hiro/ray
1
0
Fork 0
mirror of https://github.com/vale981/ray synced 2025-03-06 18:41:40 -05:00
Code Issues Projects Releases Packages Wiki Activity Actions
ray/doc/source/ray-air/examples/rl_online_example.ipynb

1367 lines
44 KiB
Text
Raw Blame History

{
"cells": [
{
"cell_type": "markdown",
"id": "3471e19a",
"metadata": {},
"source": [
"# Online reinforcement learning with Ray AIR\n",
"In this example, we'll train a reinforcement learning agent using online training.\n",
"\n",
"Online training means that the data from the environment is sampled while we are running the algorithm. In contrast, offline training uses data that has been stored on disk before."
]
},
{
"cell_type": "markdown",
"id": "f5083f08",
"metadata": {},
"source": [
"Let's start with installing our dependencies:"
]
},
{
"cell_type": "code",
"execution_count": 1,
"id": "01f914d2",
"metadata": {},
"outputs": [],
"source": [
"!pip install -qU \"ray[rllib]\" gym"
]
},
{
"cell_type": "markdown",
"id": "980cea70",
"metadata": {},
"source": [
"Now we can run some imports:"
]
},
{
"cell_type": "code",
"execution_count": 2,
"id": "db0a45ff",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"2022-05-19 13:54:16,520\tWARNING deprecation.py:47 -- DeprecationWarning: `ray.rllib.execution.buffers` has been deprecated. Use `ray.rllib.utils.replay_buffers` instead. This will raise an error in the future!\n",
"2022-05-19 13:54:16,531\tWARNING deprecation.py:47 -- DeprecationWarning: `ray.rllib.agents.marwil` has been deprecated. Use `ray.rllib.algorithms.marwil` instead. This will raise an error in the future!\n"
]
}
],
"source": [
"import argparse\n",
"import gym\n",
"import os\n",
"\n",
"import numpy as np\n",
"import ray\n",
"from ray.air import Checkpoint\n",
"from ray.air.config import RunConfig\n",
"from ray.train.rl.rl_predictor import RLPredictor\n",
"from ray.train.rl.rl_trainer import RLTrainer\n",
"from ray.air.config import ScalingConfig\n",
"from ray.air.result import Result\n",
"from ray.rllib.agents.marwil import BCTrainer\n",
"from ray.tune.tuner import Tuner"
]
},
{
"cell_type": "markdown",
"id": "a13db7e4",
"metadata": {},
"source": [
"Here we define the training function. It will create an `RLTrainer` using the `PPO` algorithm and kick off training on the `CartPole-v0` environment:"
]
},
{
"cell_type": "code",
"execution_count": 3,
"id": "87fca4b1",
"metadata": {},
"outputs": [],
"source": [
"def train_rl_ppo_online(num_workers: int, use_gpu: bool = False) -> Result:\n",
" print(\"Starting online training\")\n",
" trainer = RLTrainer(\n",
" run_config=RunConfig(stop={\"training_iteration\": 5}),\n",
" scaling_config=ScalingConfig(num_workers=num_workers, use_gpu=use_gpu),\n",
" algorithm=\"PPO\",\n",
" config={\n",
" \"env\": \"CartPole-v0\",\n",
" \"framework\": \"tf\",\n",
" },\n",
" )\n",
" # Todo (krfricke/xwjiang): Enable checkpoint config in RunConfig\n",
" # result = trainer.fit()\n",
" tuner = Tuner(\n",
" trainer,\n",
" _tuner_kwargs={\"checkpoint_at_end\": True},\n",
" )\n",
" result = tuner.fit()[0]\n",
" return result"
]
},
{
"cell_type": "markdown",
"id": "f7a5d5c2",
"metadata": {},
"source": [
"Once we trained our RL policy, we want to evaluate it on a fresh environment. For this, we will also define a utility function:"
]
},
{
"cell_type": "code",
"execution_count": 4,
"id": "2628f3b0",
"metadata": {},
"outputs": [],
"source": [
"def evaluate_using_checkpoint(checkpoint: Checkpoint, num_episodes) -> list:\n",
" predictor = RLPredictor.from_checkpoint(checkpoint)\n",
"\n",
" env = gym.make(\"CartPole-v0\")\n",
"\n",
" rewards = []\n",
" for i in range(num_episodes):\n",
" obs = env.reset()\n",
" reward = 0.0\n",
" done = False\n",
" while not done:\n",
" action = predictor.predict(np.array([obs]))\n",
" obs, r, done, _ = env.step(action[0])\n",
" reward += r\n",
" rewards.append(reward)\n",
"\n",
" return rewards"
]
},
{
"cell_type": "markdown",
"id": "d226d6aa",
"metadata": {},
"source": [
"Let's put it all together. First, we run training:"
]
},
{
