TF-Agents: A Flexible Reinforcement Learning Library for TensorFlow (Google I/O'19)

Поділитися
Вставка
  • Опубліковано 12 січ 2025

КОМЕНТАРІ • 13

  • @THEcucufate
    @THEcucufate 5 років тому

    Excellent presentation. Looks like you have a lot of fun working on this! I would love to learn more and will try to piece all this together and hope I get it. Great job!

  • @Gottii92
    @Gottii92 5 років тому +3

    videos like this always make me enthusiastic, is there a group where you can collaborate to do such things just as a hobby?

  • @PatrickNgobiro
    @PatrickNgobiro 5 років тому

    awesome presentation

  • @tagoma3504
    @tagoma3504 5 років тому

    Great talk, thank you. Are the slides of this presentation available somewhere, please?

  • @stgpepper
    @stgpepper 5 років тому

    This is awesome, thank you Google. Once I train the model how can I deploy it, in the robot example they trained it in the sim then deployed it to the actual robot. From a quick search I see pickle is one way but are there other ways to deploy a model?

    • @SergioGuadarrama
      @SergioGuadarrama 5 років тому

      We use TF.SavedModel which can be run from C++.

  • @franklinc.fernandez985
    @franklinc.fernandez985 5 років тому

    Hi, very good video!!! How can I install it with TF 1.14 ??

  • @KrishnaChaitanyakosaraju
    @KrishnaChaitanyakosaraju 5 років тому

    Hi, could you please share the code used in the talk.

    • @SergioGuadarrama
      @SergioGuadarrama 5 років тому

      The full code is available at github.com/tensorflow/agents
      Check the Colabs.

    • @shreepads
      @shreepads 5 років тому

      Sergio I can't find the second example, with dual networks on Breakout in the colabs, could you please point me to the right file? Thanks!

  • @RajatKumar_1502
    @RajatKumar_1502 5 років тому

    Hi Sergio and Eugene
    ,
    Great session, amazing library to start with for a beginner. Thanks for sharing that.
    Regarding the Example 2 for Breakout-v0, I am facing some issues when implementing the colab.
    I am refereeing to the 1_dqn_tutorial.ipynb (github.com/tensorflow/agents/blob/master/tf_agents/colabs/1_dqn_tutorial.ipynb), can you please help me with it’s hyperparameter or if there anything else that need to be changed in the code to train it for Breakout-v0?
    Thanks
    and I would really appreciate your help on this.