add PickAndPlace training

examples/PickAndPlace_train.ipynb

@@ -0,0 +1,107 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "%pip install stable-baselines3"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Train agent"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import gymnasium as gym\n",
+    "from stable_baselines3 import DDPG, SAC, HerReplayBuffer\n",
+    "\n",
+    "import panda_gym\n",
+    "\n",
+    "env = gym.make(\"PandaPickAndPlace\", render_mode=\"human\")\n",
+    "\n",
+    "model = SAC(\n",
+    "    policy=\"MultiInputPolicy\", \n",
+    "    env=env, \n",
+    "    replay_buffer_class=HerReplayBuffer, \n",
+    "    verbose=1, \n",
+    "    buffer_size=1000000,\n",
+    "    replay_buffer_kwargs=dict(\n",
+    "        n_sampled_goal=4,\n",
+    "        goal_selection_strategy='future',\n",
+    "    ),\n",
+    "    gamma=0.95,\n",
+    "    learning_starts=1000,\n",
+    "    train_freq=1\n",
+    "    )\n",
+    "\n",
+    "model.learn(total_timesteps=4000000)\n",
+    "model.save(\"./her_robot\")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### Evaluate"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import gymnasium as gym\n",
+    "from stable_baselines3 import SAC, HerReplayBuffer\n",
+    "import panda_gym\n",
+    "from time import sleep\n",
+    "\n",
+    "env = gym.make(\"PandaPickAndPlace\", render_mode=\"human\")\n",
+    "\n",
+    "model = SAC.load(\"./her_robot\", env=env)\n",
+    "vec_env = model.get_env()\n",
+    "obs = vec_env.reset()\n",
+    "for i in range(10000):\n",
+    "    sleep(0.08)\n",
+    "    action, _states = model.predict(obs, deterministic=True)\n",
+    "    obs, reward, done, info = vec_env.step(action)\n",
+    "    vec_env.render()\n",
+    "\n",
+    "    if done:\n",
+    "       obs = vec_env.reset()"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.9.6"
+  },
+  "orig_nbformat": 4
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}