I like this framework much.. Now at inference time, I would like…

I could get something to work by using the keyword argument decoder_input_ids : # Create the input for the encoder : the sentence with typo encoder_inp = tokenizer(["I like this framewok"], max_length=model.config.max_position_embeddings, padding=True, truncation=True, return_tensors="pt") # Creat…

Encoder-decoder `generate()` with forced start for decoder?

🤗Transformers

joaogante November 18, 2022, 10:44am 4

Hi @astariul – yup, using decoder_input_ids is precisely what you should do

2 Likes

Topic		Replies	Views
Generate 'continuation' for seq2seq models Intermediate	1	1870	February 22, 2021
Rewriting generate function for manual decoder input 🤗Transformers	7	3568	July 11, 2022
Generate without using the generate method Intermediate	8	6275	January 17, 2025
Encoder-Decoder model only generates bos_token's [<s><s><s>] Models	17	3176	December 6, 2022
Decoder_start_token_id per sample or per batch during training 🤗Transformers	0	231	February 16, 2024

Encoder-decoder `generate()` with forced start for decoder?

Related topics