diff --git a/.travis.yml b/.travis.yml
index 75c2c1351..6ca50d954 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -1,7 +1,7 @@
 language: cpp
 cache: ccache
 sudo: required
-dist: trusty
+dist: xenial 
 services:
   - docker
 os:
@@ -13,11 +13,12 @@ addons:
   apt:
     packages:
       - git
-      - python
-      - python-pip
-      - python2.7-dev
+      - python3-pip
+      - python3-dev
 
 before_install:
+  -  python3 --version
+  -  python3 -m pip --version
   -  sudo pip install -U virtualenv pre-commit pip
   -  docker pull paddlepaddle/paddle:latest
 
diff --git a/.travis/unittest.sh b/.travis/unittest.sh
index f27dc481a..e42764368 100755
--- a/.travis/unittest.sh
+++ b/.travis/unittest.sh
@@ -17,7 +17,7 @@ unittest(){
     fi
     find . -name 'tests' -type d -print0 | \
         xargs -0 -I{} -n1 bash -c \
-        'python -m unittest discover -v -s {}'
+        'python3 -m unittest discover -v -s {}'
     cd - > /dev/null
 }
 
diff --git a/README.md b/README.md
index 3d8b9e600..521351c37 100644
--- a/README.md
+++ b/README.md
@@ -6,12 +6,12 @@
 
 ## Table of Contents
 - [Installation](#installation)
+- [Running in Docker Container](#running-in-docker-container)
 - [Getting Started](#getting-started)
 - [Data Preparation](#data-preparation)
 - [Training a Model](#training-a-model)
 - [Data Augmentation Pipeline](#data-augmentation-pipeline)
 - [Inference and Evaluation](#inference-and-evaluation)
-- [Running in Docker Container](#running-in-docker-container)
 - [Hyper-parameters Tuning](#hyper-parameters-tuning)
 - [Training for Mandarin Language](#training-for-mandarin-language)
 - [Trying Live Demo with Your Own Voice](#trying-live-demo-with-your-own-voice)
@@ -26,20 +26,20 @@
 To avoid the trouble of environment setup, [running in Docker container](#running-in-docker-container) is highly recommended. Otherwise follow the guidelines below to install the dependencies manually.
 
 ### Prerequisites
-- Python 2.7 only supported
+- Python >= 3.6
 - PaddlePaddle 1.8.0 or later (please refer to the [Installation Guide](https://www.paddlepaddle.org.cn/documentation/docs/en/beginners_guide/index_en.html))
 
 ### Setup
 - Make sure these libraries or tools installed: `pkg-config`, `flac`, `ogg`, `vorbis`, `boost` and `swig`, e.g. installing them via `apt-get`:
 
 ```bash
-sudo apt-get install -y pkg-config libflac-dev libogg-dev libvorbis-dev libboost-dev swig python-dev
+sudo apt-get install -y pkg-config libflac-dev libogg-dev libvorbis-dev libboost-dev swig python3-dev
 ```
 
 or, installing them via `yum`:
 
 ```bash
-sudo yum install pkgconfig libogg-devel libvorbis-devel boost-devel python-devel
+sudo yum install pkgconfig libogg-devel libvorbis-devel boost-devel python3-devel
 wget https://ftp.osuosl.org/pub/xiph/releases/flac/flac-1.3.1.tar.xz
 xz -d flac-1.3.1.tar.xz
 tar -xvf flac-1.3.1.tar
@@ -57,6 +57,39 @@ cd DeepSpeech
 sh setup.sh
 ```
 
+### Running in Docker Container
+
+Docker is an open source tool to build, ship, and run distributed applications in an isolated environment. A Docker image for this project has been provided in [hub.docker.com](https://hub.docker.com) with all the dependencies installed, including the pre-built PaddlePaddle, CTC decoders, and other necessary Python and third-party packages. This Docker image requires the support of NVIDIA GPU, so please make sure its availiability and the [nvidia-docker](https://github.com/NVIDIA/nvidia-docker) has been installed.
+
+Take several steps to launch the Docker image:
+
+- Download the Docker image
+
+```bash
+nvidia-docker pull hub.baidubce.com/paddlepaddle/deep_speech_fluid:latest-gpu
+```
+
+- Clone this repository
+
+```
+git clone https://github.com/PaddlePaddle/DeepSpeech.git
+```
+
+- Run the Docker image
+
+```bash
+sudo nvidia-docker run -it -v $(pwd)/DeepSpeech:/DeepSpeech hub.baidubce.com/paddlepaddle/deep_speech_fluid:latest-gpu /bin/bash
+```
+Now go back and start from the [Getting Started](#getting-started) section, you can execute training, inference and hyper-parameters tuning similarly in the Docker container.
+
+
+- Install PaddlePaddle
+
+For example, for CUDA 10.1, CuDNN7.5:
+```bash
+python3 -m pip install paddlepaddle-gpu==1.8.0.post107
+```
+
 ## Getting Started
 
 Several shell scripts provided in `./examples` will help us to quickly give it a try, for most major modules, including data preparation, model training, case inference and model evaluation, with a few public dataset (e.g. [LibriSpeech](http://www.openslr.org/12/), [Aishell](http://www.openslr.org/33)). Reading these examples will also help you to understand how to make it work with your own data.
@@ -132,7 +165,7 @@ For how to generate such manifest files, please refer to `data/librispeech/libri
 To perform z-score normalization (zero-mean, unit stddev) upon audio features, we have to estimate in advance the mean and standard deviation of the features, with some training samples:
 
 ```bash
-python tools/compute_mean_std.py \
+python3 tools/compute_mean_std.py \
 --num_samples 2000 \
 --specgram_type linear \
 --manifest_path data/librispeech/manifest.train \
@@ -147,7 +180,7 @@ It will compute the mean and standard deviatio of power spectrum feature with 20
 A vocabulary of possible characters is required to convert the transcription into a list of token indices for training, and in decoding, to convert from a list of indices back to text again. Such a character-based vocabulary can be built with `tools/build_vocab.py`.
 
 ```bash
-python tools/build_vocab.py \
+python3 tools/build_vocab.py \
 --count_threshold 0 \
 --vocab_path data/librispeech/eng_vocab.txt \
 --manifest_paths data/librispeech/manifest.train
@@ -160,9 +193,9 @@ It will write a vocabuary file `data/librispeeech/eng_vocab.txt` with all transc
 For more help on arguments:
 
 ```bash
-python data/librispeech/librispeech.py --help
-python tools/compute_mean_std.py --help
-python tools/build_vocab.py --help
+python3 data/librispeech/librispeech.py --help
+python3 tools/compute_mean_std.py --help
+python3 tools/build_vocab.py --help
 ```
 
 ## Training a model
@@ -172,26 +205,26 @@ python tools/build_vocab.py --help
 - Start training from scratch with 8 GPUs:
 
     ```
-    CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 python train.py
+    CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 python3 train.py
     ```
 
 - Start training from scratch with CPUs:
 
     ```
-    python train.py --use_gpu False
+    python3 train.py --use_gpu False
     ```
 - Resume training from a checkpoint:
 
     ```
     CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
-    python train.py \
+    python3 train.py \
     --init_from_pretrained_model CHECKPOINT_PATH_TO_RESUME_FROM
     ```
 
 For more help on arguments:
 
 ```bash
-python train.py --help
+python3 train.py --help
 ```
 or refer to `example/librispeech/run_train.sh`.
 
@@ -273,13 +306,13 @@ An inference module caller `infer.py` is provided to infer, decode and visualize
 - Inference with GPU:
 
     ```bash
-    CUDA_VISIBLE_DEVICES=0 python infer.py
+    CUDA_VISIBLE_DEVICES=0 python3 infer.py
     ```
 
 - Inference with CPUs:
 
     ```bash
-    python infer.py --use_gpu False
+    python3 infer.py --use_gpu False
     ```
 
 We provide two types of CTC decoders: *CTC greedy decoder* and *CTC beam search decoder*. The *CTC greedy decoder* is an implementation of the simple best-path decoding algorithm, selecting at each timestep the most likely token, thus being greedy and locally optimal. The [*CTC beam search decoder*](https://arxiv.org/abs/1408.2873) otherwise utilizes a heuristic breadth-first graph search for reaching a near global optimality; it also requires a pre-trained KenLM language model for better scoring and ranking. The decoder type can be set with argument `--decoding_method`.
@@ -287,7 +320,7 @@ We provide two types of CTC decoders: *CTC greedy decoder* and *CTC beam search
 For more help on arguments:
 
 ```
-python infer.py --help
+python3 infer.py --help
 ```
 or refer to `example/librispeech/run_infer.sh`.
 
@@ -298,13 +331,13 @@ To evaluate a model's performance quantitatively, please run:
 - Evaluation with GPUs:
 
     ```bash
-    CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 python test.py
+    CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 python3 test.py
     ```
 
 - Evaluation with CPUs:
 
     ```bash
-    python test.py --use_gpu False
+    python3 test.py --use_gpu False
     ```
 
 The error rate (default: word error rate; can be set with `--error_rate_type`) will be printed.
@@ -312,7 +345,7 @@ The error rate (default: word error rate; can be set with `--error_rate_type`) w
 For more help on arguments:
 
 ```bash
-python test.py --help
+python3 test.py --help
 ```
 or refer to `example/librispeech/run_test.sh`.
 
