# What is the difference between lm\_labels and decoder\_input\_ids

**URL:** <https://discuss.huggingface.co/t/what-is-the-difference-between-lm-labels-and-decoder-input-ids/15664>\
**Category:** Models\
**Created:** [March 13, 2022, 11:07pm UTC](https://discuss.huggingface.co/t/what-is-the-difference-between-lm-labels-and-decoder-input-ids/15664 "2022-03-13T23:07:48Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![FirstKaiSa](https://avatars.discourse-cdn.com/v4/letter/f/839c29/32.png) [@FirstKaiSa](https://discuss.huggingface.co/u/FirstKaiSa)\
**Post date:** [March 13, 2022, 11:07pm UTC](https://discuss.huggingface.co/t/what-is-the-difference-between-lm-labels-and-decoder-input-ids/15664/1 "2022-03-13T23:07:48Z")

</div>

I didn’t find anything about **lm\_labels** on T5 docs. How to explain the below code:

```auto
model = T5ForConfiditionalGeneration.from_pretrained('t5-base')
outputs = model(inputs_ids=input_ids, attention_mask=mask, decoder_input_ids=y_ids, lm_labels=lm_labels)

```

How to distinguish the difference between decoder\_input\_ids and lm\_labels?