Adding functionality to save the best model

Author: DivyanshiSrivastava

trainNN/train.py

@@ -108,9 +108,10 @@ def train_bichrom(data_paths, outdir, seq_len, bin_size):
     probas_out_sc = outdir + '/msc/' + 'test_probs.txt'
     records_file_path = outdir + '/metrics'
     print(records_file_path)
+    # save the best msc model
+    msc.save(outdir + 'full_model.best.hdf5')
 
     evaluate_models(sequence_len=seq_len, path=data_paths['test'],
                     probas_out_seq=probas_out_seq, probas_out_sc=probas_out_sc,
                     model_seq=mseq, model_sc=msc,
                     records_file_path=records_file_path)
-

trainNN/train_sc.py

@@ -1,15 +1,15 @@
-from __future__ import division
 import numpy as np
 import pandas as pd
+import tensorflow as tf
 
 from sklearn.metrics import average_precision_score as auprc
 from tensorflow.keras.models import Model
 from tensorflow.keras.layers import Dense, concatenate, Input, LSTM
 from tensorflow.keras.layers import Conv1D, Reshape, Lambda
 from tensorflow.keras.optimizers import SGD
-import tensorflow.keras.backend as K
 from tensorflow.keras.callbacks import Callback
 from tensorflow.keras.callbacks import ModelCheckpoint
+import tensorflow.keras.backend as K
 
 from iterutils import train_generator
 
@@ -135,7 +135,7 @@ def transfer(train_path, val_path, basemodel, model, steps_per_epoch,
     checkpointer = ModelCheckpoint(records_path + 'model_epoch{epoch}.hdf5',
                                    verbose=1, save_best_only=False)
 
-    hist = model.fit_generator(epochs=15, steps_per_epoch=steps_per_epoch,
+    hist = model.fit_generator(epochs=1, steps_per_epoch=steps_per_epoch,
                                generator=train_data_generator,
                                validation_data=validation_data,
                                callbacks=[precision_recall_history,

trainNN/train_seq.py

@@ -104,7 +104,7 @@ def train(model, train_path, val_path, steps_per_epoch, batch_size,
     # earlystop = EarlyStopping(monitor='val_loss', mode='min', verbose=1,
     #                           patience=5)
     # training the model..
-    hist = model.fit_generator(epochs=15, steps_per_epoch=steps_per_epoch,
+    hist = model.fit_generator(epochs=1, steps_per_epoch=steps_per_epoch,
                                generator=train_generator,
                                validation_data=validation_data,
                                callbacks=[precision_recall_history,