@@ -326,7 +359,7 @@ The hyper-parameters $\alpha$ (language model weight) and $\beta$ (word insertio
 
     ```bash
     CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
-    python tools/tune.py \
+    python3 tools/tune.py \
     --alpha_from 1.0 \
     --alpha_to 3.2 \
     --num_alphas 45 \
@@ -338,7 +371,7 @@ The hyper-parameters $\alpha$ (language model weight) and $\beta$ (word insertio
 - Tuning with CPU:
 
     ```bash
-    python tools/tune.py --use_gpu False
+    python3 tools/tune.py --use_gpu False
     ```
  The grid search will print the WER (word error rate) or CER (character error rate) at each point in the hyper-parameters space, and draw the error surface optionally. A proper hyper-parameters range should include the global minima of the error surface for WER/CER, as illustrated in the following figure.
 
@@ -352,36 +385,10 @@ Usually, as the figure shows, the variation of language model weight ($\alpha$)
 After tuning, you can reset $\alpha$ and $\beta$ in the inference and evaluation modules to see if they really help improve the ASR performance. For more help
 
 ```bash
-python tune.py --help
+python3 tune.py --help
 ```
 or refer to `example/librispeech/run_tune.sh`.
 
-## Running in Docker Container
-
-Docker is an open source tool to build, ship, and run distributed applications in an isolated environment. A Docker image for this project has been provided in [hub.docker.com](https://hub.docker.com) with all the dependencies installed, including the pre-built PaddlePaddle, CTC decoders, and other necessary Python and third-party packages. This Docker image requires the support of NVIDIA GPU, so please make sure its availiability and the [nvidia-docker](https://github.com/NVIDIA/nvidia-docker) has been installed.
-
-Take several steps to launch the Docker image:
-
-- Download the Docker image
-
-```bash
-nvidia-docker pull hub.baidubce.com/paddlepaddle/deep_speech_fluid:latest-gpu
-```
-
-- Clone this repository
-
-```
-git clone https://github.com/PaddlePaddle/DeepSpeech.git
-```
-
-- Run the Docker image
-
-```bash
-sudo nvidia-docker run -it -v $(pwd)/DeepSpeech:/DeepSpeech hub.baidubce.com/paddlepaddle/deep_speech_fluid:latest-gpu /bin/bash
-```
-Now go back and start from the [Getting Started](#getting-started) section, you can execute training, inference and hyper-parameters tuning similarly in the Docker container.
-
-
 ## Training for Mandarin Language
 
 The key steps of training for Mandarin language are same to that of English language and we have also provided an example for Mandarin training with Aishell in ```examples/aishell```. As mentioned above, please execute ```sh run_data.sh```, ```sh run_train.sh```, ```sh run_test.sh``` and ```sh run_infer.sh``` to do data preparation, training, testing and inference correspondingly. We have also prepared a pre-trained model (downloaded by ./models/aishell/download_model.sh) for users to try with ```sh run_infer_golden.sh``` and ```sh run_test_golden.sh```. Notice that, different from English LM, the Mandarin LM is character-based and please run ```tools/tune.py``` to find an optimal setting.
@@ -394,7 +401,7 @@ To start the demo's server, please run this in one console:
 
 ```bash
 CUDA_VISIBLE_DEVICES=0 \
-python deploy/demo_server.py \
+python3 deploy/demo_server.py \
 --host_ip localhost \
 --host_port 8086
 ```
@@ -413,7 +420,7 @@ Then to start the client, please run this in another console:
 
 ```bash
 CUDA_VISIBLE_DEVICES=0 \
-python -u deploy/demo_client.py \
+python3 -u deploy/demo_client.py \
 --host_ip 'localhost' \
 --host_port 8086
 ```
@@ -427,8 +434,8 @@ Please also refer to `examples/deploy_demo/run_english_demo_server.sh`, which wi
 For more help on arguments:
 
 ```bash
-python deploy/demo_server.py --help
-python deploy/demo_client.py --help
+python3 deploy/demo_server.py --help
+python3 deploy/demo_client.py --help
 ```
 
 ## Released Models
diff --git a/README_cn.md b/README_cn.md
index 03243c01d..176f0cd77 100644
--- a/README_cn.md
+++ b/README_cn.md
@@ -7,12 +7,12 @@
 
 ## 目录
 - [安装](#安装)
+- [在 Docker 容器上运行](#在Docker容器上运行)
 - [开始](#开始)
 - [数据准备](#数据准备)
 - [训练模型](#训练模型)
 - [数据增强流水线](#数据增强流水线)
 - [推断和评价](#推断和评价)
-- [在 Docker 容器上运行](#在Docker容器上运行)
 - [超参数调整](#超参数调整)
 - [训练汉语语言](#训练汉语语言)
 - [用自己的声音尝试现场演示](#用自己的声音尝试现场演示)
@@ -24,20 +24,20 @@
 为了避免环境配置问题，强烈建议在[Docker容器上运行](#在Docker容器上运行)，否则请按照下面的指南安装依赖项。
 
 ### 前提
-- 只支持Python 2.7
+- Python >= 3.6
 - PaddlePaddle 1.8.0 版本及以上（请参考[安装指南](https://www.paddlepaddle.org.cn/install/quick)）
 
 ### 安装
 - 请确保以下库或工具已安装完毕：`pkg-config`, `flac`, `ogg`, `vorbis`, `boost` 和 `swig`, 如可以通过`apt-get`安装：
 
 ```bash
-sudo apt-get install -y pkg-config libflac-dev libogg-dev libvorbis-dev libboost-dev swig python-dev
+sudo apt-get install -y pkg-config libflac-dev libogg-dev libvorbis-dev libboost-dev swig python3-dev
 ```
 
 或者，也可以通过`yum`安装：
 
 ```bash
-sudo yum install pkgconfig libogg-devel libvorbis-devel boost-devel python-devel
+sudo yum install pkgconfig libogg-devel libvorbis-devel boost-devel python3-devel
 wget https://ftp.osuosl.org/pub/xiph/releases/flac/flac-1.3.1.tar.xz
 xz -d flac-1.3.1.tar.xz
 tar -xvf flac-1.3.1.tar
@@ -55,6 +55,39 @@ cd DeepSpeech
 sh setup.sh
 ```
 
+### 在Docker容器上运行
+
+Docker 是一个开源工具，用于在孤立的环境中构建、发布和运行分布式应用程序。此项目的 Docker 镜像已在[hub.docker.com](https://hub.docker.com)中提供，并安装了所有依赖项，其中包括预先构建的PaddlePaddle，CTC解码器以及其他必要的 Python 和第三方库。这个 Docker 映像需要NVIDIA GPU的支持，所以请确保它的可用性并已完成[nvidia-docker](https://github.com/NVIDIA/nvidia-docker)的安装。
+
+采取以下步骤来启动 Docker 镜像：
+
+- 下载 Docker 镜像
+
+```bash
+nvidia-docker pull hub.baidubce.com/paddlepaddle/deep_speech_fluid:latest-gpu
+```
+
+- git clone 这个资源库
+
+```
+git clone https://github.com/PaddlePaddle/DeepSpeech.git
+```
+
+- 运行 Docker 镜像
+
+```bash
+sudo nvidia-docker run -it -v $(pwd)/DeepSpeech:/DeepSpeech hub.baidubce.com/paddlepaddle/deep_speech_fluid:latest-gpu /bin/bash
+```
+
+现在返回并从[开始](#开始)部分开始，您可以在Docker容器中同样执行模型训练，推断和超参数调整。
+
+- 安装 PaddlePaddle
+
+例如 CUDA 10.1, CuDNN7.5:
+```bash
+python3 -m pip install paddlepaddle-gpu==1.8.0.post107
+```
+
 ## 开始
 
 `./examples`里的一些 shell 脚本将帮助我们在一些公开数据集(比如：[LibriSpeech](http://www.openslr.org/12/), [Aishell](http://www.openslr.org/33)) 进行快速尝试，包括了数据准备，模型训练，案例推断和模型评价。阅读这些例子将帮助你理解如何使用你的数据集训练模型。
@@ -130,7 +163,7 @@ sh setup.sh
 为了对音频特征进行 z-score 归一化（零均值，单位标准差），我们必须预估训练样本特征的均值和标准差：
 
 ```bash
-python tools/compute_mean_std.py \
+python3 tools/compute_mean_std.py \
 --num_samples 2000 \
 --specgram_type linear \
 --manifest_path data/librispeech/manifest.train \
@@ -144,7 +177,7 @@ python tools/compute_mean_std.py \
 我们需要一个包含可能会出现的字符集合的词表来在训练的时候将字符转换成索引，并在解码的时候将索引转换回文本。`tools/build_vocab.py`脚本将生成这种基于字符的词表。
 
 ```bash
-python tools/build_vocab.py \
+python3 tools/build_vocab.py \
 --count_threshold 0 \
 --vocab_path data/librispeech/eng_vocab.txt \
 --manifest_paths data/librispeech/manifest.train
@@ -157,9 +190,9 @@ python tools/build_vocab.py \
 获得更多帮助：
 
 ```bash
-python data/librispeech/librispeech.py --help
-python tools/compute_mean_std.py --help
-python tools/build_vocab.py --help
+python3 data/librispeech/librispeech.py --help
+python3 tools/compute_mean_std.py --help
+python3 tools/build_vocab.py --help
 ```
 
 ## 训练模型
@@ -169,27 +202,27 @@ python tools/build_vocab.py --help
 - 开始使用 8 片 GPU 训练：
 
     ```
-    CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 python train.py
+    CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 python3 train.py
     ```
 
 - 开始使用 CPU 训练：
 
     ```
-    python train.py --use_gpu False
+    python3 train.py --use_gpu False
     ```
 
 - 从检查点恢复训练：
 
     ```
     CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
-    python train.py \
+    python3 train.py \
     --init_from_pretrained_model CHECKPOINT_PATH_TO_RESUME_FROM
     ```
 
 获得更多帮助：
 
 ```bash
-python train.py --help
+python3 train.py --help
 ```
 或参考 `example/librispeech/run_train.sh`.
 
