Blastar newbind pr

config/ocean/blastar.ini

@@ -3,13 +3,38 @@ package = ocean
 env_name = puffer_blastar 
 policy_name = Policy
 rnn_name = Recurrent
-vec = multiprocessing
 
 [env]
-num_envs = 2048
+num_envs = 4096
 
 [train]
-total_timesteps = 100_000_000
+anneal_lr = False
+batch_size = 131072 # 65536
+bptt_horizon = 8
+checkpoint_interval = 100
+clip_coef = 0.2
+clip_vloss = True
+compile = False
+compile_mode = reduce-overhead
+cpu_offload = False
+data_dir = experiments
+device = cuda
+ent_coef = 0.002511261007200052
+env_batch_size = 1
+gae_lambda = 0.9212875655241286
+gamma = 0.9410283509696092
+learning_rate = 0.0010700459984905535
+max_grad_norm = 0.9702296257019044
+minibatch_size = 4096
+norm_adv = True
+num_envs = 1
+num_workers = 1
+torch_deterministic = True
+total_timesteps = 800000000
+update_epochs = 1
+vf_clip_coef = 0.2
+vf_coef = 0.47285471525325906
+zero_copy = True
 
 [sweep.metric]
 name = environment/enemy_crossed_screen

pufferlib/ocean/blastar/binding.c

@@ -0,0 +1,22 @@
+#include "blastar.h"
+#define Env Blastar
+#include "../env_binding.h"
+
+static int my_init(Env* env, PyObject* args, PyObject* kwargs) {
+    env->num_obs = unpack(kwargs, "num_obs");
+    init(env, env->num_obs);
+    return 0;
+}
+
+static int my_log(PyObject* dict, Log* log) {
+    assign_to_dict(dict, "episode_return", log->episode_return);
+    assign_to_dict(dict, "episode_length", log->episode_length);
+    assign_to_dict(dict, "lives", log->lives);
+    assign_to_dict(dict, "score", log->score);
+    assign_to_dict(dict, "vertical_closeness_rew", log->vertical_closeness_rew);
+    assign_to_dict(dict, "fired_bullet_rew", log->fired_bullet_rew);
+    assign_to_dict(dict, "kill_streak", (float)log->kill_streak);
+    assign_to_dict(dict, "hit_enemy_with_bullet_rew", log->hit_enemy_with_bullet_rew);
+    assign_to_dict(dict, "avg_score_difference", log->avg_score_difference);
+    return 0;
+}

pufferlib/ocean/blastar/blastar.c

@@ -1,18 +1,16 @@
 #include "blastar.h"
-
 #include <assert.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <time.h>
-
 #include "puffernet.h"
 
 const char* WEIGHTS_PATH = "resources/blastar/blastar_weights.bin";
 #define OBSERVATIONS_SIZE 10
 #define ACTIONS_SIZE 6
 #define NUM_WEIGHTS 134407
 
-void get_input(BlastarEnv* env) {
+void get_input(Blastar* env) {
     if (IsKeyDown(KEY_LEFT) || IsKeyDown(KEY_A)) {
         env->actions[0] = 1;  // Left
     } else if (IsKeyDown(KEY_RIGHT) || IsKeyDown(KEY_D)) {
@@ -31,7 +29,7 @@ void get_input(BlastarEnv* env) {
 int demo() {
     Weights* weights = load_weights(WEIGHTS_PATH, NUM_WEIGHTS);
     LinearLSTM* net = make_linearlstm(weights, 1, OBSERVATIONS_SIZE, ACTIONS_SIZE);
-    BlastarEnv env = {
+    Blastar env = {
         .num_obs = OBSERVATIONS_SIZE,
     };
     allocate(&env, env.num_obs);
@@ -47,7 +45,7 @@ int demo() {
             forward_linearlstm(net, env.observations, env.actions);  // AI input
         }
         c_step(&env);
-        c_render(client, &env);
+        c_render(&env);
         if (WindowShouldClose() || env.game_over) {
             running = 0;
         }
@@ -60,7 +58,7 @@ int demo() {
 }
 
 void perftest(float test_time) {
-    BlastarEnv env = {
+    Blastar env = {
         .num_obs = OBSERVATIONS_SIZE,
     };
     allocate(&env, env.num_obs);