CapDec About using the Sentcap dataset and setting parameters for train.py

Dear author, I would like to use this code on the Senticap dataset. May I ask for the parameter settings for train.py, such as the number of epochs, the backbone of CLIP, noise_ variance et al

Apr 27 '23 02:04 Rouchestand

Thank you for your interest in our work. Generally, I recommend using the default hyperparameters as a starting point, which are the values defined in the arg-parser. The number of epochs may depend on the size of the dataset, so I suggest training for approximately 20 epochs, but then choose the weights from the epoch where the test error begins to increase.

Apr 27 '23 06:04 DavidHuji

Thank you very much for your reply.

Apr 27 '23 06:04 Rouchestand

Thank you very much for your reply.

Hello, I also need to use the Senticap dataset, but I don't know how to use it. I would like to ask for your advice. Thank you

May 05 '23 02:05 chun178

You can use the pre trained weights provided by the author, and then use the SentiCap dataset for fine-tuning.

May 05 '23 08:05 Rouchestand

You can use the pre trained weights provided by the author, and then use the SentiCap dataset for fine-tuning.

OK,thank you very much

May 10 '23 02:05 chun178