# Debugging and performance tuning **URL:** https://discuss.ray.io/c/rllib/debugging-and-performance-tuning/24.md [Latest](https://discuss.ray.io/latest.md) · [Categories](https://discuss.ray.io/categories.md) · [Tags](https://discuss.ray.io/tags.md) --- ## [About the Debugging and performance tuning category](https://discuss.ray.io/t/about-the-debugging-and-performance-tuning-category/7743)
**Author:** [@christy](https://discuss.ray.io/u/christy)\ **Replies:** 0
--- ## [RLlib benchmarking](https://discuss.ray.io/t/rllib-benchmarking/23378)
**Author:** [@Diplomat](https://discuss.ray.io/u/Diplomat)\ **Replies:** 1\ **Last updated:** [December 17, 2025, 11:52am UTC](https://discuss.ray.io/t/rllib-benchmarking/23378 "2025-12-17T11:52:57Z")
I’m searching for RLlib benchmark results. (I’d simply like to validate e.g. PPO on Mujoco’s Walker2d where I experience significantly inferior RLlib performance compared to e.g. SB3) GitHub - ray-project/rl-experimen… --- ## [Extremely oscillating MAPPO reward on custom env](https://discuss.ray.io/t/extremely-oscillating-mappo-reward-on-custom-env/15573)
**Author:** [@Muhab\_Abubaker](https://discuss.ray.io/u/Muhab_Abubaker)\ **Replies:** 1\ **Last updated:** [August 12, 2025, 8:12pm UTC](https://discuss.ray.io/t/extremely-oscillating-mappo-reward-on-custom-env/15573 "2025-08-12T20:12:55Z")
Hi everyone, I am currently working on a custom PettingZoo environment with multiple agents trying to control a spreading wildfire by applying a fire retardant action. Here is a sample of the environment grid-world (the… --- ## [Error: No available node types can fulfill resource request defaultdict(\, {'GPU': 8.0, 'CPU': 8.0, 'priority': 16.0})](https://discuss.ray.io/t/error-no-available-node-types-can-fulfill-resource-request-defaultdict-class-float-gpu-8-0-cpu-8-0-priority-16-0/22410)
**Author:** [@evangeline](https://discuss.ray.io/u/evangeline)\ **Replies:** 1\ **Last updated:** [May 1, 2025, 9:12pm UTC](https://discuss.ray.io/t/error-no-available-node-types-can-fulfill-resource-request-defaultdict-class-float-gpu-8-0-cpu-8-0-priority-16-0/22410 "2025-05-01T21:12:00Z")
Job submission server address: http://127.0.0.1:8265 \[JobDetails(type=\, job\_id=‘01000000’, submission\_id=None, driver\_info=DriverInfo(id=‘01000000’, node\_ip\_address=‘172.22.225.33’, pid=‘346… --- ## [PPO: GPU available, but not utilized](https://discuss.ray.io/t/ppo-gpu-available-but-not-utilized/21546)
**Author:** [@Magnus411](https://discuss.ray.io/u/Magnus411)\ **Replies:** 4\ **Last updated:** [April 1, 2025, 7:10am UTC](https://discuss.ray.io/t/ppo-gpu-available-but-not-utilized/21546 "2025-04-01T07:10:20Z")
Hello! I am quite new to using RLlib, but have managed to set up my I’m relatively new to using RLlib but have managed to set up a training pipeline using the PPO algorithm. My system has a GPU (NVIDIA RTX 3080), and RL… --- ## [Encountering the Tracked Actor not managed by this event error](https://discuss.ray.io/t/encountering-the-tracked-actor-not-managed-by-this-event-error/21021)
**Author:** [@maulish34](https://discuss.ray.io/u/maulish34)\ **Replies:** 0\ **Last updated:** [December 11, 2024, 5:14am UTC](https://discuss.ray.io/t/encountering-the-tracked-actor-not-managed-by-this-event-error/21021 "2024-12-11T05:14:20Z")
i am facing the following error while tuning using ASHAScheduler: ValueError: Tracked actor is not managed by this event manager: \ I have changed the code a bit fro… --- ## [RLlib perform worse when rollout\_worker/env\_runner increased?](https://discuss.ray.io/t/rllib-perform-worse-when-rollout-worker-env-runner-increased/20315)
