Polixir Revive

• What is REVIVE SDK?
• Installation

Tutorial

• Quick Start
• Prepare Data
  • Write *.yaml file
  • Constructing Array Data
  • Validation Data (Optional)
• Define Reward Function
• Training Model
• Using Model
  • Using the saved .pkl model
  • Using the saved .onnx model

Task Examples

• Example of applying REVIVE to gym-pendulum
  • Inverted pendulum balancing task description
  • Training Control Strategy with REVIVE SDK
• Example of applying REVIVE to Lander Hover
  • Description
  • Train Control Policy Using REVIVE SDK
  • Showing the Control Effect of Different Policies
• Use REVIVE to control your refrigerator
  • Introduction to Refrigerator Temperature Control System
  • Train refrigerator temperature control policy using REVIVE SDK
  • Testing the Trained Policy Model
• Use REVIVE to control industrial machines
  • Industrial machines controlling description
  • Training control policies using REVIVE SDK
• Controlling Mujoco-HalfCheetah using REVIVE SDK
  • Mujoco-HalfCheetah Task Description
  • Use the REVIVE SDK to train control strategies

Advanced Tools

• Add External Factor
• Add Expert Functions
• Multi-Time-Steps Data Nodes
• Append Time Step Index as Node Input
• Freeze network parameters of some nodes
• Heterogeneous Decision Flow Loading
• Introduction to Empty Nodes
• Loss function of user-defined node
• Generate Decision Flow through Causal Inference
• Customize Network Decision Nodes
  • 1. Configure custom_nodes in .yaml file:
  • 2. Edit Custom Network Node file:
• Introducing expert rule constraints
• Training Control Policies with Multiple Nodes
• Introduction to Multi-Discriminator

Training Results

• Virtual Environment Training Results
  • Interpretation of Metrics in Virtual Environment
  • Rollout Chart Interpretation
• Policy Training Results
  • Interpretation of Double Environment Validation
• Response Curve
  • What is Response Curve
  • Draw Response Curve using REVIVE SDK

Revive API

• revive.server
  • ReviveServer
    • ReviveServer.revive_config_file_path
    • ReviveServer.train()
    • ReviveServer.train_venv()
    • ReviveServer.train_policy()
    • ReviveServer.tune_parameter()
    • ReviveServer.stop_train()
    • ReviveServer.get_virtualenv_env()
    • ReviveServer.get_policy_model()
    • ReviveServer.get_parameter()
• revive.computation
  • VirtualEnv
    • VirtualEnv.to()
    • VirtualEnv.check_version()
    • VirtualEnv.reset()
    • VirtualEnv.set_env()
    • VirtualEnv.target_policy_name
    • VirtualEnv.set_target_policy_name()
    • VirtualEnv.replace_policy()
    • VirtualEnv.infer_k_steps()
    • VirtualEnv.infer_one_step()
    • VirtualEnv.node_pre_computation()
    • VirtualEnv.node_post_computation()
    • VirtualEnv.node_infer()
    • VirtualEnv.node_dist()
    • VirtualEnv.export2onnx()
  • PolicyModel
    • PolicyModel.to()
    • PolicyModel.check_version()
    • PolicyModel.reset()
    • PolicyModel.target_policy_name
    • PolicyModel.infer()
    • PolicyModel.export2onnx()
• revive.conf
  • base_config
    • global_seed
    • val_split_ratio
    • val_split_mode
    • ignore_check
    • venv_rollout_horizon
    • venv_gpus_per_worker
    • venv_metric
    • venv_algo
    • rollout_plt_frequency
    • rollout_dataset_mode
    • policy_gpus_per_worker
    • behavioral_policy_init
    • policy_algo
    • test_horizon
    • train_venv_trials
    • train_policy_trials
  • venv_algo_config
    • revive_p
      • revive_batch_size
      • revive_epoch
      • fintune
      • finetune_fre
      • policy_hidden_features
      • policy_hidden_layers
      • policy_backbone
      • transition_hidden_features
      • transition_hidden_layers
      • transition_backbone
      • matcher_hidden_features
      • matcher_hidden_layers
      • g_steps
      • d_steps
      • g_lr
      • d_lr
    • bc
      • bc_batch_size
      • bc_epoch
      • policy_hidden_features
      • policy_hidden_layers
      • policy_backbone
      • g_lr
      • loss_type
  • policy_algo_config
    • ppo
      • ppo_batch_size
      • ppo_epoch
      • ppo_rollout_horizon
      • policy_hidden_features
      • policy_hidden_layers
      • policy_backbone
      • g_lr

Other

• FAQ
• Releases
  • 0.9.3
  • 0.9.2
  • 0.9.1
  • 0.9.0
  • 0.8.0
  • 0.7.3
  • 0.7.2
  • 0.7.1
  • 0.7.0
  • 0.6.0