@@ -272,13 +305,13 @@ bash download_lm_ch.sh
 - GPU 版本的推断：
 
     ```bash
-    CUDA_VISIBLE_DEVICES=0 python infer.py
+    CUDA_VISIBLE_DEVICES=0 python3 infer.py
     ```
 
 - CPU 版本的推断：
 
     ```bash
-    python infer.py --use_gpu False
+    python3 infer.py --use_gpu False
     ```
 
 我们提供两种类型的 CTC 解码器：*CTC贪心解码器*和*CTC波束搜索解码器*。*CTC贪心解码器*是简单的最佳路径解码算法的实现，在每个时间步选择最可能的字符，因此是贪心的并且是局部最优的。[*CTC波束搜索解码器*](https://arxiv.org/abs/1408.2873)另外使用了启发式广度优先图搜索以达到近似全局最优; 它也需要预先训练的KenLM语言模型以获得更好的评分和排名。解码器类型可以用参数`--decoding_method`设置。
@@ -286,7 +319,7 @@ bash download_lm_ch.sh
 获得更多帮助：
 
 ```
-python infer.py --help
+python3 infer.py --help
 ```
 或参考`example/librispeech/run_infer.sh`.
 
@@ -297,13 +330,13 @@ python infer.py --help
 - GPU 版本评估
 
     ```bash
-    CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 python test.py
+    CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 python3 test.py
     ```
 
 - CPU 版本评估
 
     ```bash
-    python test.py --use_gpu False
+    python3 test.py --use_gpu False
     ```
 
 错误率（默认：误字率;可以用--error_rate_type设置）将被打印出来。
@@ -311,7 +344,7 @@ python infer.py --help
 获得更多帮助：
 
 ```bash
-python test.py --help
+python3 test.py --help
 ```
 或参考`example/librispeech/run_test.sh`.
 
@@ -325,7 +358,7 @@ python test.py --help
 
     ```bash
     CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
-    python tools/tune.py \
+    python3 tools/tune.py \
     --alpha_from 1.0 \
     --alpha_to 3.2 \
     --num_alphas 45 \
@@ -337,7 +370,7 @@ python test.py --help
 - CPU 版的调整：
 
     ```bash
-    python tools/tune.py --use_gpu False
+    python3 tools/tune.py --use_gpu False
     ```
 网格搜索将会在超参数空间的每个点处打印出 WER (误字率)或者 CER (字符错误率)，并且可绘出误差曲面。一个合适的超参数范围应包括 WER/CER 误差表面的全局最小值，如下图所示。
 
@@ -351,37 +384,10 @@ python test.py --help
 调整之后，您可以在推理和评价模块中重置$\alpha$和$\beta$，以检查它们是否真的有助于提高 ASR 性能。更多帮助如下：
 
 ```bash
-python tune.py --help
+python3 tune.py --help
 ```
 或参考`example/librispeech/run_tune.sh`.
 
-## 在Docker容器上运行
-
-Docker 是一个开源工具，用于在孤立的环境中构建、发布和运行分布式应用程序。此项目的 Docker 镜像已在[hub.docker.com](https://hub.docker.com)中提供，并安装了所有依赖项，其中包括预先构建的PaddlePaddle，CTC解码器以及其他必要的 Python 和第三方库。这个 Docker 映像需要NVIDIA GPU的支持，所以请确保它的可用性并已完成[nvidia-docker](https://github.com/NVIDIA/nvidia-docker)的安装。
-
-采取以下步骤来启动 Docker 镜像：
-
-- 下载 Docker 镜像
-
-```bash
-nvidia-docker pull hub.baidubce.com/paddlepaddle/deep_speech_fluid:latest-gpu
-```
-
-- git clone 这个资源库
-
-```
-git clone https://github.com/PaddlePaddle/DeepSpeech.git
-```
-
-- 运行 Docker 镜像
-
-```bash
-sudo nvidia-docker run -it -v $(pwd)/DeepSpeech:/DeepSpeech hub.baidubce.com/paddlepaddle/deep_speech_fluid:latest-gpu /bin/bash
-```
-
-现在返回并从[开始](#开始)部分开始，您可以在Docker容器中同样执行模型训练，推断和超参数调整。
-
-
 ## 训练普通话语言
 
 普通话语言训练与英语训练的关键步骤相同，我们提供了一个使用 Aishell 进行普通话训练的例子```examples/aishell```。如上所述，请执行```sh run_data.sh```, ```sh run_train.sh```, ```sh run_test.sh```和```sh run_infer.sh```做相应的数据准备，训练，测试和推断。我们还准备了一个预训练过的模型（执行./models/aishell/download_model.sh下载）供用户使用```run_infer_golden.sh```和```run_test_golden.sh```来。请注意，与英语语言模型不同，普通话语言模型是基于汉字的，请运行```tools/tune.py```来查找最佳设置。
@@ -394,7 +400,7 @@ sudo nvidia-docker run -it -v $(pwd)/DeepSpeech:/DeepSpeech hub.baidubce.com/pad
 
 ```bash
 CUDA_VISIBLE_DEVICES=0 \
-python deploy/demo_server.py \
+python3 deploy/demo_server.py \
 --host_ip localhost \
 --host_port 8086
 ```
@@ -413,7 +419,7 @@ pip install keyboard
 
 ```bash
 CUDA_VISIBLE_DEVICES=0 \
-python -u deploy/demo_client.py \
+python3 -u deploy/demo_client.py \
 --host_ip 'localhost' \
 --host_port 8086
 ```
@@ -427,8 +433,8 @@ python -u deploy/demo_client.py \
 获得更多帮助：
 
 ```bash
-python deploy/demo_server.py --help
-python deploy/demo_client.py --help
+python3 deploy/demo_server.py --help
+python3 deploy/demo_client.py --help
 ```
 
 ## 发布模型
diff --git a/data/aishell/aishell.py b/data/aishell/aishell.py
index e81d20853..6290712f2 100644
--- a/data/aishell/aishell.py
+++ b/data/aishell/aishell.py
@@ -5,9 +5,6 @@ Manifest file is a json-format file with each line containing the
 meta data (i.e. audio filepath, transcript and audio duration)
 of each audio file in the data set.
 """
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import os
 import codecs
diff --git a/data/librispeech/librispeech.py b/data/librispeech/librispeech.py
index 770b71108..8a136453b 100644
--- a/data/librispeech/librispeech.py
+++ b/data/librispeech/librispeech.py
@@ -5,9 +5,6 @@ Manifest file is a json-format file with each line containing the
 meta data (i.e. audio filepath, transcript and audio duration)
 of each audio file in the data set.
 """
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import distutils.util
 import os
diff --git a/data/noise/chime3_background.py b/data/noise/chime3_background.py
index 1aa7f8df8..30a2e14e7 100644
--- a/data/noise/chime3_background.py
+++ b/data/noise/chime3_background.py
@@ -5,9 +5,6 @@ Manifest file is a json-format file with each line containing the
 meta data (i.e. audio filepath, transcript and audio duration)
 of each audio file in the data set.
 """
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import distutils.util
 import os
diff --git a/data/voxforge/voxforge.py b/data/voxforge/voxforge.py
index b7cc38bdb..8478166fb 100644
--- a/data/voxforge/voxforge.py
+++ b/data/voxforge/voxforge.py
@@ -5,9 +5,6 @@ Manifest file is a json-format file with each line containing the
 meta data (i.e. audio filepath, transcript and audio duration)
 of each audio file in the data set.
 """
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import os
 import codecs
diff --git a/data_utils/audio.py b/data_utils/audio.py
index e0feb21f3..d7b88952d 100644
--- a/data_utils/audio.py
+++ b/data_utils/audio.py
@@ -1,7 +1,4 @@
 """Contains the audio segment class."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import numpy as np
 import io
@@ -62,11 +59,11 @@ class AudioSegment(object):
         """Create audio segment from audio file.
         
         :param filepath: Filepath or file object to audio file.
-        :type filepath: basestring|file
+        :type filepath: str|file
         :return: Audio segment instance.
         :rtype: AudioSegment
         """
-        if isinstance(file, basestring) and re.findall(r".seqbin_\d+$", file):
+        if isinstance(file, str) and re.findall(r".seqbin_\d+$", file):
             return cls.from_sequence_file(file)
         else:
             samples, sample_rate = soundfile.read(file, dtype='float32')
@@ -78,7 +75,7 @@ class AudioSegment(object):
         the entire file into the memory which can be incredibly wasteful.
 
         :param file: Input audio filepath or file object.
-        :type file: basestring|file
+        :type file: str|file
         :param start: Start time in seconds. If start is negative, it wraps
                       around from the end. If not provided, this function 
                       reads from the very beginning.
@@ -97,7 +94,7 @@ class AudioSegment(object):
         sample_rate = sndfile.samplerate
         duration = float(len(sndfile)) / sample_rate
         start = 0. if start is None else start
-        end = 0. if end is None else end
+        end = duration if end is None else end
         if start < 0.0:
             start += duration
         if end < 0.0:
@@ -143,7 +140,7 @@ class AudioSegment(object):
         sequence file (starting from 1).
 