"cell_type": "code",
"execution_count": 5,
"id": "cae1337e",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"2022-05-19 13:54:16,582\tWARNING deprecation.py:47 -- DeprecationWarning: `ray.rllib.agents.dqn.dqn.DEFAULT_CONFIG` has been deprecated. Use `ray.rllib.agents.dqn.dqn.DQNConfig(...)` instead. This will raise an error in the future!\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Starting online training\n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2022-05-19 13:54:19,326\tINFO services.py:1483 -- View the Ray dashboard at \u001b[1m\u001b[32mhttp://127.0.0.1:8267\u001b[39m\u001b[22m\n"
]
},
{
"data": {
"text/html": [
"== Status ==<br>Current time: 2022-05-19 13:54:57 (running for 00:00:35.99)<br>Memory usage on this node: 9.6/16.0 GiB<br>Using FIFO scheduling algorithm.<br>Resources requested: 0/16 CPUs, 0/0 GPUs, 0.0/4.54 GiB heap, 0.0/2.0 GiB objects<br>Result logdir: /Users/kai/ray_results/AIRPPOTrainer_2022-05-19_13-54-16<br>Number of trials: 1/1 (1 TERMINATED)<br><table>\n",
"<thead>\n",
"<tr><th>Trial name </th><th>status </th><th>loc </th><th style=\"text-align: right;\"> iter</th><th style=\"text-align: right;\"> total time (s)</th><th style=\"text-align: right;\"> ts</th><th style=\"text-align: right;\"> reward</th><th style=\"text-align: right;\"> episode_reward_max</th><th style=\"text-align: right;\"> episode_reward_min</th><th style=\"text-align: right;\"> episode_len_mean</th></tr>\n",
"</thead>\n",
"<tbody>\n",
"<tr><td>AIRPPOTrainer_cd8d6_00000</td><td>TERMINATED</td><td>127.0.0.1:14174</td><td style=\"text-align: right;\"> 5</td><td style=\"text-align: right;\"> 16.7029</td><td style=\"text-align: right;\">20000</td><td style=\"text-align: right;\"> 124.79</td><td style=\"text-align: right;\"> 200</td><td style=\"text-align: right;\"> 9</td><td style=\"text-align: right;\"> 124.79</td></tr>\n",
"</tbody>\n",
"</table><br><br>"
],
"text/plain": [
"<IPython.core.display.HTML object>"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"\u001b[2m\u001b[33m(raylet)\u001b[0m 2022-05-19 13:54:23,061\tINFO context.py:70 -- Exec'ing worker with command: exec /Users/kai/.pyenv/versions/3.7.7/bin/python3.7 /Users/kai/coding/ray/python/ray/workers/default_worker.py --node-ip-address=127.0.0.1 --node-manager-port=63729 --object-store-name=/tmp/ray/session_2022-05-19_13-54-16_649144_14093/sockets/plasma_store --raylet-name=/tmp/ray/session_2022-05-19_13-54-16_649144_14093/sockets/raylet --redis-address=None --storage=None --temp-dir=/tmp/ray --metrics-agent-port=63909 --logging-rotate-bytes=536870912 --logging-rotate-backup-count=5 --gcs-address=127.0.0.1:65260 --redis-password=5241590000000000 --startup-token=16 --runtime-env-hash=-2010331134\n",
"\u001b[2m\u001b[36m(pid=14174)\u001b[0m 2022-05-19 13:54:30,271\tWARNING deprecation.py:47 -- DeprecationWarning: `ray.rllib.execution.buffers` has been deprecated. Use `ray.rllib.utils.replay_buffers` instead. This will raise an error in the future!\n",
"\u001b[2m\u001b[36m(AIRPPOTrainer pid=14174)\u001b[0m 2022-05-19 13:54:30,749\tINFO trainer.py:1728 -- Your framework setting is 'tf', meaning you are using static-graph mode. Set framework='tf2' to enable eager execution with tf2.x. You may also then want to set eager_tracing=True in order to reach similar execution speed as with static-graph mode.\n",
"\u001b[2m\u001b[36m(AIRPPOTrainer pid=14174)\u001b[0m 2022-05-19 13:54:30,750\tINFO ppo.py:361 -- In multi-agent mode, policies will be optimized sequentially by the multi-GPU optimizer. Consider setting simple_optimizer=True if this doesn't work for you.\n",
"\u001b[2m\u001b[36m(AIRPPOTrainer pid=14174)\u001b[0m 2022-05-19 13:54:30,750\tINFO trainer.py:328 -- Current log_level is WARN. For more information, set 'log_level': 'INFO' / 'DEBUG' or use the -v and -vv flags.\n",
