fix sweep , breakout config for sweeps

Author: Tim-phant

config/boxoban.ini

@@ -2,9 +2,9 @@
 env_name = boxoban
 
 [vec]
-total_agents = 32768
-num_buffers = 4
-num_threads = 1
+total_agents = 4096
+num_buffers = 8
+num_threads = 8
 
 [env]
 num_agents = 1
@@ -14,31 +14,32 @@ target_loss_pen_coeff = 0
 max_steps = 150
 
 [policy]
-hidden_size = 512
-num_layers = 3.32422
-expansion_factor = 1
+num_layers = 2
+hidden_size = 64
 
 [train]
 gpus = 1
-seed = 42
-total_timesteps = 880580001
-learning_rate = 0.00134234
-anneal_lr = 1
-min_lr_ratio = 0.37872
-gamma = 0.989717
-gae_lambda = 0.759273
-replay_ratio = 1.6234
-clip_coef = 0.01
-vf_coef = 5
-vf_clip_coef = 5
-max_grad_norm = 1.20325
-ent_coef = 0.000188411
-beta1 = 0.995526
-beta2 = 0.999536
-eps = 1e-14
-minibatch_size = 32768
-horizon = 64
-vtrace_rho_clip = 3.13347
-vtrace_c_clip = 2.75328
-prio_alpha = 0.453827
-prio_beta0 = 0.765589
+total_timesteps = 94_000_000
+beta1 = 0.7279714073125252
+beta2 = 0.9986265112492152
+clip_coef = 0.6746497927896418
+ent_coef = 0.0033240721522812535
+eps = 0.00008339460257113628
+gae_lambda = 0.948721675814334
+gamma = 0.9721246598992744
+learning_rate = 0.1
+max_grad_norm = 1.8109182724544075
+minibatch_size = 65_536
+prio_alpha = 0.1
+prio_beta0 = 0.8247156461060179
+replay_ratio = 1.4242098997083206
+vf_clip_coef = 1.2291681640124468
+vf_coef = 1.2195502588297364
+vtrace_c_clip = 1.0830442742115065
+vtrace_rho_clip = 2.1017317041552603
+
+[sweep]
+gpus = 1
+max_suggestion_cost = 10800
+
+

pufferlib/sweep.py

@@ -146,7 +146,8 @@ def _params_from_puffer_sweep(sweep_config, only_include=None):
 
     for name, param in sweep_config.items():
         if name in ('method', 'metric', 'metric_distribution', 'goal', 'downsample', 'use_gpu', 'prune_pareto',
-                    'sweep_only', 'max_suggestion_cost', 'early_stop_quantile', 'gpus', 'max_runs'):
+                    'sweep_only', 'max_suggestion_cost', 'early_stop_quantile', 'gpus', 'max_runs', 'match_enemy_model_path', 'match_num_games',
+                  'match_enemy_hidden_size', 'match_enemy_num_layers'):
             continue
 
         assert isinstance(param, dict), f'Param {name} is not a dict'