随笔分类 - Tensorflow
计算机视觉 Tensorflow
摘要:Seq2seq with Attention (Example: Neural machine translation) Beyond RNNs: Transformer, Tensor2Tensor Dialogue agents (A3) Reinforcement Learning in Te
阅读全文
摘要:1. 理论review 1.1 RNN 1. Agenda From feed forward to recurrent / Tricks & treats / Presidential tweets 2. BPTT Use sum of gradients each all timesteps t
阅读全文
摘要:1. VAE 1. lecture link: https://docs.google.com/presentation/d/1VSNlkGcR b39tMcuREjzZdhYOPvoZudpcbuNlf5hOIM/edit slide=id.g334db163d4_0_41 2. todo, 我打
阅读全文
摘要:1. 一些先修知识 1.1 TFrecord 1. 优点:tf官方推荐格式,兼容大部分格式,采用二进制保存文件,在tf中处理效率最高 2. cs20的一个good example: 写tfrecord的pipeline 读tfrecord的pipeline 一个示例: 3. 一个过时的queue的例
阅读全文
摘要:1. ConvNet 1.1 一些小知识点 1. tf中的tf.nn.conv2d的几个新感悟 Strides often [1,x, x, 1]: 因为stride 分别指定:batch, w, h, c 四个方向上的滑动,一般情况下我们肯定不想跳过某些batch和channel,所以1 th,
阅读全文
摘要:1. Assignment_1 1.1 1d解答 1. tf.where() 我的实践代码: 参考: [1] https://blog.csdn.net/A_a_ron/article/details/79048446 (讲得不错) [2] https://blog.csdn.net/ustbbsy
阅读全文
摘要:1. namspace(scope) 1.1 name_scope: create namespaces, ops分组, 进而tensorboard可以各个ops进行分组显示,进而使得tensorboard的可视化效果更好,进而有助于debug 1.2 variable_scope: 使用tf.ge
阅读全文
摘要:0. 华师 张凯旭问我启发的一个学习资料 如何加载/保存原模型?fine tune如何实践?如何实践transfer learning ? 当前我的答案:https://cv tricks.com/tensorflow tutorial/save restore tensorflow models
阅读全文
摘要:1. eager速记 1. 可以用大部分python的方法,而不仅限于tf.xxx is compatible with Python debugging tools: provides immediate error reporting,而不需要sess.run(xxxerror) permits
阅读全文
摘要:1. tf.data 1. 相比 feed_in和placeholder的优势:数据的一些操作(比如shuffle/batch/repeat/map)集成在tf中,所以效率高速度快,而且属于high level api,使用方便 2. 3. 输出一些dataset的types/shape以做sani
阅读全文
摘要:1. Linear Regression: Predict life expectancy from birth rate 1.1 Problem 1. 描述问题: I recently came across the visualization of the relationship betwee
阅读全文
摘要:0. Review 1. Computation graph TensorFlow separates definition of computations from their execution Phase 1: assemble a graph Phase 2: use a session t
阅读全文
摘要:1. Tensorboard 1. tensorboard使用 Note : If you've run your code several times, there will be multiple event files in your [logdir]. TF will show only t
阅读全文
摘要:1. Agenda Basic operations Tensor types Importing data Lazy loading 2. Basic operations 1. tensorborad使用介绍 2. Constants, Sequences, Variables, Ops con
阅读全文
摘要:1. Why Tensorflow 1. key idea: For a framework to be useful in production : it needs to be efficient, scalable, and maintainable . For research : the
阅读全文
摘要:1. 基本特点 1.1 save computation(惰性运行) 如上,因为 不需要用到 ,所以 的运算不会被执行(save computation) 1.2 Distributed Computation 1. 大图分解为子图多GPU并行计算, e.g. freamework举例 2. Mul
阅读全文