Add double pendulum environment

config/double_pendulum.ini

@@ -0,0 +1,57 @@
+[base]
+env_name = double_pendulum
+
+[vec]
+total_agents = 4096
+num_buffers = 4
+num_threads = 16
+
+[env]
+cart_mass = 1.0
+link1_mass = 0.1
+link2_mass = 0.1
+link1_length = 0.5
+link2_length = 0.5
+gravity = 9.8
+force_mag = 10.0
+dt = 0.02
+
+[policy]
+hidden_size = 128
+num_layers = 2
+expansion_factor = 1
+
+[torch]
+network = MLP
+encoder = DefaultEncoder
+decoder = DefaultDecoder
+
+[train]
+gpus = 1
+seed = 42
+total_timesteps = 20000000
+learning_rate = 0.001
+anneal_lr = 1
+min_lr_ratio = 0
+gamma = 0.99
+gae_lambda = 0.95
+replay_ratio = 1
+clip_coef = 0.2
+vf_coef = 0.5
+vf_clip_coef = 1.0
+max_grad_norm = 0.5
+ent_coef = 0.001
+beta1 = 0.9
+beta2 = 0.999
+eps = 1e-8
+minibatch_size = 16384
+horizon = 64
+vtrace_rho_clip = 1.0
+vtrace_c_clip = 1.0
+prio_alpha = 0.5
+prio_beta0 = 0.5
+
+[sweep]
+method = Protein
+metric = perf
+goal = maximize

ocean/double_pendulum/binding.c

@@ -0,0 +1,33 @@
+#include "double_pendulum.h"
+
+#define OBS_SIZE DP_OBS_SIZE
+#define NUM_ATNS 1
+#define ACT_SIZES {DP_ACTIONS}
+#define OBS_TENSOR_T FloatTensor
+
+#define Env DoublePendulum
+#include "vecenv.h"
+
+void my_init(Env* env, Dict* kwargs) {
+    env->num_agents = 1;
+    env->cart_mass = dict_get(kwargs, "cart_mass")->value;
+    env->link1_mass = dict_get(kwargs, "link1_mass")->value;
+    env->link2_mass = dict_get(kwargs, "link2_mass")->value;
+    env->link1_length = dict_get(kwargs, "link1_length")->value;
+    env->link2_length = dict_get(kwargs, "link2_length")->value;
+    env->gravity = dict_get(kwargs, "gravity")->value;
+    env->force_mag = dict_get(kwargs, "force_mag")->value;
+    env->dt = dict_get(kwargs, "dt")->value;
+    init(env);
+}
+
+void my_log(Log* log, Dict* out) {
+    dict_set(out, "score", log->score);
+    dict_set(out, "perf", log->perf);
+    dict_set(out, "episode_return", log->episode_return);
+    dict_set(out, "episode_length", log->episode_length);
+    dict_set(out, "x_threshold_termination", log->x_threshold_termination);
+    dict_set(out, "max_steps_termination", log->max_steps_termination);
+    dict_set(out, "hold_time", log->hold_time);
+    dict_set(out, "n", log->n);
+}

ocean/double_pendulum/double_pendulum.c

@@ -0,0 +1,38 @@
+#include "double_pendulum.h"
+
+int main(void) {
+    float observations[DP_OBS_SIZE] = {0};
+    float actions[1] = {0};
+    float rewards[1] = {0};
+    float terminals[1] = {0};
+
+    DoublePendulum env = {
+        .observations = observations,
+        .actions = actions,
+        .rewards = rewards,
+        .terminals = terminals,
+        .num_agents = 1,
+        .rng = 1,
+        .cart_mass = 1.0f,
+        .link1_mass = 0.1f,
+        .link2_mass = 0.1f,
+        .link1_length = 0.5f,
+        .link2_length = 0.5f,
+        .gravity = 9.8f,
+        .force_mag = 10.0f,
+        .dt = 0.02f,
+    };
+
+    init(&env);
+    c_reset(&env);
+    c_render(&env);
+    while (!WindowShouldClose()) {
+        if (IsKeyDown(KEY_LEFT) || IsKeyDown(KEY_A)) actions[0] = 0;
+        else if (IsKeyDown(KEY_RIGHT) || IsKeyDown(KEY_D)) actions[0] = 2;
+        else actions[0] = 1;
+        c_step(&env);
+        c_render(&env);
+    }
+    c_close(&env);
+    return 0;
+}