         :param filepath: Filepath of sequence file.
-        :type filepath: basestring
+        :type filepath: str
         :return: Audio segment instance.
         :rtype: AudioSegment
         """
@@ -236,7 +233,7 @@ class AudioSegment(object):
         
         :param filepath: WAV filepath or file object to save the
                          audio segment.
-        :type filepath: basestring|file
+        :type filepath: str|file
         :param dtype: Subtype for audio file. Options: 'int16', 'int32',
                       'float32', 'float64'. Default is 'float32'.
         :type dtype: str
diff --git a/data_utils/augmentor/augmentation.py b/data_utils/augmentor/augmentation.py
index 5c30b627e..37d6bc1dd 100644
--- a/data_utils/augmentor/augmentation.py
+++ b/data_utils/augmentor/augmentation.py
@@ -1,7 +1,4 @@
 """Contains the data augmentation pipeline."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import json
 import random
diff --git a/data_utils/augmentor/base.py b/data_utils/augmentor/base.py
index a323165aa..adeda6465 100644
--- a/data_utils/augmentor/base.py
+++ b/data_utils/augmentor/base.py
@@ -1,7 +1,4 @@
 """Contains the abstract base class for augmentation models."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 from abc import ABCMeta, abstractmethod
 
diff --git a/data_utils/augmentor/impulse_response.py b/data_utils/augmentor/impulse_response.py
index 536b4d6a4..937e578bd 100644
--- a/data_utils/augmentor/impulse_response.py
+++ b/data_utils/augmentor/impulse_response.py
@@ -1,7 +1,4 @@
 """Contains the impulse response augmentation model."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 from data_utils.augmentor.base import AugmentorBase
 from data_utils.utility import read_manifest
@@ -14,7 +11,7 @@ class ImpulseResponseAugmentor(AugmentorBase):
     :param rng: Random generator object.
     :type rng: random.Random
     :param impulse_manifest_path: Manifest path for impulse audio data.
-    :type impulse_manifest_path: basestring
+    :type impulse_manifest_path: str
     """
 
     def __init__(self, rng, impulse_manifest_path):
diff --git a/data_utils/augmentor/noise_perturb.py b/data_utils/augmentor/noise_perturb.py
index 96e0ff4de..952ca3fb9 100644
--- a/data_utils/augmentor/noise_perturb.py
+++ b/data_utils/augmentor/noise_perturb.py
@@ -1,7 +1,4 @@
 """Contains the noise perturb augmentation model."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 from data_utils.augmentor.base import AugmentorBase
 from data_utils.utility import read_manifest
@@ -18,7 +15,7 @@ class NoisePerturbAugmentor(AugmentorBase):
     :param max_snr_dB: Maximal signal noise ratio, in decibels.
     :type max_snr_dB: float
     :param noise_manifest_path: Manifest path for noise audio data.
-    :type noise_manifest_path: basestring
+    :type noise_manifest_path: str
     """
 
     def __init__(self, rng, min_snr_dB, max_snr_dB, noise_manifest_path):
diff --git a/data_utils/augmentor/online_bayesian_normalization.py b/data_utils/augmentor/online_bayesian_normalization.py
index e488ac7d6..deba9a77a 100644
--- a/data_utils/augmentor/online_bayesian_normalization.py
+++ b/data_utils/augmentor/online_bayesian_normalization.py
@@ -1,7 +1,4 @@
 """Contain the online bayesian normalization augmentation model."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 from data_utils.augmentor.base import AugmentorBase
 
diff --git a/data_utils/augmentor/resample.py b/data_utils/augmentor/resample.py
index 8df17f3a8..e26d06b46 100644
--- a/data_utils/augmentor/resample.py
+++ b/data_utils/augmentor/resample.py
@@ -1,7 +1,4 @@
 """Contain the resample augmentation model."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 from data_utils.augmentor.base import AugmentorBase
 
diff --git a/data_utils/augmentor/shift_perturb.py b/data_utils/augmentor/shift_perturb.py
index c4cbe3e17..a9b732161 100644
--- a/data_utils/augmentor/shift_perturb.py
+++ b/data_utils/augmentor/shift_perturb.py
@@ -1,7 +1,4 @@
 """Contains the volume perturb augmentation model."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 from data_utils.augmentor.base import AugmentorBase
 
diff --git a/data_utils/augmentor/speed_perturb.py b/data_utils/augmentor/speed_perturb.py
index cc5738bd1..78dfd0c65 100644
--- a/data_utils/augmentor/speed_perturb.py
+++ b/data_utils/augmentor/speed_perturb.py
@@ -1,7 +1,4 @@
 """Contain the speech perturbation augmentation model."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 from data_utils.augmentor.base import AugmentorBase
 
diff --git a/data_utils/augmentor/volume_perturb.py b/data_utils/augmentor/volume_perturb.py
index 758676d55..b1d8c190b 100644
--- a/data_utils/augmentor/volume_perturb.py
+++ b/data_utils/augmentor/volume_perturb.py
@@ -1,7 +1,4 @@
 """Contains the volume perturb augmentation model."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 from data_utils.augmentor.base import AugmentorBase
 
diff --git a/data_utils/data.py b/data_utils/data.py
index 4a5224efa..ed65a0947 100644
--- a/data_utils/data.py
+++ b/data_utils/data.py
@@ -1,9 +1,6 @@
 """Contains data generator for orgnaizing various audio data preprocessing
 pipeline and offering data reader interface of PaddlePaddle requirements.
 """
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import random
 import tarfile
@@ -25,9 +22,9 @@ class DataGenerator(object):
 
     :param vocab_filepath: Vocabulary filepath for indexing tokenized
                            transcripts.
-    :type vocab_filepath: basestring
+    :type vocab_filepath: str
     :param mean_std_filepath: File containing the pre-computed mean and stddev.
-    :type mean_std_filepath: None|basestring
+    :type mean_std_filepath: None|str
     :param augmentation_config: Augmentation configuration in json string.
                                 Details see AugmentationPipeline.__doc__.
     :type augmentation_config: str
@@ -104,14 +101,14 @@ class DataGenerator(object):
         """Load, augment, featurize and normalize for speech data.
 
         :param audio_file: Filepath or file object of audio file.
-        :type audio_file: basestring | file
+        :type audio_file: str | file
         :param transcript: Transcription text.
-        :type transcript: basestring
+        :type transcript: str
         :return: Tuple of audio feature tensor and data of transcription part,
                  where transcription part could be token ids or text.
         :rtype: tuple of (2darray, list)
         """
-        if isinstance(audio_file, basestring) and audio_file.startswith('tar:'):
+        if isinstance(audio_file, str) and audio_file.startswith('tar:'):
             speech_segment = SpeechSegment.from_file(
                 self._subfile_from_tar(audio_file), transcript)
         else:
@@ -137,7 +134,7 @@ class DataGenerator(object):
         same shape, or a user-defined shape.
 
         :param manifest_path: Filepath of manifest for audio files.
-        :type manifest_path: basestring
+        :type manifest_path: str
         :param batch_size: Number of instances in a batch.
         :type batch_size: int
         :param padding_to:  If set -1, the maximun shape in the batch
@@ -361,7 +358,7 @@ class DataGenerator(object):
         """
         manifest.sort(key=lambda x: x["duration"])
         shift_len = self._rng.randint(0, batch_size - 1)
-        batch_manifest = zip(*[iter(manifest[shift_len:])] * batch_size)
+        batch_manifest = list(zip(*[iter(manifest[shift_len:])] * batch_size))
         self._rng.shuffle(batch_manifest)
         batch_manifest = [item for batch in batch_manifest for item in batch]
         if not clipped:
diff --git a/data_utils/featurizer/audio_featurizer.py b/data_utils/featurizer/audio_featurizer.py
index 0a54701bf..c5f20a16b 100644
--- a/data_utils/featurizer/audio_featurizer.py
+++ b/data_utils/featurizer/audio_featurizer.py
@@ -1,7 +1,4 @@
 """Contains the audio featurizer class."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import numpy as np
 from data_utils.utility import read_manifest
diff --git a/data_utils/featurizer/speech_featurizer.py b/data_utils/featurizer/speech_featurizer.py
index 4555dc31d..a4484ae44 100644
--- a/data_utils/featurizer/speech_featurizer.py
+++ b/data_utils/featurizer/speech_featurizer.py
@@ -1,7 +1,4 @@
 """Contains the speech featurizer class."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 from data_utils.featurizer.audio_featurizer import AudioFeaturizer
 from data_utils.featurizer.text_featurizer import TextFeaturizer
@@ -18,7 +15,7 @@ class SpeechFeaturizer(object):
 
     :param vocab_filepath: Filepath to load vocabulary for token indices
                            conversion.
-    :type specgram_type: basestring
+    :type specgram_type: str
     :param specgram_type: Specgram feature type. Options: 'linear', 'mfcc'.
     :type specgram_type: str
     :param stride_ms: Striding size (in milliseconds) for generating frames.
diff --git a/data_utils/featurizer/text_featurizer.py b/data_utils/featurizer/text_featurizer.py
index 89202163c..0c48b38e0 100644
--- a/data_utils/featurizer/text_featurizer.py
+++ b/data_utils/featurizer/text_featurizer.py
@@ -1,7 +1,4 @@
 """Contains the text featurizer class."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import os
 import codecs
@@ -16,7 +13,7 @@ class TextFeaturizer(object):
 
     :param vocab_filepath: Filepath to load vocabulary for token indices
                            conversion.
-    :type specgram_type: basestring
+    :type specgram_type: str
     """
 
     def __init__(self, vocab_filepath):
@@ -28,7 +25,7 @@ class TextFeaturizer(object):
         that the token indexing order follows the given vocabulary file.
 
