Working common pool resource

config/ocean/cpr.ini

@@ -7,7 +7,7 @@ rnn_name = Recurrent
 [env]
 num_envs = 512
 vision = 3
-num_agents = [12]
+num_agents = [8]
 report_interval=1
 reward_food = 0.1
 interactive_food_reward = 0.2
@@ -19,7 +19,7 @@ num_envs = 1
 num_workers = 1
 total_timesteps = 60_000_000
 device = cpu
-batch_size = 32768
+batch_size = 65536
 minibatch_size = 8192
 bptt_horizon = 16
 checkpoint_interval = 200
@@ -64,4 +64,4 @@ distribution = log_normal
 min = 50e6
 max = 75e6
 mean = 60e6
-scale = time
+scale = time

pufferlib/ocean/cpr/cpr.c → pufferlib/ocean/common_pool_resource/common_pool_resource.c

@@ -9,7 +9,7 @@ int main() {
   int render_cell_size = 32;
 
   CCpr env = {
-      .num_agents = 1,
+      .num_agents = 8,
       .width = width,
       .height = height,
       .vision = 3,

pufferlib/ocean/cpr/cpr.h → pufferlib/ocean/common_pool_resource/cpr.h

@@ -489,7 +489,7 @@ void step_agent(CCpr *env, int i) {
   Agent *agent = &env->agents[i];
 
   int action = env->actions[i];
-  env->logs[i].episode_length += 1;
+  env->log->episode_length += 1;
 
   int dr = 0;
   int dc = 0;
@@ -690,7 +690,7 @@ Renderer *init_renderer(int cell_size, int width, int height) {
   InitWindow(width * cell_size, height * cell_size, "CPR");
   SetTargetFPS(10);
 
-  renderer->puffer = LoadTexture("resources/cpr/puffers_128.png");
+  renderer->puffer = LoadTexture("resources/common_pool_resource/puffers_128.png");
   return renderer;
 }
 

pufferlib/ocean/cpr/cpr.py → pufferlib/ocean/common_pool_resource/cpr.py

@@ -2,7 +2,7 @@
 import numpy as np 
 
 import pufferlib 
-from pufferlib.ocean.cpr.cy_cpr import CyEnv
+from pufferlib.ocean.common_pool_resource.cy_cpr import CyEnv
 
 class PyCPR(pufferlib.PufferEnv):
     def __init__(self, 
@@ -85,12 +85,12 @@ def close(self):
 
     import time 
     start = time.time()
-    # while time.time() - start < timeout:
-    while tick < 500:
+    while time.time() - start < timeout:
+    # while tick < 500:
         atns = actions[tick % 1024]
         env.step(atns)
-        if -1 in env.rewards:
-            breakpoint()
+        # if -1 in env.rewards:
+        #     breakpoint()
         # env.render()
         tick += 1
 

pufferlib/ocean/environment.py

@@ -135,7 +135,7 @@ def make_multiagent(buf=None, **kwargs):
     'trash_pickup':  lambda: lazy_import('pufferlib.ocean.trash_pickup.trash_pickup', 'TrashPickupEnv'),
     'tower_climb':   lambda: lazy_import('pufferlib.ocean.tower_climb.tower_climb', 'TowerClimb'),
     'grid':          lambda: lazy_import('pufferlib.ocean.grid.grid', 'Grid'),
-    'cpr':           lambda: lazy_import('pufferlib.ocean.cpr.cpr', 'PyCPR'),
+    'cpr':           lambda: lazy_import('pufferlib.ocean.common_pool_resource.cpr', 'PyCPR'),
     'impulse_wars':  lambda: lazy_import('pufferlib.ocean.impulse_wars.impulse_wars', 'ImpulseWars'),
     'gpudrive':      lambda: lazy_import('pufferlib.ocean.gpudrive.gpudrive', 'GPUDrive'),
     #'rocket_lander': rocket_lander.RocketLander,

setup.py

@@ -280,7 +280,8 @@
     # 'pufferlib/ocean/go/cy_go',
     'pufferlib/ocean/rware/cy_rware',
     'pufferlib/ocean/trash_pickup/cy_trash_pickup',
-    'pufferlib/ocean/cpr/cy_cpr',
+    'pufferlib/ocean/common_pool_resource/cy_cpr',
+    # 'pufferlib/ocean/tower_climb/cy_tower_climb',
     'pufferlib/ocean/tower_climb/cy_tower_climb',
     'pufferlib/ocean/gpudrive/cy_gpudrive',
 ]