Update README.md

README.md

@@ -95,7 +95,7 @@ Using the data with additional tokens, I generated two datasets for Rachel in En
 !["raw dialogs"](./images/replicas.png)
 
 # Training
-For training the model to transfer style of Rachel to chatbot I used several models.
+For training the model to transfer style of Rachel to chatbot I used several models. Models are trained in two stages. First stage, when the model tries to catch the personal of Rachel and learns her monologues. Second step the model tries to learn how the Rachel behaves in dialoges, so at this stage model is trained on dialogues.
 1. First one is GPT2. For datasets I used TextDataset by PyTorch and library transformers by huggingface.
    The results are shown on the image below
    !["gpt2-results"](./images/gpt2-results.png)
@@ -113,4 +113,16 @@ For training the model to transfer style of Rachel to chatbot I used several mod
    !["gpt2-large-mono-train"](./images/gpt2-large-replics-train.png)
    Result of the training is shown below
    !["gpt2-large-results"](./images/gpt2-large-results.png)
+
+# Architecture
+
+- PrepareData.ipynb <- Parser data from the web, clean, tokenize and prepare for datasets
+- train_data <- datasets folder with monologues and dialogs
+- Training_gpt2_medium.ipynb <- training gpt2-medium
+- en_gpt2-medium_rachel_replics <- gpt2-medium model
+- Training_gpt2_large.ipynb <- training gpt2-large
+- en_gpt2-large_rachel_replics <- gpt2-large model
+- images <- images for README.md
+- app.py <- main file
+- requirements.txt <- needed libraries