         :param text: Text to process.
-        :type text: basestring
+        :type text: str
         :return: List of char-level token indices.
         :rtype: list
         """
diff --git a/data_utils/normalizer.py b/data_utils/normalizer.py
index 7c2e05c9d..378714cc0 100644
--- a/data_utils/normalizer.py
+++ b/data_utils/normalizer.py
@@ -1,7 +1,4 @@
 """Contains feature normalizers."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import numpy as np
 import random
@@ -18,9 +15,9 @@ class FeatureNormalizer(object):
     should be given for on-the-fly mean and stddev computing.
 
     :param mean_std_filepath: File containing the pre-computed mean and stddev.
-    :type mean_std_filepath: None|basestring
+    :type mean_std_filepath: None|str
     :param manifest_path: Manifest of instances for computing mean and stddev.
-    :type meanifest_path: None|basestring
+    :type meanifest_path: None|str
     :param featurize_func: Function to extract features. It should be callable
                            with ``featurize_func(audio_segment)``.
     :type featurize_func: None|callable
@@ -63,7 +60,7 @@ class FeatureNormalizer(object):
         """Write the mean and stddev to the file.
 
         :param filepath: File to write mean and stddev.
-        :type filepath: basestring
+        :type filepath: str
         """
         np.savez(filepath, mean=self._mean, std=self._std)
 
diff --git a/data_utils/speech.py b/data_utils/speech.py
index e64109dc8..3ea6ec296 100644
--- a/data_utils/speech.py
+++ b/data_utils/speech.py
@@ -1,7 +1,4 @@
 """Contains the speech segment class."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import numpy as np
 from data_utils.audio import AudioSegment
@@ -16,7 +13,7 @@ class SpeechSegment(AudioSegment):
     :param sample_rate: Audio sample rate.
     :type sample_rate: int
     :param transcript: Transcript text for the speech.
-    :type transript: basestring
+    :type transript: str
     :raises TypeError: If the sample data type is not float or int.
     """
 
@@ -42,9 +39,9 @@ class SpeechSegment(AudioSegment):
         """Create speech segment from audio file and corresponding transcript.
         
         :param filepath: Filepath or file object to audio file.
-        :type filepath: basestring|file
+        :type filepath: str|file
         :param transcript: Transcript text for the speech.
-        :type transript: basestring
+        :type transript: str
         :return: Speech segment instance.
         :rtype: SpeechSegment
         """
@@ -59,7 +56,7 @@ class SpeechSegment(AudioSegment):
         :param bytes: Byte string containing audio samples.
         :type bytes: str
         :param transcript: Transcript text for the speech.
-        :type transript: basestring
+        :type transript: str
         :return: Speech segment instance.
         :rtype: Speech Segment
         """
@@ -100,7 +97,7 @@ class SpeechSegment(AudioSegment):
         the entire file into the memory which can be incredibly wasteful.
 
         :param filepath: Filepath or file object to audio file.
-        :type filepath: basestring|file
+        :type filepath: str|file
         :param start: Start time in seconds. If start is negative, it wraps
                       around from the end. If not provided, this function 
                       reads from the very beginning.
@@ -111,7 +108,7 @@ class SpeechSegment(AudioSegment):
         :type end: float
         :param transcript: Transcript text for the speech. if not provided, 
                            the defaults is an empty string.
-        :type transript: basestring
+        :type transript: str
         :return: SpeechSegment instance of the specified slice of the input
                  speech file.
         :rtype: SpeechSegment
@@ -139,6 +136,6 @@ class SpeechSegment(AudioSegment):
         """Return the transcript text.
 
         :return: Transcript text for the speech.
-        :rtype: basestring
+        :rtype: str
         """
         return self._transcript
diff --git a/data_utils/utility.py b/data_utils/utility.py
index 7143f7ded..2377bc221 100644
--- a/data_utils/utility.py
+++ b/data_utils/utility.py
@@ -1,14 +1,10 @@
 """Contains data helper functions."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import json
 import codecs
 import os
 import tarfile
 import time
-from Queue import Queue
 from threading import Thread
 from multiprocessing import Process, Manager, Value
 from paddle.dataset.common import md5file
@@ -21,7 +17,7 @@ def read_manifest(manifest_path, max_duration=float('inf'), min_duration=0.0):
     filtered out.
 
     :param manifest_path: Manifest file to load and parse.
-    :type manifest_path: basestring
+    :type manifest_path: str
     :param max_duration: Maximal duration in seconds for instance filter.
     :type max_duration: float
     :param min_duration: Minimal duration in seconds for instance filter.
diff --git a/decoders/decoders_deprecated.py b/decoders/decoders_deprecated.py
index b9248b58b..0dad44c4a 100644
--- a/decoders/decoders_deprecated.py
+++ b/decoders/decoders_deprecated.py
@@ -1,7 +1,4 @@
 """Contains various CTC decoders."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 from itertools import groupby
 import numpy as np
@@ -123,7 +120,7 @@ def ctc_beam_search_decoder(probs_seq,
             prob_idx = prob_idx[0:cutoff_len]
 
         for l in prefix_set_prev:
-            if not prefix_set_next.has_key(l):
+            if l not in prefix_set_next:
                 probs_b_cur[l], probs_nb_cur[l] = 0.0, 0.0
 
             # extend prefix by travering prob_idx
@@ -137,7 +134,7 @@ def ctc_beam_search_decoder(probs_seq,
                     last_char = l[-1]
                     new_char = vocabulary[c]
                     l_plus = l + new_char
-                    if not prefix_set_next.has_key(l_plus):
+                    if l_plus not in prefix_set_next:
                         probs_b_cur[l_plus], probs_nb_cur[l_plus] = 0.0, 0.0
 
                     if new_char == last_char:
@@ -164,7 +161,7 @@ def ctc_beam_search_decoder(probs_seq,
 
         ## store top beam_size prefixes
         prefix_set_prev = sorted(
-            prefix_set_next.iteritems(), key=lambda asd: asd[1], reverse=True)
+            prefix_set_next.items(), key=lambda asd: asd[1], reverse=True)
         if beam_size < len(prefix_set_prev):
             prefix_set_prev = prefix_set_prev[:beam_size]
         prefix_set_prev = dict(prefix_set_prev)
diff --git a/decoders/scorer_deprecated.py b/decoders/scorer_deprecated.py
index c6a661030..266df559b 100644
--- a/decoders/scorer_deprecated.py
+++ b/decoders/scorer_deprecated.py
@@ -1,7 +1,4 @@
 """External Scorer for Beam Search Decoder."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import os
 import kenlm
@@ -20,7 +17,7 @@ class Scorer(object):
                 count when beta = 0.
     :type beta: float
     :model_path: Path to load language model.
-    :type model_path: basestring
+    :type model_path: str
     """
 
     def __init__(self, alpha, beta, model_path):
@@ -53,7 +50,7 @@ class Scorer(object):
         and return the final one.
 
         :param sentence: The input sentence for evalutation
-        :type sentence: basestring
+        :type sentence: str
         :param log: Whether return the score in log representation.
         :type log: bool
         :return: Evaluation score, in the decimal or log.
diff --git a/decoders/swig/_init_paths.py b/decoders/swig/_init_paths.py
index ddabb535b..3bb2fd197 100644
--- a/decoders/swig/_init_paths.py
+++ b/decoders/swig/_init_paths.py
@@ -1,7 +1,4 @@
 """Set up paths for DS2"""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import os.path
 import sys
diff --git a/decoders/swig/setup.py b/decoders/swig/setup.py
index a4bb2e9da..c5a7c4ca6 100644
--- a/decoders/swig/setup.py
+++ b/decoders/swig/setup.py
@@ -1,7 +1,4 @@
 """Script to build and install decoder package."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 from setuptools import setup, Extension, distutils
 import glob
diff --git a/decoders/swig/setup.sh b/decoders/swig/setup.sh
index eb5937186..73fa7aea7 100644
--- a/decoders/swig/setup.sh
+++ b/decoders/swig/setup.sh
@@ -21,4 +21,4 @@ if [ ! -d ThreadPool ]; then
 fi
 
 echo "Install decoders ..."
-python setup.py install --num_processes 4
+python3 setup.py install --num_processes 4
diff --git a/decoders/swig_wrapper.py b/decoders/swig_wrapper.py
index 21aed03c1..b32893b92 100644
--- a/decoders/swig_wrapper.py
+++ b/decoders/swig_wrapper.py
@@ -1,7 +1,4 @@
 """Wrapper for various CTC decoders in SWIG."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import swig_decoders
 
