Commit 08ff2ded authored by nilabha's avatar nilabha

Add ccppo configs as part of baselines

parent cdfb8f5d
Pipeline #5000 failed with stage
in 1 minute and 42 seconds
flatland-random-sparse-small-tree-fc-cctransformer:
run: CcTransformer
env: flatland_sparse
stop:
timesteps_total: 15000000 # 1.5e7
checkpoint_freq: 50
checkpoint_at_end: True
keep_checkpoints_num: 100
checkpoint_score_attr: episode_reward_mean
num_samples: 3
config:
clip_rewards: True
clip_param: 0.1
# 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: 2
num_envs_per_worker: 5
batch_mode: complete_episodes
vf_share_layers: True
num_gpus: 1
env_config:
observation: tree
sparse_reward: True
done_reward: 1
not_finished_reward: -1
observation_config:
max_depth: 2
shortest_path_max_depth: 30
generator: sparse_rail_generator
generator_config: small_v0
wandb:
project: flatland-paper
entity: aicrowd
tags: ["small_v0", "tree_obs","ccppo"] # TODO should be set programmatically
model:
custom_model: cc_transformer
custom_options:
max_num_agents: 15
actor:
activation_fn: relu
hidden_layers:
- 512
- 512
- 512
critic:
centralized: True
embedding_size: 32
num_heads: 4
d_model: 32
use_scale: True
activation_fn: relu
hidden_layers:
- 512
- 512
- 512
embedding:
activation_fn: relu
hidden_layers:
- 512
- 512
- 512
fcnet_activation: relu
fcnet_hiddens: [512, 512,512]
vf_share_layers: True # False
flatland-random-sparse-small-tree-fc-cctransformer:
run: CcConcatenate
env: flatland_sparse
stop:
timesteps_total: 15000000 # 1.5e7
checkpoint_freq: 50
checkpoint_at_end: True
keep_checkpoints_num: 100
checkpoint_score_attr: episode_reward_mean
num_samples: 3
config:
clip_rewards: True
clip_param: 0.1
# 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: 2
num_envs_per_worker: 5
batch_mode: complete_episodes
vf_share_layers: True
num_gpus: 1
env_config:
observation: tree
observation_config:
max_depth: 2
shortest_path_max_depth: 30
# skip_no_choice_cells: True
# accumulate_skipped_rewards: True
generator: sparse_rail_generator
generator_config: small_v0
wandb:
project: flatland-paper
entity: aicrowd
tags: ["small_v0", "tree_obs", "ccppo", "transformer"] # TODO should be set programmatically
model:
custom_model: cc_concatenate
custom_options:
max_num_agents: 15
actor:
activation_fn: relu
hidden_layers:
- 512
- 512
- 512
critic:
centralized: True
embedding_size: 32
num_heads: 4
d_model: 32
use_scale: True
activation_fn: relu
hidden_layers:
- 512
- 512
- 512
embedding:
activation_fn: relu
hidden_layers:
- 512
- 512
- 512
fcnet_activation: relu
fcnet_hiddens: [512, 512, 512]
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