"\u001b[2m\u001b[33m(raylet)\u001b[0m 2022-05-19 13:54:31,857\tINFO context.py:70 -- Exec'ing worker with command: exec /Users/kai/.pyenv/versions/3.7.7/bin/python3.7 /Users/kai/coding/ray/python/ray/workers/default_worker.py --node-ip-address=127.0.0.1 --node-manager-port=63729 --object-store-name=/tmp/ray/session_2022-05-19_13-54-16_649144_14093/sockets/plasma_store --raylet-name=/tmp/ray/session_2022-05-19_13-54-16_649144_14093/sockets/raylet --redis-address=None --storage=None --temp-dir=/tmp/ray --metrics-agent-port=63909 --logging-rotate-bytes=536870912 --logging-rotate-backup-count=5 --gcs-address=127.0.0.1:65260 --redis-password=5241590000000000 --startup-token=17 --runtime-env-hash=-2010331134\n",
"\u001b[2m\u001b[33m(raylet)\u001b[0m 2022-05-19 13:54:31,857\tINFO context.py:70 -- Exec'ing worker with command: exec /Users/kai/.pyenv/versions/3.7.7/bin/python3.7 /Users/kai/coding/ray/python/ray/workers/default_worker.py --node-ip-address=127.0.0.1 --node-manager-port=63729 --object-store-name=/tmp/ray/session_2022-05-19_13-54-16_649144_14093/sockets/plasma_store --raylet-name=/tmp/ray/session_2022-05-19_13-54-16_649144_14093/sockets/raylet --redis-address=None --storage=None --temp-dir=/tmp/ray --metrics-agent-port=63909 --logging-rotate-bytes=536870912 --logging-rotate-backup-count=5 --gcs-address=127.0.0.1:65260 --redis-password=5241590000000000 --startup-token=18 --runtime-env-hash=-2010331134\n",
"\u001b[2m\u001b[36m(RolloutWorker pid=14179)\u001b[0m 2022-05-19 13:54:39,442\tWARNING deprecation.py:47 -- DeprecationWarning: `ray.rllib.execution.buffers` has been deprecated. Use `ray.rllib.utils.replay_buffers` instead. This will raise an error in the future!\n",
"\u001b[2m\u001b[36m(RolloutWorker pid=14180)\u001b[0m 2022-05-19 13:54:39,492\tWARNING deprecation.py:47 -- DeprecationWarning: `ray.rllib.execution.buffers` has been deprecated. Use `ray.rllib.utils.replay_buffers` instead. This will raise an error in the future!\n",
"\u001b[2m\u001b[36m(AIRPPOTrainer pid=14174)\u001b[0m 2022-05-19 13:54:40,836\tINFO trainable.py:163 -- Trainable.setup took 10.087 seconds. If your trainable is slow to initialize, consider setting reuse_actors=True to reduce actor creation overheads.\n",
"\u001b[2m\u001b[36m(AIRPPOTrainer pid=14174)\u001b[0m 2022-05-19 13:54:40,836\tWARNING util.py:65 -- Install gputil for GPU system monitoring.\n",
"\u001b[2m\u001b[36m(AIRPPOTrainer pid=14174)\u001b[0m 2022-05-19 13:54:42,569\tWARNING deprecation.py:47 -- DeprecationWarning: `slice` has been deprecated. Use `SampleBatch[start:stop]` instead. This will raise an error in the future!\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Result for AIRPPOTrainer_cd8d6_00000:\n",
" agent_timesteps_total: 4000\n",
" counters:\n",
" num_agent_steps_sampled: 4000\n",
" num_agent_steps_trained: 4000\n",
" num_env_steps_sampled: 4000\n",
" num_env_steps_trained: 4000\n",
" custom_metrics: {}\n",
" date: 2022-05-19_13-54-44\n",
" done: false\n",
" episode_len_mean: 22.11731843575419\n",
" episode_media: {}\n",
" episode_reward_max: 87.0\n",
" episode_reward_mean: 22.11731843575419\n",
" episode_reward_min: 8.0\n",
" episodes_this_iter: 179\n",
" episodes_total: 179\n",
" experiment_id: 158c57d8b6e142ad85b393db57c8bdff\n",
" hostname: Kais-MacBook-Pro.local\n",
" info:\n",
" learner:\n",
" default_policy:\n",
" custom_metrics: {}\n",
" learner_stats:\n",
" cur_kl_coeff: 0.20000000298023224\n",
" cur_lr: 4.999999873689376e-05\n",
" entropy: 0.6653298139572144\n",
" entropy_coeff: 0.0\n",
" kl: 0.02798665314912796\n",
" model: {}\n",
" policy_loss: -0.0422092080116272\n",
" total_loss: 8.986403465270996\n",
" vf_explained_var: -0.06533512473106384\n",
" vf_loss: 9.023015022277832\n",
" num_agent_steps_trained: 128.0\n",
" num_agent_steps_sampled: 4000\n",
" num_agent_steps_trained: 4000\n",
" num_env_steps_sampled: 4000\n",
" num_env_steps_trained: 4000\n",
" iterations_since_restore: 1\n",
" node_ip: 127.0.0.1\n",
" num_agent_steps_sampled: 4000\n",
" num_agent_steps_trained: 4000\n",
" num_env_steps_sampled: 4000\n",
" num_env_steps_sampled_this_iter: 4000\n",
" num_env_steps_trained: 4000\n",
" num_env_steps_trained_this_iter: 4000\n",
" num_healthy_workers: 2\n",
