[Docs] Premade scripts

Signed-off-by: Matteo Bettini <[email protected]>

README.md

@@ -235,6 +235,14 @@ And the ones that are _work in progress_
 | [GNN](https://github.com/facebookresearch/BenchMARL/pull/18) |       ✅       |               ✅               |               ❌               | 
 | CNN                                                          |       ✅       |               ✅               |               ✅               | 
 
+## Premade scripts
+> [!WARNING]  
+> This section is under a work in progress. We are constantly working on fine-tuning
+> our algorithms to enable our users to have access to state-of-the-art benchmarks.
+> If you would like to collaborate in this effort, please reach out to us.
+
+In the [premade_scripts](premade_scripts) folder we are collecting some tested hyperparameters for
+specific environments to enable users to bootstrap their benchmarking.
 
 ## Reporting and plotting
 

premade_scripts/smacv2/config.yaml

@@ -14,5 +14,40 @@ hydra:
 seed: 0
 
 experiment:
-    sampling_device: "cuda"
-    train_device: "cuda"
+  sampling_device: "cuda"
+  train_device: "cuda"
+
+  share_policy_params: True
+  prefer_continuous_actions: True
+
+  gamma: 0.99
+  lr: 0.00005
+  clip_grad_norm: True
+  clip_grad_val: 5
+
+  soft_target_update: True
+  polyak_tau: 0.005
+  hard_target_update_frequency: 5
+
+  exploration_eps_init: 0.8
+  exploration_eps_end: 0.01
+  exploration_anneal_frames: null
+
+  max_n_iters: null
+  max_n_frames: 30_000_000
+
+  on_policy_collected_frames_per_batch: 6000
+  on_policy_n_envs_per_worker: 10
+  on_policy_n_minibatch_iters: 45
+  on_policy_minibatch_size: 400
+
+  off_policy_collected_frames_per_batch: 6000
+  off_policy_n_envs_per_worker: 10
+  off_policy_n_optimizer_steps: 1000
+  off_policy_train_batch_size: 100
+  off_policy_memory_size: 1_000_000
+
+  evaluation: True
+  render: True
+  evaluation_interval: 50
+  evaluation_episodes: 20

premade_scripts/vmas/config.yaml

@@ -14,9 +14,41 @@ hydra:
 seed: 0
 
 experiment:
-    sampling_device: "cuda"
-    train_device: "cuda"
-    collected_frames_per_batch: 60_000
-    n_envs_per_worker: 600
-    on_policy_minibatch_size: 4096
-    evaluation_episodes: 200
+
+  sampling_device: "cuda"
+  train_device: "cuda"
+
+  share_policy_params: True
+  prefer_continuous_actions: True
+
+  gamma: 0.9
+  lr: 0.00005
+  clip_grad_norm: True
+  clip_grad_val: 5
+
+  soft_target_update: True
+  polyak_tau: 0.005
+  hard_target_update_frequency: 5
+
+  exploration_eps_init: 0.8
+  exploration_eps_end: 0.01
+  exploration_anneal_frames: 1_000_000
+
+  max_n_iters: null
+  max_n_frames: 10_000_000
+
+  on_policy_collected_frames_per_batch: 60_000
+  on_policy_n_envs_per_worker: 600
+  on_policy_n_minibatch_iters: 45
+  on_policy_minibatch_size: 4096
+
+  off_policy_collected_frames_per_batch: 6000
+  off_policy_n_envs_per_worker: 60
+  off_policy_n_optimizer_steps: 6000
+  off_policy_train_batch_size: 128
+  off_policy_memory_size: 1_000_000
+
+  evaluation: True
+  render: True
+  evaluation_interval: 20
+  evaluation_episodes: 200