**Author:** [@Morphlng](https://discuss.ray.io/u/Morphlng)\ **Replies:** 0\ **Last updated:** [November 1, 2024, 3:10am UTC](https://discuss.ray.io/t/rllib-perform-worse-when-rollout-worker-env-runner-increased/20315 "2024-11-01T03:10:39Z")
I’m using Ray 2.8.1 (I’ve also tried the latest 2.38.0), and it seems like RLlib’s performance is worse when adding rollout workers (env\_runners). The above results were trained with 2.8.1 by the following script: f… --- ## [How to Log Episode Value Targets in New API](https://discuss.ray.io/t/how-to-log-episode-value-targets-in-new-api/15967)
**Author:** [@o17](https://discuss.ray.io/u/o17)\ **Replies:** 0\ **Last updated:** [October 8, 2024, 4:15pm UTC](https://discuss.ray.io/t/how-to-log-episode-value-targets-in-new-api/15967 "2024-10-08T16:15:30Z")
How severe does this issue affect your experience of using Ray? None I am trying to log the value targets for an episode. Is there a way to access or compute that on the new API stack? --- ## [Process keep stuck in pending status](https://discuss.ray.io/t/process-keep-stuck-in-pending-status/15919)
**Author:** [@Hawwwkim](https://discuss.ray.io/u/Hawwwkim)\ **Replies:** 0\ **Last updated:** [September 30, 2024, 6:20am UTC](https://discuss.ray.io/t/process-keep-stuck-in-pending-status/15919 "2024-09-30T06:20:26Z")
I’m new to here. I’m trying to train a muti-agent reinforce learning model, with an environment created with PettingZoo. But in training, when I try to use tune.grid\_search the gpu in my device seems not in using, and th… --- ## [PPO: Value estimate off for goal state](https://discuss.ray.io/t/ppo-value-estimate-off-for-goal-state/15841)
**Author:** [@matthias-brucklacher](https://discuss.ray.io/u/matthias-brucklacher)\ **Replies:** 0\ **Last updated:** [September 18, 2024, 4:55pm UTC](https://discuss.ray.io/t/ppo-value-estimate-off-for-goal-state/15841 "2024-09-18T16:55:50Z")
Hi! I have some issues getting PPO to run on my custom environment. I observe that the value estimate at the goal state is negative (while it is correctly positive and large at the preceding state) and suspect that this … --- ## [NotImplementedError: Unsupported args](https://discuss.ray.io/t/notimplementederror-unsupported-args/15802)
**Author:** [@InLine6261947](https://discuss.ray.io/u/InLine6261947)\ **Replies:** 0\ **Last updated:** [September 13, 2024, 6:30pm UTC](https://discuss.ray.io/t/notimplementederror-unsupported-args/15802 "2024-09-13T18:30:35Z")
I am trying to implement a MAPPO Multi-Agent boid environment. Using Gym and Ray Rllib. Whenever I run the training file it gives the error (PPO pid=24516) NotImplementedError: Unsupported args: Box(-5.0, 5.0, (2,), f… --- ## ['MultiAgentBatch' object has no attribute 'get' when using DQN and storing sequences in the Replay Buffer](https://discuss.ray.io/t/multiagentbatch-object-has-no-attribute-get-when-using-dqn-and-storing-sequences-in-the-replay-buffer/14665)
