Resume training from pre-trained model

[Takuya Uehara]

Hi All,

I'm trying to resume training from a pre-trained model, but each time it starts from the beginning and doesn't utilize the pre-trained model. Could you guide me on how to correct this? Here are my settings. Is there any setting that I'm missing? I tried specifying the full path to the pre-trained model, but it didn't seem to work.

• Moved pre-trained model to "project_dir/model/pretrained_model/VAME_epoch_45.pkl"
• Modified config.yaml as follows

# RNN model general hyperparameter:
pretrained_model: "VAME_epoch_45.pkl"
pretrained_weights: true

...
# RNN loss hyperparameter:
mse_reconstruction_reduction: sum
mse_prediction_reduction: sum
kmeans_loss: 30
kmeans_lambda: 0.1
anneal_function: linear
kl_start: 0
annealtime: 1

Thanks!