" off_policy_estimator: {}\n",
" perf:\n",
" cpu_util_percent: 24.849999999999998\n",
" ram_util_percent: 61.199999999999996\n",
" pid: 14174\n",
" policy_reward_max: {}\n",
" policy_reward_mean: {}\n",
" policy_reward_min: {}\n",
" sampler_perf:\n",
" mean_action_processing_ms: 0.06886580197141673\n",
" mean_env_render_ms: 0.0\n",
" mean_env_wait_ms: 0.05465748139159193\n",
" mean_inference_ms: 0.6132523881103351\n",
" mean_raw_obs_processing_ms: 0.10609273714105154\n",
" sampler_results:\n",
" custom_metrics: {}\n",
" episode_len_mean: 22.11731843575419\n",
" episode_media: {}\n",
" episode_reward_max: 87.0\n",
" episode_reward_mean: 22.11731843575419\n",
" episode_reward_min: 8.0\n",
" episodes_this_iter: 179\n",
" hist_stats:\n",
" episode_lengths:\n",
" - 28\n",
" - 9\n",
" - 12\n",
" - 23\n",
" - 13\n",
" - 21\n",
" - 15\n",
" - 16\n",
" - 19\n",
" - 44\n",
" - 14\n",
" - 19\n",
" - 19\n",
" - 17\n",
" - 17\n",
" - 12\n",
" - 9\n",
" - 48\n",
" - 43\n",
" - 15\n",
" - 21\n",
" - 25\n",
" - 16\n",
" - 14\n",
" - 22\n",
" - 21\n",
" - 24\n",
" - 53\n",
" - 21\n",
" - 16\n",
" - 17\n",
" - 14\n",
" - 20\n",
" - 22\n",
" - 18\n",
" - 17\n",
" - 14\n",
" - 11\n",
" - 46\n",
" - 12\n",
" - 18\n",
" - 21\n",
" - 13\n",
" - 58\n",
" - 10\n",
" - 20\n",
" - 14\n",
" - 25\n",
" - 22\n",
" - 33\n",
" - 23\n",
" - 10\n",
" - 25\n",
" - 11\n",
" - 32\n",
" - 48\n",
" - 12\n",
" - 12\n",
" - 10\n",
" - 24\n",
" - 15\n",
" - 28\n",
" - 14\n",
" - 16\n",
" - 14\n",
" - 21\n",
" - 12\n",
" - 13\n",
" - 8\n",
" - 12\n",
" - 13\n",
" - 10\n",
" - 10\n",
" - 14\n",
" - 30\n",
" - 16\n",
" - 23\n",
" - 47\n",
" - 14\n",
" - 22\n",
" - 11\n",
" - 18\n",
" - 12\n",
" - 21\n",
" - 21\n",
" - 20\n",
" - 18\n",
" - 29\n",
" - 18\n",
" - 24\n",
" - 50\n",
" - 87\n",
" - 21\n",
" - 41\n",
" - 21\n",
" - 34\n",
" - 47\n",
" - 20\n",
" - 26\n",
" - 14\n",
" - 9\n",
" - 24\n",
" - 16\n",
" - 18\n",
" - 44\n",
" - 28\n",
" - 37\n",
" - 10\n",
" - 19\n",
" - 11\n",
" - 56\n",
" - 11\n",
" - 28\n",
" - 16\n",
" - 14\n",
" - 19\n",
" - 23\n",
" - 11\n",
" - 22\n",
" - 63\n",
" - 22\n",
" - 13\n",
" - 29\n",
" - 11\n",
" - 64\n",
" - 44\n",
" - 45\n",
" - 38\n",
" - 17\n",
" - 18\n",
" - 21\n",
" - 13\n",
" - 12\n",
" - 13\n",
" - 10\n",
" - 17\n",
" - 14\n",
" - 16\n",
" - 10\n",
" - 19\n",
" - 25\n",
" - 15\n",
" - 50\n",
" - 13\n",
" - 10\n",
" - 15\n",
" - 12\n",
" - 15\n",
" - 11\n",
" - 14\n",
" - 17\n",
" - 17\n",
" - 14\n",
" - 49\n",
" - 18\n",
" - 13\n",
" - 28\n",
" - 31\n",
" - 19\n",
" - 26\n",
" - 31\n",
" - 29\n",
" - 21\n",
" - 23\n",
" - 17\n",
" - 23\n",
" - 32\n",
" - 35\n",
" - 10\n",
" - 11\n",
" - 30\n",
" - 21\n",
" - 16\n",
" - 15\n",
" - 23\n",
" - 40\n",
" - 24\n",
" - 24\n",
" - 14\n",
" episode_reward:\n",
" - 28.0\n",
" - 9.0\n",
" - 12.0\n",
" - 23.0\n",
" - 13.0\n",
" - 21.0\n",
" - 15.0\n",
" - 16.0\n",
" - 19.0\n",
" - 44.0\n",
" - 14.0\n",
" - 19.0\n",
" - 19.0\n",
" - 17.0\n",
" - 17.0\n",
" - 12.0\n",
" - 9.0\n",
" - 48.0\n",
" - 43.0\n",
" - 15.0\n",
" - 21.0\n",
" - 25.0\n",
" - 16.0\n",
" - 14.0\n",
" - 22.0\n",
" - 21.0\n",
" - 24.0\n",
" - 53.0\n",
" - 21.0\n",
" - 16.0\n",
" - 17.0\n",
" - 14.0\n",
" - 20.0\n",
" - 22.0\n",
" - 18.0\n",
" - 17.0\n",
" - 14.0\n",
" - 11.0\n",
" - 46.0\n",
" - 12.0\n",
" - 18.0\n",
" - 21.0\n",
" - 13.0\n",
" - 58.0\n",
" - 10.0\n",
" - 20.0\n",
" - 14.0\n",
" - 25.0\n",
" - 22.0\n",
" - 33.0\n",
" - 23.0\n",
" - 10.0\n",
" - 25.0\n",
" - 11.0\n",
" - 32.0\n",
" - 48.0\n",
" - 12.0\n",
" - 12.0\n",
" - 10.0\n",
" - 24.0\n",
" - 15.0\n",
" - 28.0\n",
" - 14.0\n",
" - 16.0\n",
" - 14.0\n",
" - 21.0\n",
" - 12.0\n",
" - 13.0\n",
" - 8.0\n",
" - 12.0\n",
" - 13.0\n",
" - 10.0\n",
" - 10.0\n",
" - 14.0\n",
" - 30.0\n",
" - 16.0\n",
" - 23.0\n",
" - 47.0\n",
" - 14.0\n",
" - 22.0\n",
" - 11.0\n",
" - 18.0\n",
" - 12.0\n",
" - 21.0\n",
" - 21.0\n",
" - 20.0\n",
" - 18.0\n",
" - 29.0\n",
" - 18.0\n",
" - 24.0\n",
" - 50.0\n",
" - 87.0\n",
" - 21.0\n",
" - 41.0\n",
" - 21.0\n",
" - 34.0\n",