@@ -16,7 +13,7 @@ class Scorer(swig_decoders.Scorer):
                  count when beta = 0.
     :type beta: float
     :model_path: Path to load language model.
-    :type model_path: basestring
+    :type model_path: str
     """
 
     def __init__(self, alpha, beta, model_path, vocabulary):
@@ -33,7 +30,7 @@ def ctc_greedy_decoder(probs_seq, vocabulary):
     :param vocabulary: Vocabulary list.
     :type vocabulary: list
     :return: Decoding result string.
-    :rtype: basestring
+    :rtype: str
     """
     result = swig_decoders.ctc_greedy_decoder(probs_seq.tolist(), vocabulary)
     return result.decode('utf-8')
@@ -117,8 +114,6 @@ def ctc_beam_search_decoder_batch(probs_split,
     batch_beam_results = swig_decoders.ctc_beam_search_decoder_batch(
         probs_split, vocabulary, beam_size, num_processes, cutoff_prob,
         cutoff_top_n, ext_scoring_func)
-    batch_beam_results = [
-        [(res[0], res[1].decode("utf-8")) for res in beam_results]
-        for beam_results in batch_beam_results
-    ]
+    batch_beam_results = [[(res[0], res[1]) for res in beam_results]
+                          for beam_results in batch_beam_results]
     return batch_beam_results
diff --git a/decoders/tests/test_decoders.py b/decoders/tests/test_decoders.py
index d522b5efa..878417432 100644
--- a/decoders/tests/test_decoders.py
+++ b/decoders/tests/test_decoders.py
@@ -1,7 +1,4 @@
 """Test decoders."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import unittest
 from decoders import decoders_deprecated as decoder
diff --git a/deploy/_init_paths.py b/deploy/_init_paths.py
index ddabb535b..3bb2fd197 100644
--- a/deploy/_init_paths.py
+++ b/deploy/_init_paths.py
@@ -1,7 +1,4 @@
 """Set up paths for DS2"""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import os.path
 import sys
diff --git a/deploy/demo_client.py b/deploy/demo_client.py
index 7f8869462..45a9d319a 100644
--- a/deploy/demo_client.py
+++ b/deploy/demo_client.py
@@ -56,7 +56,7 @@ def callback(in_data, frame_count, time_info, status):
         print('Speech[length=%d] Sent.' % len(sent))
         # Receive data from the server and shut down
         received = sock.recv(1024)
-        print "Recognition Results: {}".format(received)
+        print("Recognition Results: {}".format(received))
         sock.close()
         data_list = []
     enable_trigger_record = True
diff --git a/deploy/demo_server.py b/deploy/demo_server.py
index 37782d377..88c2e9ef9 100644
--- a/deploy/demo_server.py
+++ b/deploy/demo_server.py
@@ -166,7 +166,7 @@ def start_server():
         place=place,
         share_rnn_weights=args.share_rnn_weights)
 
-    vocab_list = [chars.encode("utf-8") for chars in data_generator.vocab_list]
+    vocab_list = [chars for chars in data_generator.vocab_list]
 
     if args.decoding_method == "ctc_beam_search":
         ds2_model.init_ext_scorer(args.alpha, args.beta, args.lang_model_path,
diff --git a/examples/aishell/run_data.sh b/examples/aishell/run_data.sh
index 93ea6c291..fc002fcd2 100644
--- a/examples/aishell/run_data.sh
+++ b/examples/aishell/run_data.sh
@@ -3,7 +3,7 @@
 cd ../.. > /dev/null
 
 # download data, generate manifests
-PYTHONPATH=.:$PYTHONPATH python data/aishell/aishell.py \
+PYTHONPATH=.:$PYTHONPATH python3 data/aishell/aishell.py \
 --manifest_prefix='data/aishell/manifest' \
 --target_dir='./dataset/aishell'
 
@@ -14,7 +14,7 @@ fi
 
 
 # build vocabulary
-python tools/build_vocab.py \
+python3 tools/build_vocab.py \
 --count_threshold=0 \
 --vocab_path='data/aishell/vocab.txt' \
 --manifest_paths 'data/aishell/manifest.train' 'data/aishell/manifest.dev'
@@ -26,7 +26,7 @@ fi
 
 
 # compute mean and stddev for normalizer
-python tools/compute_mean_std.py \
+python3 tools/compute_mean_std.py \
 --manifest_path='data/aishell/manifest.train' \
 --num_samples=2000 \
 --specgram_type='linear' \
diff --git a/examples/aishell/run_infer.sh b/examples/aishell/run_infer.sh
index 6683b770c..ef382e310 100644
--- a/examples/aishell/run_infer.sh
+++ b/examples/aishell/run_infer.sh
@@ -13,7 +13,7 @@ cd - > /dev/null
 
 # infer
 CUDA_VISIBLE_DEVICES=0 \
-python -u infer.py \
+python3 -u infer.py \
 --num_samples=10 \
 --beam_size=300 \
 --num_proc_bsearch=8 \
diff --git a/examples/aishell/run_infer_golden.sh b/examples/aishell/run_infer_golden.sh
index 56d3365d9..dabdc0c68 100644
--- a/examples/aishell/run_infer_golden.sh
+++ b/examples/aishell/run_infer_golden.sh
@@ -22,7 +22,7 @@ cd - > /dev/null
 
 # infer
 CUDA_VISIBLE_DEVICES=0 \
-python -u infer.py \
+python3 -u infer.py \
 --num_samples=10 \
 --beam_size=300 \
 --num_proc_bsearch=8 \
diff --git a/examples/aishell/run_test.sh b/examples/aishell/run_test.sh
index 2867444be..b56d164b9 100644
--- a/examples/aishell/run_test.sh
+++ b/examples/aishell/run_test.sh
@@ -13,7 +13,7 @@ cd - > /dev/null
 
 # evaluate model
 CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
-python -u test.py \
+python3 -u test.py \
 --batch_size=128 \
 --beam_size=300 \
 --num_proc_bsearch=8 \
diff --git a/examples/aishell/run_test_golden.sh b/examples/aishell/run_test_golden.sh
index 799f382f5..2f79cc647 100644
--- a/examples/aishell/run_test_golden.sh
+++ b/examples/aishell/run_test_golden.sh
@@ -22,7 +22,7 @@ cd - > /dev/null
 
 # evaluate model
 CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
-python -u test.py \
+python3 -u test.py \
 --batch_size=128 \
 --beam_size=300 \
 --num_proc_bsearch=8 \
diff --git a/examples/aishell/run_train.sh b/examples/aishell/run_train.sh
index 335473fcf..889e19047 100644
--- a/examples/aishell/run_train.sh
+++ b/examples/aishell/run_train.sh
@@ -6,7 +6,7 @@ cd ../.. > /dev/null
 # if you wish to resume from an exists model, uncomment --init_from_pretrained_model
 export FLAGS_sync_nccl_allreduce=0
 CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
-python -u train.py \
+python3 -u train.py \
 --batch_size=64 \
 --num_epoch=50 \
 --num_conv_layers=2 \
diff --git a/examples/baidu_en8k/run_infer_golden.sh b/examples/baidu_en8k/run_infer_golden.sh
index 2f3f0acf7..85fea5f47 100644
--- a/examples/baidu_en8k/run_infer_golden.sh
+++ b/examples/baidu_en8k/run_infer_golden.sh
@@ -22,7 +22,7 @@ cd - > /dev/null
 
 # infer
 CUDA_VISIBLE_DEVICES=0 \
-python -u infer.py \
+python3 -u infer.py \
 --num_samples=10 \
 --beam_size=500 \
 --num_proc_bsearch=5 \
diff --git a/examples/baidu_en8k/run_test_golden.sh b/examples/baidu_en8k/run_test_golden.sh
index 612e71a01..1ed633624 100644
--- a/examples/baidu_en8k/run_test_golden.sh
+++ b/examples/baidu_en8k/run_test_golden.sh
@@ -22,7 +22,7 @@ cd - > /dev/null
 
 # evaluate model
 CUDA_VISIBLE_DEVICES=0,1,2,3 \
-python -u test.py \
+python3 -u test.py \
 --batch_size=128 \
 --beam_size=500 \
 --num_proc_bsearch=8 \
diff --git a/examples/deploy_demo/run_demo_client.sh b/examples/deploy_demo/run_demo_client.sh
index 6ae4ddcab..7c2609511 100644
--- a/examples/deploy_demo/run_demo_client.sh
+++ b/examples/deploy_demo/run_demo_client.sh
@@ -4,7 +4,7 @@ cd ../.. > /dev/null
 
 # start demo client
 CUDA_VISIBLE_DEVICES=0 \
-python -u deploy/demo_client.py \
+python3 -u deploy/demo_client.py \
 --host_ip='localhost' \
 --host_port=8086 \
 
diff --git a/examples/deploy_demo/run_english_demo_server.sh b/examples/deploy_demo/run_english_demo_server.sh
index d67559f33..5d86ab6e1 100644
--- a/examples/deploy_demo/run_english_demo_server.sh
+++ b/examples/deploy_demo/run_english_demo_server.sh
@@ -23,7 +23,7 @@ cd - > /dev/null
 
 # start demo server
 CUDA_VISIBLE_DEVICES=0 \
-python -u deploy/demo_server.py \
+python3 -u deploy/demo_server.py \
 --host_ip='localhost' \
 --host_port=8086 \
 --num_conv_layers=2 \
diff --git a/examples/librispeech/run_data.sh b/examples/librispeech/run_data.sh
index e4db1ac9b..788b415de 100644
--- a/examples/librispeech/run_data.sh
+++ b/examples/librispeech/run_data.sh
@@ -3,7 +3,7 @@
 cd ../.. > /dev/null
 
 # download data, generate manifests
-PYTHONPATH=.:$PYTHONPATH python data/librispeech/librispeech.py \
+PYTHONPATH=.:$PYTHONPATH python3 data/librispeech/librispeech.py \
 --manifest_prefix='data/librispeech/manifest' \
 --target_dir='./dataset/librispeech' \
 --full_download='True'
@@ -17,7 +17,7 @@ cat data/librispeech/manifest.train-* | shuf > data/librispeech/manifest.train
 
 
 # build vocabulary
-python tools/build_vocab.py \
+python3 tools/build_vocab.py \
 --count_threshold=0 \
 --vocab_path='data/librispeech/vocab.txt' \
 --manifest_paths='data/librispeech/manifest.train'
@@ -29,7 +29,7 @@ fi
 
