Hello,
I know how to obtain the classification scores of the GPT2DoubleHeadsModel before the softmax layer:
model_gpt2 = GPT2DoubleHeadsModel.from_pretrained('gpt2-medium', output_hidden_states = True)
# retrieve the classification scores before the softmax
classification_scores = model_gpt2(input_ids=input_ids, mc_labels=mc_labels)[2]
How can I obtain the classification scores of the GPT2DoubleHeadsModel
after the softmax layer, along with the gradient information?
Thank you,