Jammy@Jetson Orin - Tensorflow & Keras Get Started: 000 setup for tutorial

原创已于 2024-04-29 10:35:48 修改 · 1.2k 阅读

10 ·

CC 4.0 BY-SA版权

文章标签：

#tensorflow #keras #人工智能

于 2024-04-23 16:59:51 首次发布

Linux 专栏收录该内容

252 篇文章

订阅专栏

本文详细介绍了在JetsonOrin上安装和配置Tensorflow、Keras、JAX等深度学习框架的步骤，以及遇到的numpy版本兼容、API编译和预测问题的解决方案。通过实际操作演示，强调了逐步学习和解决问题的重要性。

部署运行你感兴趣的模型镜像

Jammy@Jetson Orin - Tensorflow & Keras Get Started: 000 setup for tutorial

1. 源由
2. 搭建环境
3. 测试DEMO
4. 总结
5. 参考资料

1. 源由

凡事开头难！入门搭建环境难！

这里就从最基本的环境搭建和大家共一起勉！

2. 搭建环境

2.1 安装IDE环境

jupyterlab环境

$ pip install jupyterlab
$ jupyter lab

jupyternotebook环境

$ pip install notebook
$ jupyter notebook

注：推荐使用jupyterlab。

2.2 安装numpy

$ pip install -U numpy  //升级到最新版本

$ python
Python 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import numpy
>>> numpy.__version__
'1.26.4'
>>>

注：升级到指定版本可以使用命令pip install numpy==1.24.3

2.3 安装keras

$ pip install --upgrade keras

$ python
Python 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import keras
>>> keras.__version__
'3.3.2'
>>>

参考：Keras Get Started

2.4 安装JAX

CPU-only (Linux/macOS/Windows)

$ pip install -U "jax[cpu]"

GPU (NVIDIA, CUDA 12, x86_64)

$ pip install -U "jax[cuda12_pip]" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html

注：更多关于JAX的硬件版本信息，详见：Installing JAX

2.5 安装tensorflow

$ pip install tensorflow

$ python
Python 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import tensorflow as tf
>>> tf.__version__
'2.16.1'
>>> tf.__path__
['/home/daniel/.local/lib/python3.10/site-packages/keras/api/_v2', '/home/daniel/.local/lib/python3.10/site-packages/keras/_tf_keras', '/home/daniel/.local/lib/python3.10/site-packages/tensorflow', '/home/daniel/.local/lib/python3.10/site-packages/tensorflow/_api/v2']
>>>

注意：目前tensorflow的安装存在一些GPU版本的问题，参考下面的链接资料。

2.6 安装PyTorch

具体安装版本因硬件差异，命令不同，详见：Install pytorch

在这里插入图片描述因为，笔者这里环境是Jetson Orin，所以选择了上面的配置版本：

$ pip install torch torchvision torchaudio

2.7 安装nbdiff

鉴于.ipynb文件会包含最后一次执行的输出信息，不像通常代码diff那样可以看的很清楚，这里需要安装一个类似diff的命令。

$ pip install nbdime

$ nbdiff --help
usage: nbdiff [-h] [--version] [--config] [--log-level {DEBUG,INFO,WARN,ERROR,CRITICAL}] [-s] [-o] [-a] [-m] [-i] [-d] [--color-words] [--no-color] [--no-git] [--no-use-diff] [--out OUT]
              [base] [remote] [paths ...]

Compute the difference between two Jupyter notebooks.

positional arguments:
  base                  the base notebook filename OR base git-revision.
  remote                the remote modified notebook filename OR remote git-revision.
  paths                 filter diffs for git-revisions based on path

options:
  -h, --help            show this help message and exit
  --version             show program's version number and exit
  --config              list the valid config keys and their current effective values
  --log-level {DEBUG,INFO,WARN,ERROR,CRITICAL}
                        set the log level by name.
  --color-words         whether to pass the --color-words flag to any internal calls to git diff
  --no-color            prevent use of ANSI color code escapes for text output
  --no-git              prevent use of git for formatting diff/merge text output
  --no-use-diff         prevent use of diff/diff3 for formatting diff/merge text output
  --out OUT             if supplied, the diff is written to this file. Otherwise it is printed to the terminal.

ignorables:
  Set which parts of the notebook (not) to process.

  -s, --sources, -S, --ignore-sources
                        process/ignore sources.
  -o, --outputs, -O, --ignore-outputs
                        process/ignore outputs.
  -a, --attachments, -A, --ignore-attachments
                        process/ignore attachments.
  -m, --metadata, -M, --ignore-metadata
                        process/ignore metadata.
  -i, --id, -I, --ignore-id
                        process/ignore identifiers.
  -d, --details, -D, --ignore-details
                        process/ignore details not covered by other options.

3. 测试DEMO

学习是一个过程，是一种大学生应该掌握的技能。手把手教那是在学校，真正的学习是不断的自我学习和提高，这种螺旋式学习技能将会受益一辈子！

即使很好的搭建了环境，代码依然会出现问题！

001_Keras-Linear-Regression

$ git log -n 2
commit 84b7f5ee7c80d9faecf79af96f8a677f47c44f0d (HEAD -> main, origin/main, origin/HEAD)
Author: Daniel Li <lida_mail@163.com>
Date:   Tue Apr 23 16:49:42 2024 +0800

    Fix Keras-Linear-Regression demo code issue with Jammy(Jetson Orin)

commit 8c89b4c2b9e9df2e854f280ce19ed3010c7ac2fc
Author: Daniel Li <lida_mail@163.com>
Date:   Tue Apr 23 15:00:46 2024 +0800

    Add raw 001_Keras-Linear-Regression/Keras-Linear-Regression.ipynb

3.1 numpy版本兼容问题

在这里插入图片描述
解决方法：numpy版本降级

$ pip install numpy==1.23.4

3.2 karas API - model.compile问题

在这里插入图片描述

解决方法：修正API入参参数

## modified /cells/15/source:
@@ -1,2 +1,2 @@
-model.compile(optimizer=tf.keras.optimizers.RMSprop(lr=.005),
+model.compile(optimizer=tf.keras.optimizers.RMSprop(learning_rate=.005),
               loss='mse')

3.3 karas API - model.predict问题

在这里插入图片描述
解决方法：修正API入参参数

## modified /cells/23/source:
@@ -1,5 +1,5 @@
 # Predict the median price of a home with [3, 4, 5, 6, 7] rooms.
-x = [3, 4, 5, 6, 7]
-y_pred = model.predict(x)
-for idx in range(len(x)):
-    print("Predicted price of a home with {} rooms: ${}K".format(x[idx], int(y_pred[idx]*10)/10))

+rooms = [3, 4, 5, 6, 7]
+y_pred = model.predict(x = np.array(rooms))
+for idx in range(len(rooms)):
+    print("Predicted price of a home with {} rooms: ${}K".format(rooms[idx], int(y_pred[idx][0]*10)/10))