" - 47.0\n",
" - 20.0\n",
" - 26.0\n",
" - 14.0\n",
" - 9.0\n",
" - 24.0\n",
" - 16.0\n",
" - 18.0\n",
" - 44.0\n",
" - 28.0\n",
" - 37.0\n",
" - 10.0\n",
" - 19.0\n",
" - 11.0\n",
" - 56.0\n",
" - 11.0\n",
" - 28.0\n",
" - 16.0\n",
" - 14.0\n",
" - 19.0\n",
" - 23.0\n",
" - 11.0\n",
" - 22.0\n",
" - 63.0\n",
" - 22.0\n",
" - 13.0\n",
" - 29.0\n",
" - 11.0\n",
" - 64.0\n",
" - 44.0\n",
" - 45.0\n",
" - 38.0\n",
" - 17.0\n",
" - 18.0\n",
" - 21.0\n",
" - 13.0\n",
" - 12.0\n",
" - 13.0\n",
" - 10.0\n",
" - 17.0\n",
" - 14.0\n",
" - 16.0\n",
" - 10.0\n",
" - 19.0\n",
" - 25.0\n",
" - 15.0\n",
" - 50.0\n",
" - 13.0\n",
" - 10.0\n",
" - 15.0\n",
" - 12.0\n",
" - 15.0\n",
" - 11.0\n",
" - 14.0\n",
" - 17.0\n",
" - 17.0\n",
" - 14.0\n",
" - 49.0\n",
" - 18.0\n",
" - 13.0\n",
" - 28.0\n",
" - 31.0\n",
" - 19.0\n",
" - 26.0\n",
" - 31.0\n",
" - 29.0\n",
" - 21.0\n",
" - 23.0\n",
" - 17.0\n",
" - 23.0\n",
" - 32.0\n",
" - 35.0\n",
" - 10.0\n",
" - 11.0\n",
" - 30.0\n",
" - 21.0\n",
" - 16.0\n",
" - 15.0\n",
" - 23.0\n",
" - 40.0\n",
" - 24.0\n",
" - 24.0\n",
" - 14.0\n",
" off_policy_estimator: {}\n",
" policy_reward_max: {}\n",
" policy_reward_mean: {}\n",
" policy_reward_min: {}\n",
" sampler_perf:\n",
" mean_action_processing_ms: 0.06886580197141673\n",
" mean_env_render_ms: 0.0\n",
" mean_env_wait_ms: 0.05465748139159193\n",
" mean_inference_ms: 0.6132523881103351\n",
" mean_raw_obs_processing_ms: 0.10609273714105154\n",
" time_since_restore: 3.7304069995880127\n",
" time_this_iter_s: 3.7304069995880127\n",
" time_total_s: 3.7304069995880127\n",
" timers:\n",
" learn_throughput: 2006.2\n",
" learn_time_ms: 1993.819\n",
" load_throughput: 24708712.813\n",
" load_time_ms: 0.162\n",
" training_iteration_time_ms: 3726.731\n",
" update_time_ms: 1.95\n",
" timestamp: 1652964884\n",
" timesteps_since_restore: 0\n",
" timesteps_total: 4000\n",
" training_iteration: 1\n",
" trial_id: cd8d6_00000\n",
" warmup_time: 10.095139741897583\n",
" \n",
"Result for AIRPPOTrainer_cd8d6_00000:\n",
" agent_timesteps_total: 12000\n",
" counters:\n",
" num_agent_steps_sampled: 12000\n",
" num_agent_steps_trained: 12000\n",
" num_env_steps_sampled: 12000\n",
" num_env_steps_trained: 12000\n",
" custom_metrics: {}\n",
" date: 2022-05-19_13-54-51\n",
" done: false\n",
" episode_len_mean: 65.15\n",
" episode_media: {}\n",
" episode_reward_max: 200.0\n",
" episode_reward_mean: 65.15\n",
" episode_reward_min: 9.0\n",
" episodes_this_iter: 44\n",
" episodes_total: 311\n",
" experiment_id: 158c57d8b6e142ad85b393db57c8bdff\n",
" hostname: Kais-MacBook-Pro.local\n",
" info:\n",
" learner:\n",
" default_policy:\n",
" custom_metrics: {}\n",
" learner_stats:\n",
" cur_kl_coeff: 0.30000001192092896\n",
" cur_lr: 4.999999873689376e-05\n",
" entropy: 0.5750519633293152\n",
" entropy_coeff: 0.0\n",
" kl: 0.012749233283102512\n",
" model: {}\n",
" policy_loss: -0.026830431073904037\n",
" total_loss: 9.414541244506836\n",
" vf_explained_var: 0.046859823167324066\n",
" vf_loss: 9.43754768371582\n",
" num_agent_steps_trained: 128.0\n",
" num_agent_steps_sampled: 12000\n",
" num_agent_steps_trained: 12000\n",
" num_env_steps_sampled: 12000\n",
" num_env_steps_trained: 12000\n",
" iterations_since_restore: 3\n",
" node_ip: 127.0.0.1\n",
" num_agent_steps_sampled: 12000\n",
" num_agent_steps_trained: 12000\n",
" num_env_steps_sampled: 12000\n",
" num_env_steps_sampled_this_iter: 4000\n",
" num_env_steps_trained: 12000\n",
" num_env_steps_trained_this_iter: 4000\n",
" num_healthy_workers: 2\n",
" off_policy_estimator: {}\n",
" perf:\n",
" cpu_util_percent: 20.9\n",
" ram_util_percent: 61.379999999999995\n",
" pid: 14174\n",
" policy_reward_max: {}\n",
" policy_reward_mean: {}\n",
" policy_reward_min: {}\n",
" sampler_perf:\n",
" mean_action_processing_ms: 0.06834399059626647\n",
" mean_env_render_ms: 0.0\n",
" mean_env_wait_ms: 0.05423359203664157\n",
" mean_inference_ms: 0.5997818239241897\n",
" mean_raw_obs_processing_ms: 0.0982917359628421\n",
" sampler_results:\n",
" custom_metrics: {}\n",
