Classification head missing
#2
by
AlexandrosChariton
- opened
Model card implies it classifies text to positive - neutral - negative, but running
output = model(input_ids, attention_mask)
after preprocessing and tokenizing returns an object with the last hidden state and the last hidden state after pooling. I can't find the sentiment analysis classification head.
>>> output['last_hidden_state'].shape
torch.Size([1, 8, 768])
>>> output['pooler_output'].shape
torch.Size([1, 768])
Am I missing something?