Reinforcement Learning Module, Part 2 by gabriel-trigo · Pull Request #22 · google/sbsim

Commits on Mar 25, 2025

Commits on Mar 27, 2025

Commits on Apr 11, 2025

fix: change learning rate of td3 algorithm
gabriel-trigo
committed
f831a1b
Revert "fix: change learning rate of td3 algorithm"
gabriel-trigo
committed
e698c6d

Commits on May 30, 2025

reinforcement learning 2, merged with changes to main
gabriel-trigo
committed
0bf3927
fix: fix merge conflicts I forgot
gabriel-trigo
committed
c1b4322

Commits on May 31, 2025

Commits on Jun 4, 2025

fix: fix merge conflicts
gabriel-trigo
committed
a802100

Commits on Jun 6, 2025

Merge branch 'copybara_push' into reinforcement_learning2-gabriel
s2t2
authored
68954e4

Navigation Menu

Uh oh!

Reinforcement Learning Module, Part 2 - #22

Reinforcement Learning Module, Part 2#22
gabriel-trigo wants to merge 26 commits into
google:copybara_pushfrom
gabriel-trigo:reinforcement_learning2-gabriel

Commits on Mar 9, 2025

feat: add reinforcement learning directory. The purpose of this directory is to have all reinforcement learning related code, and scripts to train and evaluate agents

Commits on Mar 10, 2025

Merge branch 'reinforcement_learning-gabriel'

Commits on Mar 23, 2025

chore: update .gitignore to igonore experiment results, replay buffer data

build: add tqdm to project dependencies. Will be used to better monitor progress of RL experiments

feat: add generate_gin_config_files.py script, which takes in a base gin config file, and generates variations changing imporant parameters like time step length, start date and number of days on e…

feat: add ddpg agent implementation to agents directory

feat: add td3 implementation to agents directory

Commits on Mar 25, 2025

feat: improve train.py script. Added support for td3 and ddpg implementations that were added in previous commits. Changed some default argument values, same for populate_starter_buffer.py script

feat: add observer that records and saves trajectories. Also added a visualization module, which this observer uses to plot the graphs

fix: minor bug in print_status_observer.py. Was using the total number of steps instead of episode steps to calculate percentage, leading to > 100% values

feat: add eval.py script, which is used to evaluate a trained policy. Added saved_model_policy.py file to policies directory -- this file has a class which is used to load and interact with policie…

docs: add example bash script to run the populate_starter_buffer script

chore: update .gitignore

Commits on Mar 27, 2025

fix: get rid of step_interval parameter in environment.py (this parameter was redundant with the base_building's time_step_sec

fix: make eval.py script use the latest learned policy checkpoint (before, it was wrongly using the first checkpoint, which gave the untrained agent performance)

tests: fix environment.py tests that were failing to conform to the changes from 111b6e2

Commits on Apr 11, 2025

fix: change learning rate of td3 algorithm

Revert "fix: change learning rate of td3 algorithm"

Commits on May 30, 2025

reinforcement learning 2, merged with changes to main

fix: fix merge conflicts I forgot

Commits on May 31, 2025

fix: forgot to add resolution of merge conflicts to previous commit. This commit fixes that

chore: update gitignore

fix: remove tqdm from isort dependencies in project file

style: styling changes from isort

Commits on Jun 4, 2025

fix: fix merge conflicts

Commits on Jun 6, 2025

Merge branch 'copybara_push' into reinforcement_learning2-gabriel