**Author:** [@Faptimus420](https://discuss.ray.io/u/Faptimus420)\ **Replies:** 0\ **Last updated:** [May 11, 2024, 10:57pm UTC](https://discuss.ray.io/t/multiagentbatch-object-has-no-attribute-get-when-using-dqn-and-storing-sequences-in-the-replay-buffer/14665 "2024-05-11T22:57:48Z")
How severe does this issue affect your experience of using Ray? Medium: It contributes to significant difficulty to complete my task, but I can work around it. I’m working with an older version of Ray here (2.4.0), so… --- ## [Getting errors while using documentation sample codes](https://discuss.ray.io/t/getting-errors-while-using-documentation-sample-codes/14473)
**Author:** [@77asadian](https://discuss.ray.io/u/77asadian)\ **Replies:** 0\ **Last updated:** [April 22, 2024, 9:21am UTC](https://discuss.ray.io/t/getting-errors-while-using-documentation-sample-codes/14473 "2024-04-22T09:21:36Z")
Hay, I was trying to run this code on my PC but some problems occured. import ray from ray import train, tune ray.init() config = PPOConfig().training(lr=tune.grid\_search(\[0.01, 0.001, 0.0001\])) tuner = tune.Tuner( … --- ## [APPO Learner spent really long time in sampling/deserialization](https://discuss.ray.io/t/appo-learner-spent-really-long-time-in-sampling-deserialization/13784)
**Author:** [@Ran\_Cao](https://discuss.ray.io/u/Ran_Cao)\ **Replies:** 1\ **Last updated:** [February 22, 2024, 7:18pm UTC](https://discuss.ray.io/t/appo-learner-spent-really-long-time-in-sampling-deserialization/13784 "2024-02-22T19:18:04Z")
How severe does this issue affect your experience of using Ray? High: It blocks me to complete my task. I’m doing an RL training locally using Rllib and Unreal in windows, I’m using APPO agent and eventually will sen… --- ## [Build\_for\_inference() in env\_runner\_v2.py created empty state\_out\_1 and lead to failure of initiation](https://discuss.ray.io/t/build-for-inference-in-env-runner-v2-py-created-empty-state-out-1-and-lead-to-failure-of-initiation/13636)
**Author:** [@Qin](https://discuss.ray.io/u/Qin)\ **Replies:** 1\ **Last updated:** [February 5, 2024, 3:44pm UTC](https://discuss.ray.io/t/build-for-inference-in-env-runner-v2-py-created-empty-state-out-1-and-lead-to-failure-of-initiation/13636 "2024-02-05T15:44:34Z")
How severe does this issue affect your experience of using Ray? High: It blocks me to complete my task. I had a RNN model inheriting modelV2 that had worked well with ray 2.2. In ray 2.9, I set the option as required… --- ## [How to get the best performance of Ray´s RLlib when running python scripts (using PPO) via a SLURM file on a HPC?](https://discuss.ray.io/t/how-to-get-the-best-performance-of-ray-s-rllib-when-running-python-scripts-using-ppo-via-a-slurm-file-on-a-hpc/13116)
**Author:** [@MRMarlies](https://discuss.ray.io/u/MRMarlies)\ **Replies:** 0\ **Last updated:** [December 12, 2023, 9:21am UTC](https://discuss.ray.io/t/how-to-get-the-best-performance-of-ray-s-rllib-when-running-python-scripts-using-ppo-via-a-slurm-file-on-a-hpc/13116 "2023-12-12T09:21:14Z")
How severe does this issue affect your experience of using Ray? High: It blocks me to complete my task. Dear Ray-team! Currently I am working with RLlib and PPO. I developed an agent that imitates an acc-controller. … --- ## [Dreamerv3 weird bug](https://discuss.ray.io/t/dreamerv3-weird-bug/12765)
