search

Kyfafyd / MirrorGAN

Reproduction of the paper MirrorGAN: Learning Text-to-image Generation by Redescription
MIT License
25 stars 4 forks source link

size mismatch for embed.weight #5

Closed WhereISmyroad closed 4 years ago

WhereISmyroad commented 4 years ago

Hello sir Do you have bird/STREAM pretrained models? 【cnn/rnn_encoder.ckpt with torch.Size([5450, 512])】, I have some difficult in training the image caption model. Thank you very much

Kyfafyd commented 4 years ago

this problem should be solved by pretraining STREAM model at first.