" episode_len_mean: 65.15\n",
" episode_media: {}\n",
" episode_reward_max: 200.0\n",
" episode_reward_mean: 65.15\n",
" episode_reward_min: 9.0\n",
" episodes_this_iter: 44\n",
" hist_stats:\n",
" episode_lengths:\n",
" - 34\n",
" - 37\n",
" - 38\n",
" - 23\n",
" - 29\n",
" - 56\n",
" - 38\n",
" - 13\n",
" - 10\n",
" - 18\n",
" - 40\n",
" - 23\n",
" - 46\n",
" - 84\n",
" - 29\n",
" - 44\n",
" - 54\n",
" - 32\n",
" - 30\n",
" - 100\n",
" - 28\n",
" - 67\n",
" - 47\n",
" - 40\n",
" - 74\n",
" - 133\n",
" - 32\n",
" - 28\n",
" - 86\n",
" - 133\n",
" - 46\n",
" - 60\n",
" - 17\n",
" - 43\n",
" - 12\n",
" - 51\n",
" - 57\n",
" - 70\n",
" - 54\n",
" - 73\n",
" - 16\n",
" - 29\n",
" - 113\n",
" - 45\n",
" - 31\n",
" - 44\n",
" - 103\n",
" - 62\n",
" - 72\n",
" - 20\n",
" - 15\n",
" - 35\n",
" - 12\n",
" - 9\n",
" - 24\n",
" - 10\n",
" - 102\n",
" - 93\n",
" - 73\n",
" - 27\n",
" - 52\n",
" - 144\n",
" - 19\n",
" - 140\n",
" - 91\n",
" - 133\n",
" - 147\n",
" - 140\n",
" - 90\n",
" - 14\n",
" - 73\n",
" - 71\n",
" - 200\n",
" - 55\n",
" - 184\n",
" - 103\n",
" - 196\n",
" - 168\n",
" - 177\n",
" - 38\n",
" - 33\n",
" - 50\n",
" - 149\n",
" - 67\n",
" - 87\n",
" - 25\n",
" - 134\n",
" - 42\n",
" - 26\n",
" - 24\n",
" - 121\n",
" - 61\n",
" - 109\n",
" - 19\n",
" - 200\n",
" - 60\n",
" - 40\n",
" - 51\n",
" - 88\n",
" - 30\n",
" episode_reward:\n",
" - 34.0\n",
" - 37.0\n",
" - 38.0\n",
" - 23.0\n",
" - 29.0\n",
" - 56.0\n",
" - 38.0\n",
" - 13.0\n",
" - 10.0\n",
" - 18.0\n",
" - 40.0\n",
" - 23.0\n",
" - 46.0\n",
" - 84.0\n",
" - 29.0\n",
" - 44.0\n",
" - 54.0\n",
" - 32.0\n",
" - 30.0\n",
" - 100.0\n",
" - 28.0\n",
" - 67.0\n",
" - 47.0\n",
" - 40.0\n",
" - 74.0\n",
" - 133.0\n",
" - 32.0\n",
" - 28.0\n",
" - 86.0\n",
" - 133.0\n",
" - 46.0\n",
" - 60.0\n",
" - 17.0\n",
" - 43.0\n",
" - 12.0\n",
" - 51.0\n",
" - 57.0\n",
" - 70.0\n",
" - 54.0\n",
" - 73.0\n",
" - 16.0\n",
" - 29.0\n",
" - 113.0\n",
" - 45.0\n",
" - 31.0\n",
" - 44.0\n",
" - 103.0\n",
" - 62.0\n",
" - 72.0\n",
" - 20.0\n",
" - 15.0\n",
" - 35.0\n",
" - 12.0\n",
" - 9.0\n",
" - 24.0\n",
" - 10.0\n",
" - 102.0\n",
" - 93.0\n",
" - 73.0\n",
" - 27.0\n",
" - 52.0\n",
" - 144.0\n",
" - 19.0\n",
" - 140.0\n",
" - 91.0\n",
" - 133.0\n",
" - 147.0\n",
" - 140.0\n",
" - 90.0\n",
" - 14.0\n",
" - 73.0\n",
" - 71.0\n",
" - 200.0\n",
" - 55.0\n",
" - 184.0\n",
" - 103.0\n",
" - 196.0\n",
" - 168.0\n",
" - 177.0\n",
" - 38.0\n",
" - 33.0\n",
" - 50.0\n",
" - 149.0\n",
" - 67.0\n",
" - 87.0\n",
" - 25.0\n",
" - 134.0\n",
" - 42.0\n",
" - 26.0\n",
" - 24.0\n",
" - 121.0\n",
" - 61.0\n",
" - 109.0\n",
" - 19.0\n",
" - 200.0\n",
" - 60.0\n",
" - 40.0\n",
" - 51.0\n",
" - 88.0\n",
" - 30.0\n",
" off_policy_estimator: {}\n",
" policy_reward_max: {}\n",
" policy_reward_mean: {}\n",
" policy_reward_min: {}\n",
" sampler_perf:\n",
" mean_action_processing_ms: 0.06834399059626647\n",
" mean_env_render_ms: 0.0\n",
" mean_env_wait_ms: 0.05423359203664157\n",
" mean_inference_ms: 0.5997818239241897\n",
" mean_raw_obs_processing_ms: 0.0982917359628421\n",
" time_since_restore: 10.289561986923218\n",
" time_this_iter_s: 3.3495230674743652\n",
" time_total_s: 10.289561986923218\n",
" timers:\n",
" learn_throughput: 2276.977\n",
" learn_time_ms: 1756.715\n",
" load_throughput: 20798201.653\n",
" load_time_ms: 0.192\n",
" training_iteration_time_ms: 3425.704\n",
" update_time_ms: 1.814\n",
" timestamp: 1652964891\n",
" timesteps_since_restore: 0\n",
" timesteps_total: 12000\n",
" training_iteration: 3\n",
" trial_id: cd8d6_00000\n",
" warmup_time: 10.095139741897583\n",
" \n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Result for AIRPPOTrainer_cd8d6_00000:\n",
" agent_timesteps_total: 20000\n",
" counters:\n",
" num_agent_steps_sampled: 20000\n",
" num_agent_steps_trained: 20000\n",
" num_env_steps_sampled: 20000\n",
" num_env_steps_trained: 20000\n",
" custom_metrics: {}\n",
" date: 2022-05-19_13-54-57\n",
" done: true\n",
" episode_len_mean: 124.79\n",
" episode_media: {}\n",
" episode_reward_max: 200.0\n",
