aspire

/

acge_text_embedding

like 150

Hey! I use this model in transformers AutoModelForSequenceClassification. By default, it uses Bertconfig, and uses the cls token for sentence embedding. In the config.json file, I find "pooler_type": "first_token_transform", while in ./1_Pooling/config.json, "pooling_mode_mean_tokens": true, it seems like you use mean pooler. So what is the final pooler type? Thank you!