**Author:** [@evc](https://discuss.ray.io/u/evc)\ **Replies:** 0\ **Last updated:** [November 9, 2023, 1:00am UTC](https://discuss.ray.io/t/dreamerv3-weird-bug/12765 "2023-11-09T01:00:57Z")
Hi, I have been running the dreamerv3 with different environments but I got similar errors where all of them includes: cuDNN launch failure : input shape (\[1,1,512,1\]) \[\[{{node mlp/layer\_normalization\_8/FusedBatchNormV… --- ## [(raylet) ModuleNotFoundError: No module named 'ray' with installed ray](https://discuss.ray.io/t/raylet-modulenotfounderror-no-module-named-ray-with-installed-ray/12625)
**Author:** [@chvbs2000](https://discuss.ray.io/u/chvbs2000)\ **Replies:** 1\ **Last updated:** [October 30, 2023, 7:16pm UTC](https://discuss.ray.io/t/raylet-modulenotfounderror-no-module-named-ray-with-installed-ray/12625 "2023-10-30T19:16:57Z")
HI, I am using Python==3.10, ray==2.6.1. I am confused why I keep getting this error when I have ray installed: (raylet) File "/home/anaconda3/envs/myvirenv\_py310/lib/python3.10/site-packages/ray/\_private/workers/def… --- ## [All worker objects need to write to one single file (which we use for logging for splunk) on server the application started to run](https://discuss.ray.io/t/all-worker-objects-need-to-write-to-one-single-file-which-we-use-for-logging-for-splunk-on-server-the-application-started-to-run/12404)
**Author:** [@ramyamaddi](https://discuss.ray.io/u/ramyamaddi)\ **Replies:** 0\ **Last updated:** [October 10, 2023, 6:46pm UTC](https://discuss.ray.io/t/all-worker-objects-need-to-write-to-one-single-file-which-we-use-for-logging-for-splunk-on-server-the-application-started-to-run/12404 "2023-10-10T18:46:44Z")
We have an application that uses ray cluster with two servers to distribute work. It is supposed to process 20-100 files. This application spawns worker processes on these two servers/nodes. We want to write log statemen… --- ## [The example pendulum-mbmpo.yaml gives an error when I run it](https://discuss.ray.io/t/the-example-pendulum-mbmpo-yaml-gives-an-error-when-i-run-it/11940)
**Author:** [@Sergey\_Filimonov](https://discuss.ray.io/u/Sergey_Filimonov)\ **Replies:** 0\ **Last updated:** [August 25, 2023, 6:59am UTC](https://discuss.ray.io/t/the-example-pendulum-mbmpo-yaml-gives-an-error-when-i-run-it/11940 "2023-08-25T06:59:31Z")
At first, when I run tuned exampl example pendulum-mbmpo.yaml, it always returns the error message: (MBMPO pid=3356237) RuntimeError: split\_with\_sizes expects split\_sizes to sum exactly to 32 (input tensor’s size at dim… --- ## [Debug MNIST tuto](https://discuss.ray.io/t/debug-mnist-tuto/11809)
**Author:** [@Antoine101](https://discuss.ray.io/u/Antoine101)\ **Replies:** 0\ **Last updated:** [August 16, 2023, 3:12pm UTC](https://discuss.ray.io/t/debug-mnist-tuto/11809 "2023-08-16T15:12:44Z")
Hi everyone, I want to start using Ray Tune to implement hyperparameters optimization for my DL models. I tried to just copy and paste the MNIST tuto to start with. It works fine on my Macbook Air M1 (although a few t… --- ## [Debugging a custom model at runtime](https://discuss.ray.io/t/debugging-a-custom-model-at-runtime/11442)
**Author:** [@Robert1](https://discuss.ray.io/u/Robert1)\ **Replies:** 0\ **Last updated:** [July 17, 2023, 7:38am UTC](https://discuss.ray.io/t/debugging-a-custom-model-at-runtime/11442 "2023-07-17T07:38:22Z")
Hi everyone :wave: I’m working on a custom model in a custom environment with ray rllib. Naturally, I need to debug frequently :bug: Is there any good way to debug a custom model during training? The current behavior … --- ## [Why is GPU usage capped at 50% when training PPO?](https://discuss.ray.io/t/why-is-gpu-usage-capped-at-50-when-training-ppo/11392)