" episode_reward_mean: 124.79\n",
" episode_reward_min: 9.0\n",
" episodes_this_iter: 20\n",
" episodes_total: 354\n",
" experiment_id: 158c57d8b6e142ad85b393db57c8bdff\n",
" hostname: Kais-MacBook-Pro.local\n",
" info:\n",
" learner:\n",
" default_policy:\n",
" custom_metrics: {}\n",
" learner_stats:\n",
" cur_kl_coeff: 0.30000001192092896\n",
" cur_lr: 4.999999873689376e-05\n",
" entropy: 0.5436986684799194\n",
" entropy_coeff: 0.0\n",
" kl: 0.0034858626313507557\n",
" model: {}\n",
" policy_loss: -0.012989979237318039\n",
" total_loss: 9.49295425415039\n",
" vf_explained_var: 0.025460055097937584\n",
" vf_loss: 9.504897117614746\n",
" num_agent_steps_trained: 128.0\n",
" num_agent_steps_sampled: 20000\n",
" num_agent_steps_trained: 20000\n",
" num_env_steps_sampled: 20000\n",
" num_env_steps_trained: 20000\n",
" iterations_since_restore: 5\n",
" node_ip: 127.0.0.1\n",
" num_agent_steps_sampled: 20000\n",
" num_agent_steps_trained: 20000\n",
" num_env_steps_sampled: 20000\n",
" num_env_steps_sampled_this_iter: 4000\n",
" num_env_steps_trained: 20000\n",
" num_env_steps_trained_this_iter: 4000\n",
" num_healthy_workers: 2\n",
" off_policy_estimator: {}\n",
" perf:\n",
" cpu_util_percent: 24.599999999999998\n",
" ram_util_percent: 59.775\n",
" pid: 14174\n",
" policy_reward_max: {}\n",
" policy_reward_mean: {}\n",
" policy_reward_min: {}\n",
" sampler_perf:\n",
" mean_action_processing_ms: 0.06817872750804764\n",
" mean_env_render_ms: 0.0\n",
" mean_env_wait_ms: 0.05424549075766555\n",
" mean_inference_ms: 0.5976919122059019\n",
" mean_raw_obs_processing_ms: 0.09603803519062176\n",
" sampler_results:\n",
" custom_metrics: {}\n",
" episode_len_mean: 124.79\n",
" episode_media: {}\n",
" episode_reward_max: 200.0\n",
" episode_reward_mean: 124.79\n",
" episode_reward_min: 9.0\n",
" episodes_this_iter: 20\n",
" hist_stats:\n",
" episode_lengths:\n",
" - 45\n",
" - 31\n",
" - 44\n",
" - 103\n",
" - 62\n",
" - 72\n",
" - 20\n",
" - 15\n",
" - 35\n",
" - 12\n",
" - 9\n",
" - 24\n",
" - 10\n",
" - 102\n",
" - 93\n",
" - 73\n",
" - 27\n",
" - 52\n",
" - 144\n",
" - 19\n",
" - 140\n",
" - 91\n",
" - 133\n",
" - 147\n",
" - 140\n",
" - 90\n",
" - 14\n",
" - 73\n",
" - 71\n",
" - 200\n",
" - 55\n",
" - 184\n",
" - 103\n",
" - 196\n",
" - 168\n",
" - 177\n",
" - 38\n",
" - 33\n",
" - 50\n",
" - 149\n",
" - 67\n",
" - 87\n",
" - 25\n",
" - 134\n",
" - 42\n",
" - 26\n",
" - 24\n",
" - 121\n",
" - 61\n",
" - 109\n",
" - 19\n",
" - 200\n",
" - 60\n",
" - 40\n",
" - 51\n",
" - 88\n",
" - 30\n",
" - 200\n",
" - 186\n",
" - 200\n",
" - 182\n",
" - 196\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 43\n",
" - 200\n",
" - 109\n",
" - 156\n",
" - 200\n",
" - 183\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 107\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 89\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 200\n",
" - 200\n",
" episode_reward:\n",
" - 45.0\n",
" - 31.0\n",
" - 44.0\n",
" - 103.0\n",
" - 62.0\n",
" - 72.0\n",
" - 20.0\n",
" - 15.0\n",
" - 35.0\n",
" - 12.0\n",
" - 9.0\n",
" - 24.0\n",
" - 10.0\n",
" - 102.0\n",
" - 93.0\n",
" - 73.0\n",
" - 27.0\n",
" - 52.0\n",
" - 144.0\n",
" - 19.0\n",
" - 140.0\n",
" - 91.0\n",
" - 133.0\n",
" - 147.0\n",
" - 140.0\n",
" - 90.0\n",
" - 14.0\n",
" - 73.0\n",
" - 71.0\n",
" - 200.0\n",
" - 55.0\n",
" - 184.0\n",
" - 103.0\n",
" - 196.0\n",
" - 168.0\n",
" - 177.0\n",
" - 38.0\n",
" - 33.0\n",
" - 50.0\n",
" - 149.0\n",
" - 67.0\n",
" - 87.0\n",
" - 25.0\n",
" - 134.0\n",
" - 42.0\n",
" - 26.0\n",
" - 24.0\n",
" - 121.0\n",
" - 61.0\n",
" - 109.0\n",
" - 19.0\n",
" - 200.0\n",
" - 60.0\n",
" - 40.0\n",
" - 51.0\n",
" - 88.0\n",
" - 30.0\n",
" - 200.0\n",
" - 186.0\n",
" - 200.0\n",
" - 182.0\n",
" - 196.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 43.0\n",
" - 200.0\n",
" - 109.0\n",
" - 156.0\n",
" - 200.0\n",
" - 183.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 107.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 89.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" - 200.0\n",
