vardhanshah / control-agent Goto Github PK

usage: main.py [-h] [--id ID] [-tr] [-ts] [-yf YAML_FILE]
[--explore_start EXPLORE_START] [--explore_stop EXPLORE_STOP]
[--decay_rate DECAY_RATE] [-aer AVG_EXPECTED_REWARD]
[--model_save MODEL_SAVE] [--memory_save MEMORY_SAVE]
[-er EPISODE_RENDER] [--max_steps MAX_STEPS]
[--max_episodes MAX_EPISODES]
[--training_frequency TRAINING_FREQUENCY]
[--batch_size BATCH_SIZE] [--pretrain_length PRETRAIN_LENGTH]
[--pretrain_init {random,agent}] [--for_all] [--plot]
[--only_plot] [--total_episodes TOTAL_EPISODES]
[--restore_model] [--restore_memory]
optional arguments:
-h, --help            show this help message and exit
--id ID               give folder id
-tr, --train          train with given folder id
-ts, --test, --evaluate, -ev
test with given folder id, folder id is folder name in
data
-yf YAML_FILE, --yaml_file YAML_FILE
hyper-parameters file
--explore_start EXPLORE_START
exploring probability at start
--explore_stop EXPLORE_STOP
minimum exploring probablity
--decay_rate DECAY_RATE
decaying rate of exploring
-aer AVG_EXPECTED_REWARD, --avg_expected_reward AVG_EXPECTED_REWARD
Average expected reward
--model_save MODEL_SAVE
Model saving after given number of episodes
--memory_save MEMORY_SAVE
Memory saving after given number of episodes
-er EPISODE_RENDER, --episode_render EPISODE_RENDER
Episode Rendering after given number of episodes
--max_steps MAX_STEPS
maximum number of steps for training the agent
--max_episodes MAX_EPISODES
maximum number of episodes for training the agent.
This has higher precedence over max_steps
--training_frequency TRAINING_FREQUENCY
training periodically after given steps in each
episode
--batch_size BATCH_SIZE
batch size of experiences to give Network
--pretrain_length PRETRAIN_LENGTH
pretraining experiences
--pretrain_init {random,agent}
pretraining randomly or with agent
--for_all             Only to be used if testing. this flag will evaluate
all saved models
--plot                Only to be used if testing. this flag will plot
rewards and losses during training
--only_plot           plots rewards and losses during training, give
folder_id
--total_episodes TOTAL_EPISODES
testing will done for given number of episodes
--restore_model       only to be used to restore model for training purpose
--restore_memory      only to be used if saved memory during training, want
to restore that memory