cleaned up

Author: paulmerceur

pufferlib/config/ocean/four_rooms.ini

@@ -4,86 +4,11 @@ env_name = puffer_four_rooms
 policy_name = Policy
 rnn_name = Recurrent
 
-[vec]
-num_workers = 12
-num_envs = 12
-
 [env]
-num_envs = 100
+num_envs = 256
 
 [train]
 total_timesteps = 10_000_000
 gamma = 0.99
 learning_rate = 0.015
 minibatch_size = 32768
-device = mps
-
-[sweep]
-method = Protein
-metric = perf
-goal = maximize
-downsample = 0
-
-[sweep.train.learning_rate]
-distribution = log_normal
-min = 0.001
-max = 0.1
-mean = 0.015
-scale = 0.5
-
-[sweep.train.gamma]
-distribution = uniform
-min = 0.95
-max = 0.999
-mean = 0.99
-scale = auto
-
-[sweep.train.total_timesteps]
-distribution = log_normal
-min = 5_000_000
-max = 50_000_000
-mean = 10_000_000
-scale = 0.5
-
-[sweep.train.minibatch_size]
-distribution = uniform_pow2
-min = 8192
-max = 65536
-mean = 32768
-scale = auto
-
-[sweep.train.ent_coef]
-distribution = log_normal
-min = 0.0001
-max = 0.01
-mean = 0.001
-scale = 0.5
-
-[sweep.train.clip_coef]
-distribution = uniform
-min = 0.1
-max = 0.3
-mean = 0.2
-scale = auto
-
-[sweep.train.vf_coef]
-distribution = uniform
-min = 0.5
-max = 4.0
-mean = 2.0
-scale = auto
-
-[sweep.env.size]
-distribution = int_uniform
-min = 13
-max = 25
-mean = 19
-scale = auto
-
-[sweep.policy.hidden_size]
-distribution = uniform_pow2
-min = 64
-max = 512
-mean = 128
-scale = auto
-

pufferlib/ocean/four_rooms/binding.c

@@ -5,7 +5,7 @@
 
 static int my_init(Env* env, PyObject* args, PyObject* kwargs) {
     env->size = unpack(kwargs, "size");
-    env->see_through_walls = 0; // Default to false for MinGrid compatibility
+    env->see_through_walls = 0;
     // Allocate grid memory for full state (stores OBJECT_IDX values)
     env->grid = (unsigned char*)calloc(env->size * env->size, sizeof(unsigned char));
     return 0;

pufferlib/ocean/four_rooms/four_rooms.h

@@ -63,7 +63,7 @@ typedef struct {
 } FourRooms;
 
 void add_log(FourRooms* env) {
-    env->log.perf += (env->rewards[0] > 0) ? 1 : 0;
+    env->log.perf += (env->rewards[0] > 0) ? 1.0 : 0.0;
     env->log.score += env->rewards[0];
     env->log.episode_length += env->tick;
     env->log.episode_return += env->rewards[0];