Commit e6c5b194 authored by umbra99's avatar umbra99

compair configs

parent d63fba40
Pipeline #4951 failed with stage
in 4 minutes and 18 seconds
......@@ -10,8 +10,6 @@ flatland-random-sparse-small-tree-fc-cctransformer:
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
......@@ -20,16 +18,16 @@ flatland-random-sparse-small-tree-fc-cctransformer:
num_sgd_iter: 10
num_workers: 2
num_envs_per_worker: 5
batch_mode: truncate_episodes
observation_filter: NoFilter
batch_mode: complete_episodes
vf_share_layers: True
lr: 0.0001
vf_loss_coeff: 0.000001
num_gpus: 1
env_config:
observation: tree
resolve_deadlocks: false
sparse_reward: True
done_reward: 1
not_finished_reward: -1
observation_config:
max_depth: 2
shortest_path_max_depth: 30
......@@ -70,5 +68,5 @@ flatland-random-sparse-small-tree-fc-cctransformer:
- 512
- 512
fcnet_activation: relu
fcnet_hiddens: [256, 256]
fcnet_hiddens: [512, 512,512]
vf_share_layers: True # False
......@@ -23,18 +23,19 @@ flatland-random-sparse-small-tree-fc-cctransformer:
num_gpus: 1
env_config:
observation: tree
resolve_deadlocks: True
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
entity:
tags: ["small_v0", "tree_obs"] # TODO should be set programmatically
entity: masterscrat
tags: ["small_v0", "tree_obs", "ccppo", "transformer"] # TODO should be set programmatically
model:
custom_model: cc_concatenate
......
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