**Author:** [@george-adams](https://discuss.ray.io/u/george-adams)\ **Replies:** 0\ **Last updated:** [July 13, 2023, 4:28pm UTC](https://discuss.ray.io/t/why-is-gpu-usage-capped-at-50-when-training-ppo/11392 "2023-07-13T16:28:46Z")
Whenever I’m training a PPO, my CPU and GPU alternate. When the CPU workers are going through the environment, my CPU is at 100%. When the GPU is updating the parameters of the neural nets, it only reaches a maximum of 5… --- ## [Debugging proof of concept env with custom GCN model](https://discuss.ray.io/t/debugging-proof-of-concept-env-with-custom-gcn-model/11233)
**Author:** [@guillermo](https://discuss.ray.io/u/guillermo)\ **Replies:** 3\ **Last updated:** [July 3, 2023, 3:31pm UTC](https://discuss.ray.io/t/debugging-proof-of-concept-env-with-custom-gcn-model/11233 "2023-07-03T15:31:42Z")
I’m working on a problem with the observation space being a undirected graph and I created a proof of concept environment to debug any issues as I’m also using a custom model. However, I’m seeing that the average reward … --- ## [HalfCheetah isnot working withMBMPO](https://discuss.ray.io/t/halfcheetah-isnot-working-withmbmpo/10349)
**Author:** [@Rakesh\_Lal](https://discuss.ray.io/u/Rakesh_Lal)\ **Replies:** 1\ **Last updated:** [June 23, 2023, 7:32pm UTC](https://discuss.ray.io/t/halfcheetah-isnot-working-withmbmpo/10349 "2023-06-23T19:32:50Z")
To run HalfCheetah env with the algorithm MBMPO a wrapper is needed. The wrapper provided in ray.rllib.examples.env.mbmpo\_env.HalfCheetahWrapper is for HalfCheetah-v2 and gives errors. I made a wrapper to workaround this… --- ## [Logging episode media with WandB](https://discuss.ray.io/t/logging-episode-media-with-wandb/11079)
**Author:** [@Sertingolix](https://discuss.ray.io/u/Sertingolix)\ **Replies:** 0\ **Last updated:** [June 19, 2023, 6:10am UTC](https://discuss.ray.io/t/logging-episode-media-with-wandb/11079 "2023-06-19T06:10:43Z")
Hi there, I’m currently logging some string/text from my environment with a CustomCallback: def on\_episode\_end(self, \*, worker, base\_env:BaseEnv, policies, episode, env\_index, \*\*kwargs): urls = \[env.get\_url… --- ## [Training keeps getting stuck](https://discuss.ray.io/t/training-keeps-getting-stuck/10428)
**Author:** [@Brian\_H](https://discuss.ray.io/u/Brian_H)\ **Replies:** 3\ **Last updated:** [May 25, 2023, 5:41pm UTC](https://discuss.ray.io/t/training-keeps-getting-stuck/10428 "2023-05-25T17:41:31Z")
I am running on an M1 Mac Pro (32 GB ram, 10 CPU) if it matters. Ray RLLib code When I run experiments sometimes they get stuck. In a training loop, I won’t get a print statement to say the loop is finished. I have trie… --- ## [Ray Tune Table location](https://discuss.ray.io/t/ray-tune-table-location/8707)
**Author:** [@Gaddiel\_Ouaknin](https://discuss.ray.io/u/Gaddiel_Ouaknin)\ **Replies:** 1\ **Last updated:** [December 20, 2022, 7:07pm UTC](https://discuss.ray.io/t/ray-tune-table-location/8707 "2022-12-20T19:07:33Z")
I am using Ray Tune and I get a table in the command line with a performance report of every simulation. I wanted to know if there is Json of txt file with this table somewhere. Thank you.