Hi, at training, I’m using the forward pass and batch_decode on the logits to get the decoded output: outputs = model( input_ids, attention_mask, dec_input_ids, dec_attention_mask, labels=dec_input_ids, ) loss, logits = outputs.loss, outputs.logi…

T5 forward pass versus generate, latter outputs non-sense

deathcrush March 25, 2021, 1:06pm 9

I happened to stumble across this post, which has a similar flavour to your error. Have a look!

Topic		Replies	Views
Untrained T5 model outputting logits that argmax to the decoder_input_ids Beginners	0	395	September 28, 2022
The output of T5 is not consistent on multiple sequences 🤗Transformers	1	719	May 11, 2022
Understanding Output of `PreTrainedModel.forward` Beginners	2	468	February 12, 2024
Decode whisper logits to transcript using forward instead of generate method 🤗Transformers	3	1320	December 20, 2022
Proper way to do conditional generation with T5 Beginners	1	1212	January 20, 2023