diff --git a/project/generated_music/readme.md b/project/generated_music/readme.md
new file mode 100644
index 0000000..e738105
--- /dev/null
+++ b/project/generated_music/readme.md
@@ -0,0 +1 @@
+In this folder generated music will be stored
diff --git a/project/midi_packs/readme.md b/project/midi_packs/readme.md
new file mode 100644
index 0000000..07f3c7d
--- /dev/null
+++ b/project/midi_packs/readme.md
@@ -0,0 +1,2 @@
+Place in this directory folders with midi.
+Each folder represent a source for making trainig sets
diff --git a/project/models/readme.md b/project/models/readme.md
new file mode 100644
index 0000000..f7c9357
--- /dev/null
+++ b/project/models/readme.md
@@ -0,0 +1 @@
+In this directory trained models will be stored
diff --git a/project/train.py b/project/train.py
index 2f70fde..b2fc9e6 100644
--- a/project/train.py
+++ b/project/train.py
@@ -4,6 +4,7 @@ import pickle
 import keras
 import argparse
 import warnings
+import pandas as pd
 from model import Seq2SeqModel
 from extract import make_folder_if_not_exist
 
@@ -42,6 +43,7 @@ def train_models(model_workflow):
         if not INSTRUMENT or INSTRUMENT == instrument:
             data_path = os.path.join('training_sets', EXPERIMENT_NAME, instrument.lower() + '_data.pkl')
             model_path = os.path.join('models', EXPERIMENT_NAME, f'{instrument.lower()}_model.h5')
+            history_path = os.path.join('models', EXPERIMENT_NAME, f'{instrument.lower()}_history.csv')
 
             x_train, y_train, _, bars_in_seq = pickle.load(open(data_path,'rb'))
             
@@ -52,8 +54,9 @@ def train_models(model_workflow):
                 model = Seq2SeqModel(x_train, y_train, LATENT_DIM, ENCODER_DROPOUT, DECODER_DROPOUT, bars_in_seq)
 
             print(f'Training: {instrument}')
-            model.fit(BATCH_SIZE, EPOCHS, callbacks=[])
+            history = model.fit(BATCH_SIZE, EPOCHS, callbacks=[])
             make_folder_if_not_exist(os.path.join('models', EXPERIMENT_NAME))
+            pd.DataFrame(history.history).to_csv(history_path, mode='a', header=False)
             model.save(model_path)
             found = True
 
@@ -88,4 +91,4 @@ if __name__ == '__main__':
     if not DECODER_DROPOUT:
         DECODER_DROPOUT = 0.0
     
-    train_models(load_workflow())
\ No newline at end of file
+    train_models(load_workflow())
diff --git a/project/training_sets/readme.md b/project/training_sets/readme.md
new file mode 100644
index 0000000..321fe89
--- /dev/null
+++ b/project/training_sets/readme.md
@@ -0,0 +1 @@
+In this directory generated training sets will be stored