 
 # compute mean and stddev for normalizer
-python tools/compute_mean_std.py \
+python3 tools/compute_mean_std.py \
 --manifest_path='data/librispeech/manifest.train' \
 --num_samples=2000 \
 --specgram_type='linear' \
diff --git a/examples/librispeech/run_infer.sh b/examples/librispeech/run_infer.sh
index 91d8ff2eb..2452b45e3 100644
--- a/examples/librispeech/run_infer.sh
+++ b/examples/librispeech/run_infer.sh
@@ -13,7 +13,7 @@ cd - > /dev/null
 
 # infer
 CUDA_VISIBLE_DEVICES=0 \
-python -u infer.py \
+python3 -u infer.py \
 --num_samples=10 \
 --beam_size=500 \
 --num_proc_bsearch=8 \
diff --git a/examples/librispeech/run_infer_golden.sh b/examples/librispeech/run_infer_golden.sh
index eb8121294..6004c6af6 100644
--- a/examples/librispeech/run_infer_golden.sh
+++ b/examples/librispeech/run_infer_golden.sh
@@ -22,7 +22,7 @@ cd - > /dev/null
 
 # infer
 CUDA_VISIBLE_DEVICES=0 \
-python -u infer.py \
+python3 -u infer.py \
 --num_samples=10 \
 --beam_size=500 \
 --num_proc_bsearch=8 \
diff --git a/examples/librispeech/run_test.sh b/examples/librispeech/run_test.sh
index 9eebbbf24..e8e65c39e 100644
--- a/examples/librispeech/run_test.sh
+++ b/examples/librispeech/run_test.sh
@@ -13,7 +13,7 @@ cd - > /dev/null
 
 # evaluate model
 CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
-python -u test.py \
+python3 -u test.py \
 --batch_size=128 \
 --beam_size=500 \
 --num_proc_bsearch=8 \
diff --git a/examples/librispeech/run_test_golden.sh b/examples/librispeech/run_test_golden.sh
index abd895925..1db36758a 100644
--- a/examples/librispeech/run_test_golden.sh
+++ b/examples/librispeech/run_test_golden.sh
@@ -22,7 +22,7 @@ cd - > /dev/null
 
 # evaluate model
 CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
-python -u test.py \
+python3 -u test.py \
 --batch_size=128 \
 --beam_size=500 \
 --num_proc_bsearch=8 \
diff --git a/examples/librispeech/run_train.sh b/examples/librispeech/run_train.sh
index a568bf221..6b8982d78 100644
--- a/examples/librispeech/run_train.sh
+++ b/examples/librispeech/run_train.sh
@@ -7,7 +7,7 @@ cd ../.. > /dev/null
 export FLAGS_sync_nccl_allreduce=0
 
 CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
-python -u train.py \
+python3 -u train.py \
 --batch_size=20 \
 --num_epoch=50 \
 --num_conv_layers=2 \
diff --git a/examples/librispeech/run_tune.sh b/examples/librispeech/run_tune.sh
index af6e9dafd..834f060fa 100644
--- a/examples/librispeech/run_tune.sh
+++ b/examples/librispeech/run_tune.sh
@@ -4,7 +4,7 @@ cd ../.. > /dev/null
 
 # grid-search for hyper-parameters in language model
 CUDA_VISIBLE_DEVICES=0,1,2,3 \
-python -u tools/tune.py \
+python3 -u tools/tune.py \
 --num_batches=-1 \
 --batch_size=128 \
 --beam_size=500 \
diff --git a/examples/tiny/run_data.sh b/examples/tiny/run_data.sh
index 1428194d3..dd75ddadc 100644
--- a/examples/tiny/run_data.sh
+++ b/examples/tiny/run_data.sh
@@ -8,7 +8,7 @@ if [ ! -e data/tiny ]; then
 fi
 
 # download data, generate manifests
-PYTHONPATH=.:$PYTHONPATH python data/librispeech/librispeech.py \
+PYTHONPATH=.:$PYTHONPATH python3 data/librispeech/librispeech.py \
 --manifest_prefix='data/tiny/manifest' \
 --target_dir='./dataset/librispeech' \
 --full_download='False'
@@ -21,7 +21,7 @@ fi
 head -n 64 data/tiny/manifest.dev-clean  > data/tiny/manifest.tiny
 
 # build vocabulary
-python tools/build_vocab.py \
+python3 tools/build_vocab.py \
 --count_threshold=0 \
 --vocab_path='data/tiny/vocab.txt' \
 --manifest_paths='data/tiny/manifest.tiny'
@@ -33,7 +33,7 @@ fi
 
 
 # compute mean and stddev for normalizer
-python tools/compute_mean_std.py \
+python3 tools/compute_mean_std.py \
 --manifest_path='data/tiny/manifest.tiny' \
 --num_samples=64 \
 --specgram_type='linear' \
diff --git a/examples/tiny/run_infer.sh b/examples/tiny/run_infer.sh
index bded0e7b6..d88f4526a 100644
--- a/examples/tiny/run_infer.sh
+++ b/examples/tiny/run_infer.sh
@@ -13,7 +13,7 @@ cd - > /dev/null
 
 # infer
 CUDA_VISIBLE_DEVICES=0 \
-python -u infer.py \
+python3 -u infer.py \
 --num_samples=10 \
 --beam_size=500 \
 --num_proc_bsearch=8 \
diff --git a/examples/tiny/run_infer_golden.sh b/examples/tiny/run_infer_golden.sh
index 33662622d..d18c21f54 100644
--- a/examples/tiny/run_infer_golden.sh
+++ b/examples/tiny/run_infer_golden.sh
@@ -22,7 +22,7 @@ cd - > /dev/null
 
 # infer
 CUDA_VISIBLE_DEVICES=0 \
-python -u infer.py \
+python3 -u infer.py \
 --num_samples=10 \
 --beam_size=500 \
 --num_proc_bsearch=8 \
diff --git a/examples/tiny/run_test.sh b/examples/tiny/run_test.sh
index 1dfc65e19..81eafe236 100644
--- a/examples/tiny/run_test.sh
+++ b/examples/tiny/run_test.sh
@@ -13,7 +13,7 @@ cd - > /dev/null
 
 # evaluate model
 CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
-python -u test.py \
+python3 -u test.py \
 --batch_size=128 \
 --beam_size=500 \
 --num_proc_bsearch=8 \
diff --git a/examples/tiny/run_test_golden.sh b/examples/tiny/run_test_golden.sh
index 542552657..d82865f42 100644
--- a/examples/tiny/run_test_golden.sh
+++ b/examples/tiny/run_test_golden.sh
@@ -22,7 +22,7 @@ cd - > /dev/null
 
 # evaluate model
 CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
-python -u test.py \
+python3 -u test.py \
 --batch_size=128 \
 --beam_size=500 \
 --num_proc_bsearch=8 \
diff --git a/examples/tiny/run_train.sh b/examples/tiny/run_train.sh
index 95ccd2bc6..fe5b62030 100644
--- a/examples/tiny/run_train.sh
+++ b/examples/tiny/run_train.sh
@@ -6,7 +6,7 @@ cd ../.. > /dev/null
 # if you wish to resume from an exists model, uncomment --init_from_pretrained_model
 export FLAGS_sync_nccl_allreduce=0
 CUDA_VISIBLE_DEVICES=0,1,2,3 \
-python -u train.py \
+python3 -u train.py \
 --batch_size=4 \
 --num_epoch=20 \
 --num_conv_layers=2 \
diff --git a/examples/tiny/run_tune.sh b/examples/tiny/run_tune.sh
index 87bcb67b1..bec71111a 100644
--- a/examples/tiny/run_tune.sh
+++ b/examples/tiny/run_tune.sh
@@ -4,7 +4,7 @@ cd ../.. > /dev/null
 
 # grid-search for hyper-parameters in language model
 CUDA_VISIBLE_DEVICES=0,1,2,3 \
-python -u tools/tune.py \
+python3 -u tools/tune.py \
 --num_batches=-1 \
 --batch_size=128 \
 --beam_size=500 \
diff --git a/infer.py b/infer.py
index 9e6a8c3b2..910140282 100644
--- a/infer.py
+++ b/infer.py
@@ -1,12 +1,6 @@
 """Inferer for DeepSpeech2 model."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import sys
-reload(sys)
-sys.setdefaultencoding('utf-8')
-
 import argparse
 import functools
 import paddle.fluid as fluid
@@ -104,7 +98,7 @@ def infer():
         init_from_pretrained_model=args.model_path)
 
