search

njucckevin / ADS-Cap

A Framework for Accurate and Diverse Stylized Captioning with Unpaired Stylistic Corpora
9 stars 2 forks source link

Discriminator #2

Open Butterfies opened 1 year ago

Butterfies commented 1 year ago

Hi, @njucckevin : In config.py parser.add_argument('--recheck_model_path', default='./models/Discriminator/model_3800.pt') the discriminator model_3800.pt could not be found. Do I need to run train_recheck.py to get the discriminator model ?

njucckevin commented 1 year ago

Hi, I just add the model_3800.pt in models/Discriminator/model_3800.pt. You can try to see if it works.

njucckevin commented 1 year ago

It is just a style classification model based on LSTM.

gWeiXP commented 11 months ago

It is just a style classification model based on LSTM.

Hi, Can you provide a set of trained model parameters, I trained the model with your code, acc is only over 70%. Then I remembered that I had an error when I was running this code, and it made me realize that my training process may be different from yours image

njucckevin commented 11 months ago

@wxpqq826615304 It seems that the embedding layer size is mismatched with the checkpoint. Have you check the vocab_size of your model? Alternatively, I upload a new vocab.pkl file in /data/vocab.pkl. And I also upload the trained model checkpoint of cvae_k0.03_s1.0_ft in https://box.nju.edu.cn/d/bff4e27f2291466a860a/. You can have a try.