" off_policy_estimator: {}\n",
" policy_reward_max: {}\n",
" policy_reward_mean: {}\n",
" policy_reward_min: {}\n",
" sampler_perf:\n",
" mean_action_processing_ms: 0.06817872750804764\n",
" mean_env_render_ms: 0.0\n",
" mean_env_wait_ms: 0.05424549075766555\n",
" mean_inference_ms: 0.5976919122059019\n",
" mean_raw_obs_processing_ms: 0.09603803519062176\n",
" time_since_restore: 16.702913284301758\n",
" time_this_iter_s: 3.1872010231018066\n",
" time_total_s: 16.702913284301758\n",
" timers:\n",
" learn_throughput: 2378.661\n",
" learn_time_ms: 1681.619\n",
" load_throughput: 16503261.853\n",
" load_time_ms: 0.242\n",
" training_iteration_time_ms: 3336.7\n",
" update_time_ms: 1.759\n",
" timestamp: 1652964897\n",
" timesteps_since_restore: 0\n",
" timesteps_total: 20000\n",
" training_iteration: 5\n",
" trial_id: cd8d6_00000\n",
" warmup_time: 10.095139741897583\n",
" \n"
]
},
{
"name": "stderr",
"output_type": "stream",
"text": [
"2022-05-19 13:54:58,548\tINFO tune.py:753 -- Total run time: 36.92 seconds (35.95 seconds for the tuning loop).\n"
]
}
],
"source": [
"result = train_rl_ppo_online(num_workers=2, use_gpu=False)"
]
},
{
"cell_type": "markdown",
"id": "6714a3d6",
"metadata": {},
"source": [
"And then, using the obtained checkpoint, we evaluate the policy on a fresh environment:"
]
},
{
"cell_type": "code",
"execution_count": 6,
"id": "b73bfa0f",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"2022-05-19 13:54:58,589\tINFO trainer.py:1728 -- Your framework setting is 'tf', meaning you are using static-graph mode. Set framework='tf2' to enable eager execution with tf2.x. You may also then want to set eager_tracing=True in order to reach similar execution speed as with static-graph mode.\n",
"2022-05-19 13:54:58,590\tWARNING deprecation.py:47 -- DeprecationWarning: `simple_optimizer` has been deprecated. This will raise an error in the future!\n",
"2022-05-19 13:54:58,591\tINFO ppo.py:361 -- In multi-agent mode, policies will be optimized sequentially by the multi-GPU optimizer. Consider setting simple_optimizer=True if this doesn't work for you.\n",
"2022-05-19 13:54:58,591\tINFO trainer.py:328 -- Current log_level is WARN. For more information, set 'log_level': 'INFO' / 'DEBUG' or use the -v and -vv flags.\n",
"\u001b[2m\u001b[36m(RolloutWorker pid=14191)\u001b[0m 2022-05-19 13:55:06,622\tWARNING deprecation.py:47 -- DeprecationWarning: `ray.rllib.execution.buffers` has been deprecated. Use `ray.rllib.utils.replay_buffers` instead. This will raise an error in the future!\n",
"\u001b[2m\u001b[36m(RolloutWorker pid=14192)\u001b[0m 2022-05-19 13:55:06,622\tWARNING deprecation.py:47 -- DeprecationWarning: `ray.rllib.execution.buffers` has been deprecated. Use `ray.rllib.utils.replay_buffers` instead. This will raise an error in the future!\n",
"2022-05-19 13:55:07,968\tWARNING util.py:65 -- Install gputil for GPU system monitoring.\n",
"2022-05-19 13:55:08,021\tINFO trainable.py:589 -- Restored on 127.0.0.1 from checkpoint: /Users/kai/ray_results/AIRPPOTrainer_2022-05-19_13-54-16/AIRPPOTrainer_cd8d6_00000_0_2022-05-19_13-54-22/checkpoint_000005/checkpoint-5\n",
"2022-05-19 13:55:08,021\tINFO trainable.py:597 -- Current state after restoring: {'_iteration': 5, '_timesteps_total': None, '_time_total': 16.702913284301758, '_episodes_total': 354}\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Average reward over 3 episodes: 200.0\n"
]
}
],
"source": [
"num_eval_episodes = 3\n",
"\n",
"rewards = evaluate_using_checkpoint(result.checkpoint, num_episodes=num_eval_episodes)\n",
"print(f\"Average reward over {num_eval_episodes} episodes: \" f\"{np.mean(rewards)}\")"
]
}
],
"metadata": {
"jupytext": {
"cell_metadata_filter": "-all",
"main_language": "python",
"notebook_metadata_filter": "-all"
},
"kernelspec": {
"display_name": "Python 3 (ipykernel)",
"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": "3.7.7"
}
},
"nbformat": 4,
"nbformat_minor": 5
}
Reference in a new issue View git blame Copy permalink