     # decoders only accept string encoded in utf-8
-    vocab_list = [chars.encode("utf-8") for chars in data_generator.vocab_list]
+    vocab_list = [chars for chars in data_generator.vocab_list]
 
     if args.decoding_method == "ctc_greedy":
         ds2_model.logger.info("start inference ...")
diff --git a/model_utils/model.py b/model_utils/model.py
index 7cb26f39f..68b963cdf 100644
--- a/model_utils/model.py
+++ b/model_utils/model.py
@@ -1,7 +1,4 @@
 """Contains DeepSpeech2 model."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import sys
 import os
@@ -10,7 +7,6 @@ import logging
 import gzip
 import copy
 import inspect
-import cPickle as pickle
 import collections
 import multiprocessing
 import numpy as np
@@ -445,7 +441,7 @@ class DeepSpeech2Model(object):
         :param vocab_list: List of tokens in the vocabulary, for decoding.
         :type vocab_list: list
         :return: List of transcription texts.
-        :rtype: List of basestring
+        :rtype: List of str
         """
         results = []
         for i, probs in enumerate(probs_split):
@@ -466,7 +462,7 @@ class DeepSpeech2Model(object):
                                     empty, the external scorer will be set to
                                     None, and the decoding method will be pure
                                     beam search without scorer.
-        :type language_model_path: basestring|None
+        :type language_model_path: str|None
         :param vocab_list: List of tokens in the vocabulary, for decoding.
         :type vocab_list: list
         """
@@ -513,7 +509,7 @@ class DeepSpeech2Model(object):
         :param num_processes: Number of processes (CPU) for decoder.
         :type num_processes: int
         :return: List of transcription texts.
-        :rtype: List of basestring
+        :rtype: List of str
         """
         if self._ext_scorer != None:
             self._ext_scorer.reset_params(beam_alpha, beam_beta)
diff --git a/model_utils/network.py b/model_utils/network.py
index e27ff02cf..b8812e609 100644
--- a/model_utils/network.py
+++ b/model_utils/network.py
@@ -1,7 +1,3 @@
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
-
 import collections
 import paddle.fluid as fluid
 import numpy as np
diff --git a/setup.sh b/setup.sh
index ec5e47ec8..8d471cbf6 100644
--- a/setup.sh
+++ b/setup.sh
@@ -2,7 +2,7 @@
 
 # install python dependencies
 if [ -f "requirements.txt" ]; then
-    pip install -r requirements.txt
+    pip3 install -r requirements.txt
 fi
 if [ $? != 0 ]; then
     echo "Install python dependencies failed !!!"
@@ -10,7 +10,7 @@ if [ $? != 0 ]; then
 fi
 
 # install package libsndfile
-python -c "import soundfile"
+python3 -c "import soundfile"
 if [ $? != 0 ]; then
     echo "Install package libsndfile into default system path."
     wget "http://www.mega-nerd.com/libsndfile/files/libsndfile-1.0.28.tar.gz"
@@ -27,7 +27,7 @@ if [ $? != 0 ]; then
 fi
 
 # install decoders
-python -c "import pkg_resources; pkg_resources.require(\"swig_decoders==1.1\")"
+python3 -c "import pkg_resources; pkg_resources.require(\"swig_decoders==1.1\")"
 if [ $? != 0 ]; then
     cd decoders/swig > /dev/null
     sh setup.sh
diff --git a/test.py b/test.py
index ec9fdfda6..053a43acd 100644
--- a/test.py
+++ b/test.py
@@ -1,7 +1,4 @@
 """Evaluation for DeepSpeech2 model."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import argparse
 import functools
@@ -99,7 +96,7 @@ def evaluate():
         init_from_pretrained_model=args.model_path)
 
     # decoders only accept string encoded in utf-8
-    vocab_list = [chars.encode("utf-8") for chars in data_generator.vocab_list]
+    vocab_list = [chars for chars in data_generator.vocab_list]
 
     if args.decoding_method == "ctc_beam_search":
         ds2_model.init_ext_scorer(args.alpha, args.beta, args.lang_model_path,
diff --git a/tools/_init_paths.py b/tools/_init_paths.py
index ddabb535b..3bb2fd197 100644
--- a/tools/_init_paths.py
+++ b/tools/_init_paths.py
@@ -1,7 +1,4 @@
 """Set up paths for DS2"""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import os.path
 import sys
diff --git a/tools/build_vocab.py b/tools/build_vocab.py
index e167e92ad..13ab843c4 100644
--- a/tools/build_vocab.py
+++ b/tools/build_vocab.py
@@ -2,9 +2,6 @@
 
 Each item in vocabulary file is a character.
 """
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import argparse
 import functools
diff --git a/tools/compute_mean_std.py b/tools/compute_mean_std.py
index 11aa856d7..203f83a2c 100644
--- a/tools/compute_mean_std.py
+++ b/tools/compute_mean_std.py
@@ -1,7 +1,4 @@
 """Compute mean and std for feature normalizer, and save to file."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import argparse
 import functools
diff --git a/tools/tune.py b/tools/tune.py
index 7996e4d53..74007f55a 100644
--- a/tools/tune.py
+++ b/tools/tune.py
@@ -1,7 +1,4 @@
 """Beam search parameters tuning for DeepSpeech2 model."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import sys
 import os
@@ -107,7 +104,7 @@ def tune():
         share_rnn_weights=args.share_rnn_weights)
 
     # decoders only accept string encoded in utf-8
-    vocab_list = [chars.encode("utf-8") for chars in data_generator.vocab_list]
+    vocab_list = [chars for chars in data_generator.vocab_list]
     errors_func = char_errors if args.error_rate_type == 'cer' else word_errors
     # create grid for search
     cand_alphas = np.linspace(args.alpha_from, args.alpha_to, args.num_alphas)
diff --git a/train.py b/train.py
index 5dae4ccdd..caa7c266e 100644
--- a/train.py
+++ b/train.py
@@ -1,7 +1,4 @@
 """Trainer for DeepSpeech2 model."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import argparse
 import functools
diff --git a/utils/error_rate.py b/utils/error_rate.py
index d84d9f875..628a84cb4 100644
--- a/utils/error_rate.py
+++ b/utils/error_rate.py
@@ -1,10 +1,6 @@
-# -*- coding: utf-8 -*-
 """This module provides functions to calculate error rate in different level.
 e.g. wer for word-level, cer for char-level.
 """
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import numpy as np
 
@@ -61,9 +57,9 @@ def word_errors(reference, hypothesis, ignore_case=False, delimiter=' '):
     hypothesis sequence in word-level.
 
     :param reference: The reference sentence.
-    :type reference: basestring
+    :type reference: str
     :param hypothesis: The hypothesis sentence.
-    :type hypothesis: basestring
+    :type hypothesis: str
     :param ignore_case: Whether case-sensitive or not.
     :type ignore_case: bool
     :param delimiter: Delimiter of input sentences.
@@ -75,8 +71,8 @@ def word_errors(reference, hypothesis, ignore_case=False, delimiter=' '):
         reference = reference.lower()
         hypothesis = hypothesis.lower()
 
-    ref_words = filter(None, reference.split(delimiter))
-    hyp_words = filter(None, hypothesis.split(delimiter))
+    ref_words = list(filter(None, reference.split(delimiter)))
+    hyp_words = list(filter(None, hypothesis.split(delimiter)))
 
     edit_distance = _levenshtein_distance(ref_words, hyp_words)
     return float(edit_distance), len(ref_words)
@@ -87,9 +83,9 @@ def char_errors(reference, hypothesis, ignore_case=False, remove_space=False):
     hypothesis sequence in char-level.
 
     :param reference: The reference sentence.
-    :type reference: basestring
+    :type reference: str
     :param hypothesis: The hypothesis sentence.
-    :type hypothesis: basestring
+    :type hypothesis: str
     :param ignore_case: Whether case-sensitive or not.
     :type ignore_case: bool
     :param remove_space: Whether remove internal space characters
@@ -105,8 +101,8 @@ def char_errors(reference, hypothesis, ignore_case=False, remove_space=False):
     if remove_space == True:
         join_char = ''
 
-    reference = join_char.join(filter(None, reference.split(' ')))
-    hypothesis = join_char.join(filter(None, hypothesis.split(' ')))
+    reference = join_char.join(list(filter(None, reference.split(' '))))
+    hypothesis = join_char.join(list(filter(None, hypothesis.split(' '))))
 
     edit_distance = _levenshtein_distance(reference, hypothesis)
     return float(edit_distance), len(reference)
@@ -132,9 +128,9 @@ def wer(reference, hypothesis, ignore_case=False, delimiter=' '):
     that empty items will be removed when splitting sentences by delimiter.
 
     :param reference: The reference sentence.
-    :type reference: basestring
+    :type reference: str
     :param hypothesis: The hypothesis sentence.
-    :type hypothesis: basestring
+    :type hypothesis: str
     :param ignore_case: Whether case-sensitive or not.
     :type ignore_case: bool
     :param delimiter: Delimiter of input sentences.
@@ -175,9 +171,9 @@ def cer(reference, hypothesis, ignore_case=False, remove_space=False):
     characters in a sentence will be replaced by one space character.
 
     :param reference: The reference sentence.
-    :type reference: basestring
+    :type reference: str
     :param hypothesis: The hypothesis sentence.
-    :type hypothesis: basestring
+    :type hypothesis: str
     :param ignore_case: Whether case-sensitive or not.
     :type ignore_case: bool
     :param remove_space: Whether remove internal space characters
diff --git a/utils/tests/test_error_rate.py b/utils/tests/test_error_rate.py
index d6bc7442e..efa04b827 100644
--- a/utils/tests/test_error_rate.py
+++ b/utils/tests/test_error_rate.py
@@ -1,8 +1,4 @@
-# -*- coding: utf-8 -*-
 """Test error rate."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import unittest
 from utils import error_rate
diff --git a/utils/utility.py b/utils/utility.py
index 2e489ade6..543f3ebce 100644
--- a/utils/utility.py
+++ b/utils/utility.py
@@ -1,7 +1,4 @@
 """Contains common utility functions."""
-from __future__ import absolute_import
-from __future__ import division
-from __future__ import print_function
 
 import distutils.util
 
@@ -22,7 +19,7 @@ def print_arguments(args):
     :type args: argparse.Namespace
     """
     print("-----------  Configuration Arguments -----------")
-    for arg, value in sorted(vars(args).iteritems()):
+    for arg, value in sorted(vars(args).items()):
         print("%s: %s" % (arg, value))
     print("------------------------------------------------")