Update agents colab to work with latest Dopamine version.

PiperOrigin-RevId: 692944575

Author: psc-g

dopamine/colab/agents.ipynb

@@ -49,9 +49,9 @@
       "source": [
         "# @title Install necessary packages.\n",
         "!pip install -U dopamine-rl\n",
-        "!pip install pandas==0.24.2  # Needed to be able to load the pickle files.\n",
-        "!pip install --upgrade gym\n",
-        "!pip install gym[atari,accept-rom-license]"
+        "!pip install -U gymnasium\n",
+        "!pip install -U baselines\n",
+        "!pip install -U ale_py\n"
       ]
     },
     {
@@ -77,20 +77,6 @@
         "GAME = 'Asterix'  # @param"
       ]
     },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {
-        "cellView": "form",
-        "id": "EFY3tTITHugq"
-      },
-      "outputs": [],
-      "source": [
-        "# @title Load baseline data\n",
-        "!gsutil -q -m cp -R gs://download-dopamine-rl/preprocessed-benchmarks/* /content/\n",
-        "experimental_data = colab_utils.load_baselines('/content')"
-      ]
-    },
     {
       "cell_type": "markdown",
       "metadata": {
@@ -140,6 +126,8 @@
         "run_experiment.Runner.num_iterations = 200\n",
         "run_experiment.Runner.training_steps = 10\n",
         "run_experiment.Runner.max_steps_per_episode = 100\n",
+        "ReplayBuffer.max_capacity = 1_000\n",
+        "ReplayBuffer.batch_size = 32\n",
         "\"\"\".format(GAME)\n",
         "gin.parse_config(random_dqn_config, skip_unknown=False)\n",
         "\n",
@@ -178,8 +166,7 @@
         "    LOG_PATH, verbose=True, summary_keys=['train_episode_returns'])\n",
         "random_dqn_data['agent'] = 'MyRandomDQN'\n",
         "random_dqn_data['run_number'] = 1\n",
-        "experimental_data[GAME] = experimental_data[GAME].merge(random_dqn_data,\n",
-        "                                                        how='outer')"
+        "experimental_data = {GAME:  random_dqn_data}\n"
       ]
     },
     {
@@ -234,6 +221,7 @@
         "  actions with probability switch_prob.\"\"\"\n",
         "  def __init__(self, sess, num_actions, switch_prob=0.1):\n",
         "    self._sess = sess\n",
+        "    self.summary_writer = None\n",
         "    self._num_actions = num_actions\n",
         "    self._switch_prob = switch_prob\n",
         "    self._last_action = np.random.randint(num_actions)\n",
@@ -273,6 +261,8 @@
         "run_experiment.Runner.num_iterations = 200\n",
         "run_experiment.Runner.training_steps = 10\n",
         "run_experiment.Runner.max_steps_per_episode = 100\n",
+        "ReplayBuffer.max_capacity = 1_000\n",
+        "ReplayBuffer.batch_size = 32\n",
         "\"\"\".format(GAME)\n",
         "gin.parse_config(sticky_config, skip_unknown=False)\n",
         "\n",