Update README.md

README.md

@@ -7,7 +7,7 @@ Our vision is to empower both industrial application and academic research on en
 
 All of our models are implemented in Tensorflow>=2.0.0.
 
-## Table of Contents
+## 1) Table of Contents
 
 - [Athena](#athena)
   - [Table of Contents](#table-of-contents)
@@ -27,15 +27,15 @@ All of our models are implemented in Tensorflow>=2.0.0.
   - [Results](#results)
   - [Directory Structure](#directory-structure)
 
-## Key Features
+## 2) Key Features
 
 - Hybrid CTC/Transformer based end-to-end ASR
 - Speech-Transformer
 - Unsupervised pretraining
 
-## Installation
+## 3) Installation
 
-### 1) Creating a virtual environment [Optional]
+### 3.1) Creating a virtual environment [Optional]
 
 This project has only been tested on Python 3. We highly recommend creating a virtual environment and installing the python requirements there.
 
@@ -45,7 +45,7 @@ python -m venv venv_athena
 source venv_athena/bin/activate
 ```
 
-### 2) Install *tensorflow* backend
+### 3.2) Install *tensorflow* backend
 
 For more information, you can checkout the [tensorflow website](https://github.com/tensorflow/tensorflow)
 
@@ -55,12 +55,12 @@ pip install --upgrade pip
 pip install tensorflow==2.0.0
 ```
 
-### 3) Install *horovod* for multiple-device training [Optional]
+### 3.3) Install *horovod* for multiple-device training [Optional]
 
 For multiple GPU/CPU training
 You have to install the *horovod*, you can find out more information from the [horovod website](https://github.com/horovod/horovod#install)
 
-### 4) Install *athena* package
+### 3.4) Install *athena* package
 
 ```bash
 git clone https://github.com/athena-team/athena.git
@@ -73,7 +73,7 @@ python -m pip install --ignore-installed dist/athena-0.1.0*.whl
 - Once athena is successfully installed , you should do `source tools/env.sh` firstly before doing other things.
 - For installing some other supporting tools, you can check the `tools/install*.sh` to install kenlm, sph2pipe, spm and ... [Optional]
 
-### 5) Test your installation
+### 3.5) Test your installation
 
 - On a single cpu/gpu
 
@@ -96,9 +96,9 @@ horovodrun -np 4 -H localhost:4 athena/horovod_main.py examples/translate/spa-en
 - If you see errors such as `ERROR: Cannot uninstall 'wrapt'` while installing TensorFlow, try updating it using command `conda update wrapt`. Same for similar dependencies such as `entrypoints`, `llvmlite` and so on.
 - You may want to make sure you have `g++` version 7 or above to make sure you can successfully install TensorFlow.
 
-## Data Preparation
+## 4) Data Preparation
 
-### Create Manifest
+### 4.1) Create Manifest
 
 Athena accepts a textual manifest file as data set interface, which describes speech data set in csv format. In such file, each line contains necessary meta data (e.g. key, audio path, transcription) of a speech audio. For custom data, such manifest file needs to be prepared first. An example is shown as follows:
 
@@ -110,9 +110,9 @@ wav_filename	wav_length_ms	transcript
 /dataset/train-clean-100-wav/374-180298-0003.wav	356044	assumed all at once an appearance of noise and disorder never believe however disinterested the love of a kept woman may be that it will cost one nothing
 ```
 
-## Training
+## 5) Training
 
-### Setting the Configuration File
+### 5.1) Setting the Configuration File
 
 All of our training/ inference configurations are written in config.json. Below is an example configuration file with comments to help you understand.
 
@@ -178,7 +178,7 @@ All of our training/ inference configurations are written in config.json. Below
 }
 ```
 
-### Train a Model
+### 5.2) Train a Model
 
 With all the above preparation done, training becomes straight-forward. `athena/main.py` is the entry point of the training module. Just run `python athena/main.py <your_config_in_json_file>`
 
@@ -190,15 +190,15 @@ To run on a machine with 4 GPUs with Athena:
 To run on 4 machines with 4 GPUs each with Athena:
 `$ horovodrun -np 16 -H server1:4,server2:4,server3:4,server4:4 python athena/horovod_main.py <your_config_in_json_file>`
 
-## Results
+## 6) Results
 
 Language  | Model Name | Training Data | Hours of Speech | Error Rate
 :-----------: | :------------: | :----------: |  -------: | -------:
 English  | Transformer | [LibriSpeech Dataset](http://www.openslr.org/12/) | 960 h |
 Mandarin | Transformer | HKUST Dataset | 151 h | 22.75% (CER)
 Mandarin | Transformer | [AISHELL Dataset](http://www.openslr.org/33/) | 178 h | 5.77% (CER)
 
-## Directory Structure
+## 7) Directory Structure
 
 Below is the basic directory structure for Athena