codesum readme

Author: code-desc

CodeSummarization/README.md

@@ -0,0 +1,17 @@
+## Dataset link:
+
+Data source: https://drive.google.com/file/d/14nHVljNMb37-tpOW59NaDY26T6z2BcXD/view?usp=sharing
+
+## training procedure
+1. Download the data source from above link. Then extract the zip file inside data/java.
+2. Start training from scripts/java/transformer.sh  
+run this command 
+```bash
+ transformer.sh 0 CoDescSum
+ ```
+3. After two epoch terminate the terminal.
+4. Rename 'train' folder to 'pretrain' and rename 'finetune' to 'train'
+5. Run this again
+```bash
+ transformer.sh 0 CoDescSum
+ ```

CodeSummarization/data/README.md

@@ -109,7 +109,7 @@ Then the following commands will train and test code search networks:
  ```
 
 # Code Summarization
-We used the original implementation of Code Summarization of [NeuralCodeSum](https://github.com/wasiahmad/NeuralCodeSum.git). Please refer to [this guide](https://github.com/code-desc/CoDesc/blob/master/CodeSummarization/data/README.md) for instructions on how to train the code summarization network.
+We used the original implementation of Code Summarization of [NeuralCodeSum](https://github.com/wasiahmad/NeuralCodeSum.git). Please refer to [this guide](https://github.com/code-desc/CoDesc/blob/master/CodeSummarization/README.md) for instructions on how to train the code summarization network.
 
 
 # Licenses

README.md

@@ -13,6 +13,10 @@ pip install numpy
 pip install jsonlines
 pip install javalang
 pip install transformers
+pip install tqdm
+pip install nltk
+pip install prettytable
+pip install torch==1.5.1
 
 # download dataset
 mkdir data/