Save the weights after each epoch

leondgarse / Keras_insightface

Insightface Keras implementation

MIT License

230 stars 56 forks source link

Save the weights after each epoch #102

Closed Abdelfatah1998 closed 1 year ago

Abdelfatah1998 commented 1 year ago

Hi, How can I edit the callbacks so that I can also save the weights after each epoch in separate folder or with different file name (so that I can have the weights for each epoch after the training finish)?

Example: Epoch1weigths.h5 Epoch2weights.h5 .. etc

leondgarse commented 1 year ago

For saving model you can set filepath for myCallbacks.py#L261 like introduced in tf.keras.callbacks.ModelCheckpoint:

- model_checkpoint = ModelCheckpoint(checkpoint, verbose=1)
+ model_checkpoint = ModelCheckpoint(os.path.splitext(checkpoint)[0] + "_epoch{epoch:02d}.h5", verbose=1)

For saving basic_model, you can comment off the os.remove line in evals.py#L131-L132.

Abdelfatah1998 commented 1 year ago

Noted with thanks! Appreciate it.