3 Adding sentiment analysis capability for Kaiaulu to use #4
Conversation
Added the following functions to api/model.py: - train_model_bert - train_model_xlnet - train_model_roberta - train_model_albert - predict_bert - predict_xlnet - predict_roberta - predict_albert Also added exec/train_or_predict.py for external syscall use.
|
@geraldmjhuff I just noticed this new PR. Why was this not added to #2 ? |
| print(classification_report(val_labels, val_predictions, digits=4)) | ||
| return model_save_path | ||
|
|
||
| def train_model(train_df, model_save_path, model_select=0): |
There was a problem hiding this comment.
@carlosparadis
On top of dropping the 4 predict functions, then creating
def predict_model(predict_df, model_saved_path, model_task, model_tokenizer, model_name)
I will drop the 4 train functions, and create a function called:
train_model(train_df, model_save_path, model_task, model_tokenizer, model_name)
train_model() currently exists as this is Haotian's train model. What should our train_model() name be?
I will also commit the changes to Haotians PR, and in that model.py, Haotians original functions are preserved (or did you want to delete those functions?)
train_model()
test_model()
It will include my functions
train_model() (Again we can change this name if necessary)
predict_model()
There was a problem hiding this comment.
I recommend you comment out his train and other functions (unless it is used somewhere on the code that will break, then let me know).
In a subsequent commit I will remove the commented code so it at least stays versioned on master if we need to refer back.
No description provided.