test env fixes

Author: jsuarez5341

config/target.ini

@@ -1,10 +1,6 @@
 [base]
 env_name = target
 
-[env]
-width = 1080
-height = 720
-
 [vec]
 total_agents = 4096
 num_buffers = 2

ocean/scape/binding.c

@@ -1,11 +1,10 @@
-#include "target.h"
+#include "scape.h"
 #define OBS_SIZE 28
 #define NUM_ATNS 2
 #define ACT_SIZES {9, 5}
-#define OBS_TYPE FLOAT
-#define ACT_TYPE DOUBLE
+#define OBS_TENSOR_T FloatTensor
 
-#define Env Target
+#define Env Scape
 #include "vecenv.h"
 
 void my_init(Env* env, Dict* kwargs) {

ocean/squared/binding.c

@@ -2,8 +2,7 @@
 #define OBS_SIZE 121
 #define NUM_ATNS 1
 #define ACT_SIZES {5}
-#define OBS_TYPE UNSIGNED_CHAR
-#define ACT_TYPE DOUBLE
+#define OBS_TENSOR_T ByteTensor
 
 #define Env Squared
 #include "vecenv.h"

ocean/squared/squared.h

@@ -24,23 +24,23 @@ typedef struct {
     float score; // Recommended unnormalized single real number perf metric
     float episode_return; // Recommended metric: sum of agent rewards over episode
     float episode_length; // Recommended metric: number of steps of agent episode
-    // Any extra fields you add here may be exported to Python in binding.c
+    // Any extra fields you add here may be exported in binding.c
     float n; // Required as the last field
 } Log;
 
 // Required that you have some struct for your env
-// Recommended that you name it the same as the env file
 typedef struct {
     Log log; // Required field. Env binding code uses this to aggregate logs
     unsigned char* observations; // Required. You can use any obs type, but make sure it matches in Python!
-    double* actions; // Required. double* for new API
+    float* actions; // Required
     float* rewards; // Required
     float* terminals; // Required
     int num_agents;
     int size;
     int tick;
     int r;
     int c;
+    unsigned int rng;
 } Squared;
 
 void add_log(Squared* env) {
@@ -61,7 +61,7 @@ void c_reset(Squared* env) {
     env->tick = 0;
     int target_idx = 0; // Deterministic for testing
     do {
-        target_idx = rand() % tiles;
+        target_idx = rand_r(&env->rng) % tiles;
     } while (target_idx == tiles/2);
     env->observations[target_idx] = TARGET;
 }
@@ -70,7 +70,7 @@ void c_reset(Squared* env) {
 void c_step(Squared* env) {
     env->tick += 1;
 
-    int action = env->actions[0];
+    int action = (int)env->actions[0];
     env->terminals[0] = 0;
     env->rewards[0] = 0;
 

pufferlib/pufferl.py

@@ -492,6 +492,8 @@ def load_config(env_name):
         prev[subkey] = value
 
     args['env_name'] = env_name
+    for section in p.sections():
+        args.setdefault(section, {})
     return dict(args)
 
 def main():