-
-
Notifications
You must be signed in to change notification settings - Fork 439
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Release v0.1.9 #23
Release v0.1.9 #23
Conversation
modified and add some models
Develop Up-to-date
I reproduced the dim mismatch error during saved model loading. |
After one day investigation, I tend to suspect this would be caused by an unknown bug in Keras during loading a model with custom layer. As long as I alternate the At least two directions could be explored:
Some references: keras-team/keras#8612 |
I have tried the |
You tried it? It works? |
I have tried this and this works fine, I think this means we could build the model first then load the saved weight later. model = KMaxCNNModel()
model.fit(x, y, epochs=2)
model.save('model')
# works fine
model.model.load_weights('model/model.model')
# Error
KMaxCNNModel.load_model('model') |
Develop - Improve `KMaxPooling` layer
Great! |
Your test may be not the exact situation. |
@alexwwang You are right, let me try a new model then load the weights to the new one. |
…drop layer to some models
I tried, the problem is not gone. :( |
Develop - bug fix and model improve
Ok, let's release this version without the |
Well, but it's really annoying. By the way, do you have any idea on supporting customized layer definition and registration to this scheme? Just like what you've done on customized embedding layers. |
I've got an idea. |
@alexwwang yes, maybe this will work. |
I've solved it! |
…pairs in hyper_parameters dict to be more flexible to config at model initialization
Develop - Fix shape mismatch bug in KMaxPooling, improve save load robust and config flexibility
add sequence length check at sequence labeling model
add
AVCNNModel
,KMaxCNNModel
,RCNNModel
,AVRNNModel
,DropoutBGRUModel
,DropoutAVRNNModel
model to classification task.