Timeline for Bert-Transformer : Why Bert transformer uses [CLS] token for classification instead of average over all tokens?
Current License: CC BY-SA 4.0
3 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jul 13, 2022 at 16:27 | comment | added | roman | @canP you're right that any method works as long as you're finetuning your model. However, with a pre-trained model without fine-tuning, there is no proper answer. It depends on how the model was trained. | |
| Jul 12, 2022 at 21:12 | comment | added | canP | Thanks for your answer. What do you mean by anyway? I can use CLS token or Mean of token outputs as embedding vector even in the pre-trained BERT model. Am i wrong? | |
| Jul 2, 2020 at 21:42 | history | answered | roman | CC BY-SA 4.0 |