Question: Does AutoModelForSequenceClassification work directly for SimCSE model

[vitthal-bhandari]

I have trained a model using @nreimers unsupervised implementation of SimCSE and saved it to the hub.

Now I wish to fine-tune the model on a particular dataset and get predictions.

I follow the implementation specified by Lewis Tunstall in his book -

1. Use AutoTokenizer to get the tokenizer
2. Use AutoModelForSequenceClassification to get the model from the hub
3. Fine-tune it using a Trainer object instance

Question: In the model card for your implementation, you use pooling after getting embeddings using AutoModel.

However, if I have to use the model as a black box using AutoModelForSequenceClassification, can I use it directly, or do I have to manually pool the outputs somewhere as with using AutoModel?