Stable dev

config/ocean/cpr.ini

@@ -5,16 +5,63 @@ vec = multiprocessing
 rnn_name = Recurrent
 
 [env]
-num_envs = 256
+num_envs = 512
 vision = 3
-widths = [32]
-heights = [32]
-num_agents = [8]
+num_agents = [12]
 report_interval=1
-reward_food = 1.0
-interactive_food_reward = 5.0
-reward_move = -0.01
+reward_food = 0.1
+interactive_food_reward = 0.2
+reward_move = +0.00
 food_base_spawn_rate = 2e-3
 
 [train]
-total_timesteps = 100_000_000
+num_envs = 1
+num_workers = 1
+total_timesteps = 60_000_000
+device = cpu
+batch_size = 32768
+minibatch_size = 8192
+bptt_horizon = 16
+checkpoint_interval = 200
+learning_rate = 0.0008524
+gamma = 0.9989
+gae_lambda = 0.99
+vf_coef = 1
+ent_coef = 0.01
+adam_beta1 = 0.9
+adam_beta2 = 0.999
+adam_eps = 1e-12
+max_grad_norm = 0.5
+vf_clip_coef = 0.1
+update_epochs = 1
+
+[workspace]
+name = boxingbytes
+project = pufferai
+
+[sweep.metric]
+goal = maximize
+name = score 
+min = -10 
+max = 10 
+
+[sweep.env.reward_food]
+distribution = log_normal
+min = 0.0001
+max = 0.01
+mean = 0.001
+scale = auto 
+
+[sweep.env.interactive_food_reward]
+distribution = log_normal
+min = 0.0001
+max = 0.02
+mean = 0.002
+scale = auto
+
+[sweep.train.total_timesteps]
+distribution = log_normal
+min = 50e6
+max = 75e6
+mean = 60e6
+scale = time

pufferlib/ocean/cpr/cpr.c

@@ -3,16 +3,16 @@
 #include <unistd.h>
 
 int main() {
-  int width = 24;
-  int height = 24;
+  int width = 32;
+  int height = 32;
 
   int render_cell_size = 32;
 
   CCpr env = {
-      .num_agents = 8,
+      .num_agents = 1,
       .width = width,
       .height = height,
-      .vision = 2,
+      .vision = 3,
       .reward_food = 1.0f,
       .interactive_food_reward = 5.0f,
       .food_base_spawn_rate = 2e-3,
@@ -21,9 +21,9 @@ int main() {
   c_reset(&env);
 
   Renderer *renderer = init_renderer(render_cell_size, width, height);
-
   while (!WindowShouldClose()) {
 
+    c_render(renderer, &env);
     int st = 0;
     // User can take control of the first puffer
     if (IsKeyDown(KEY_LEFT_SHIFT)) {
@@ -41,12 +41,10 @@ int main() {
       sleep(2);
     }
     for (int i = st; i < env.num_agents; i++) {
-      env.actions[i] = rand() % 4;
+      env.actions[i] = rand() % 5;
       // printf("Agent %d gets actions %d\n", i, env->actions[i]);
     }
     c_step(&env);
-
-    c_render(renderer, &env);
   }
   close_renderer(renderer);
   free_CCpr(&env);