[Notbooks] Refresh notebooks (#90)

* amend

* amend

* amend

* amend

Author: matteobettini

notebooks/VMAS_RLlib.ipynb

@@ -44,7 +44,8 @@
     "! git clone https://github.com/proroklab/VectorizedMultiAgentSimulator.git\n",
     "%cd /content/VectorizedMultiAgentSimulator\n",
     "!pip install -e .\n",
-    "!pip install \"ray[rllib]\"==2.2 wandb"
+    "!pip install \"ray[rllib]\"==2.2 wandb\n",
+    "!pip install \"pydantic<2\" numpy==1.23.5"
    ]
   },
   {

notebooks/VMAS_Use_vmas_environment.ipynb

@@ -77,38 +77,43 @@
     "import time\n",
     "import torch\n",
     "from vmas import make_env\n",
+    "from vmas.simulator.core import Agent\n",
+    "\n",
+    "def _get_deterministic_action(agent: Agent, continuous: bool, env):\n",
+    "    if continuous:\n",
+    "        action = -agent.action.u_range_tensor.expand(env.batch_dim, agent.action_size)\n",
+    "    else:\n",
+    "        action = (\n",
+    "            torch.tensor([1], device=env.device, dtype=torch.long)\n",
+    "            .unsqueeze(-1)\n",
+    "            .expand(env.batch_dim, 1)\n",
+    "        )\n",
+    "    return action.clone()\n",
     "\n",
     "def use_vmas_env(\n",
-    "    render: bool = False,\n",
-    "    save_render: bool = False,\n",
-    "    num_envs: int = 32,\n",
-    "    n_steps: int = 100,\n",
-    "    device: str = \"cpu\",\n",
-    "    scenario: Union[str, BaseScenario]= \"waterfall\",\n",
-    "    n_agents: int = 4,\n",
-    "    continuous_actions: bool = True,\n",
+    "    render: bool,\n",
+    "    num_envs: int,\n",
+    "    n_steps: int,\n",
+    "    device: str,\n",
+    "    scenario: Union[str, BaseScenario],\n",
+    "    continuous_actions: bool,\n",
+    "    random_action: bool,\n",
     "    **kwargs\n",
     "):\n",
-    "    \"\"\"Example function to use a vmas environment\n",
+    "    \"\"\"Example function to use a vmas environment.\n",
+    "    \n",
+    "    This is a simplification of the function in `vmas.examples.use_vmas_env.py`.\n",
     "\n",
     "    Args:\n",
     "        continuous_actions (bool): Whether the agents have continuous or discrete actions\n",
-    "        n_agents (int): Number of agents\n",
     "        scenario (str): Name of scenario\n",
     "        device (str): Torch device to use\n",
     "        render (bool): Whether to render the scenario\n",
-    "        save_render (bool):  Whether to save render of the scenario\n",
     "        num_envs (int): Number of vectorized environments\n",
     "        n_steps (int): Number of steps before returning done\n",
-    "\n",
-    "    Returns:\n",
+    "        random_action (bool): Use random actions or have all agents perform the down action\n",
     "\n",
     "    \"\"\"\n",
-    "    assert not (save_render and not render), \"To save the video you have to render it\"\n",
-    "\n",
-    "    simple_2d_action = (\n",
-    "        [0, -1.0] if continuous_actions else [3]\n",
-    "    )  # Simple action for an agent with 2d actions\n",
     "\n",
     "    scenario_name = scenario if isinstance(scenario,str) else scenario.__class__.__name__\n",
     "\n",
@@ -117,10 +122,8 @@
     "        num_envs=num_envs,\n",
     "        device=device,\n",
     "        continuous_actions=continuous_actions,\n",
-    "        wrapper=None,\n",
-    "        seed=None,\n",
+    "        seed=0,\n",
     "        # Environment specific variables\n",
-    "        n_agents=n_agents,\n",
     "        **kwargs\n",
     "    )\n",
     "\n",
@@ -134,31 +137,29 @@
     "\n",
     "        actions = []\n",
     "        for i, agent in enumerate(env.agents):\n",
-    "            action = torch.tensor(\n",
-    "                simple_2d_action,\n",
-    "                device=device,\n",
-    "            ).repeat(num_envs, 1)\n",
+    "            if not random_action:\n",
+    "                action = _get_deterministic_action(agent, continuous_actions, env)\n",
+    "            else:\n",
+    "                action = env.get_random_action(agent)\n",
     "\n",
     "            actions.append(action)\n",
     "\n",
     "        obs, rews, dones, info = env.step(actions)\n",
     "\n",
     "        if render:\n",
     "            frame = env.render(\n",
-    "                mode=\"rgb_array\" if save_render else \"human\",\n",
+    "                mode=\"rgb_array\",\n",
     "                agent_index_focus=None,  # Can give the camera an agent index to focus on\n",
-    "                visualize_when_rgb=True,\n",
     "            )\n",
-    "            if save_render:\n",
-    "                frame_list.append(frame)\n",
+    "            frame_list.append(frame)\n",
     "\n",
     "    total_time = time.time() - init_time\n",
     "    print(\n",
     "        f\"It took: {total_time}s for {n_steps} steps of {num_envs} parallel environments on device {device} \"\n",
     "        f\"for {scenario_name} scenario.\"\n",
     "    )\n",
     "\n",
-    "    if render and save_render:\n",
+    "    if render:\n",
     "        from moviepy.editor import ImageSequenceClip\n",
     "        fps=30\n",
     "        clip = ImageSequenceClip(frame_list, fps=fps)\n",
@@ -177,11 +178,11 @@
     "use_vmas_env(\n",
     "    scenario=scenario_name,\n",
     "    render=True,\n",
-    "    save_render=True,\n",
     "    num_envs=32,\n",
-    "    n_steps=150,\n",
+    "    n_steps=100,\n",
     "    device=\"cuda\",\n",
-    "    continuous_actions=True,\n",
+    "    continuous_actions=False,\n",
+    "    random_action=False,\n",
     "    # Environment specific variables\n",
     "    n_agents=4,\n",
     ")"

vmas/examples/use_vmas_env.py

@@ -20,7 +20,7 @@ def _get_deterministic_action(agent: Agent, continuous: bool, env):
             .unsqueeze(-1)
             .expand(env.batch_dim, 1)
         )
-    return action
+    return action.clone()
 
 
 def use_vmas_env(