patonw/tqc-PandaPickAndPlace-v3

TQC Agent playing PandaPickAndPlace-v3

This is a trained model of a TQC agent playing PandaPickAndPlace-v3 using the stable-baselines3 library.

Usage (with Stable-baselines3)

TODO: Add your code

# 1 - 2
env_id = "PandaPickAndPlace-v3"

env = gym.make(env_id)

# 4
from stable_baselines3 import HerReplayBuffer, SAC
model = TQC(policy = "MultiInputPolicy",
            env = env,
            batch_size=2048,
            gamma=0.95,
            learning_rate=1e-4,
            train_freq=64,
            gradient_steps=64,
            tau=0.05,
            replay_buffer_class=HerReplayBuffer,
            replay_buffer_kwargs=dict(
                n_sampled_goal=4,
                goal_selection_strategy="future",
            ),
            policy_kwargs=dict(
                net_arch=[512, 512, 512],
                n_critics=2,
            ),
            tensorboard_log=f"runs/{wandb_run.id}",
           )

# 5
model.learn(1_000_000, progress_bar=True, callback=WandbCallback(verbose=2))
wandb_run.finish()

Weights & Biases charts: https://wandb.ai/patonw/PandaPickAndPlace-v3/runs/w7lzlwnx/workspace?workspace=user-patonw