Commit 050238d1 authored by MasterScrat's avatar MasterScrat

Added local conflict config using new syntax

parent c42a05a6
......@@ -49,9 +49,11 @@ class FlatlandRllibWrapper(object):
for agent, done in dones.items():
if agent != '__all__' and not agent in obs:
continue # skip agent if there is no observation
# FIXME the check below should be kept in MARL training
#if agent not in self._agents_done:
if True or agent not in self._agents_done:
# Use this if using a single policy for multiple agents
# TODO find better way to handle this
#if True or agent not in self._agents_done:
if agent not in self._agents_done:
if agent != '__all__':
if done:
self._agents_done.append(agent)
......
flatland-random-sparse-small-local-conflict-fc-ppo:
run: PPO
env: flatland_sparse
stop:
timesteps_total: 10000000 # 1e7
checkpoint_freq: 10
checkpoint_at_end: True
keep_checkpoints_num: 5
checkpoint_score_attr: episode_reward_mean
config:
clip_rewards: True
clip_param: 0.1
vf_clip_param: 500.0
entropy_coeff: 0.01
# effective batch_size: train_batch_size * num_agents_in_each_environment [5, 10]
# see https://github.com/ray-project/ray/issues/4628
train_batch_size: 1000 # 5000
rollout_fragment_length: 50 # 100
sgd_minibatch_size: 100 # 500
num_sgd_iter: 10
num_workers: 11
num_envs_per_worker: 1
batch_mode: truncate_episodes
observation_filter: NoFilter
vf_share_layers: True
vf_loss_coeff: 0.5
num_gpus: 1
env_config:
observation: localConflict
observation_config:
max_depth: 2
shortest_path_max_depth: 30
n_local: 5
generator: sparse_rail_generator
generator_config: small_v0
wandb:
project: flatland
entity: masterscrat
tags: ["small_v0", "local_conflict_obs"] # TODO should be set programmatically
model:
fcnet_activation: relu
fcnet_hiddens: [256, 256]
vf_share_layers: True # False
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment