Badges
Activity
Challenge Categories
Challenges Entered
Completed
Institute of Global Health
LifeCLEF
Classify images of snake species from around the world
Latest submissions
Stanford Neuromuscular Biomechanics Laboratory
Reinforcement Learning on Musculoskeletal Models
Latest submissions
MineRL Labs - Carnegie Mellon University
Sample-efficient reinforcement learning in Minecraft
Latest submissions
Unity Technologies
A new benchmark for Artificial Intelligence (AI) research in Reinforcement Learning
Latest submissions
See All| graded | 8603 | ||
| graded | 8558 | ||
| failed | 8555 |
Humanity & Inclusion
Building Missing Maps with Machine Learning
Latest submissions
| Participant | Rating |
|---|
| Participant | Rating |
|---|
SnakeCLEF2021 - Snake Species Identification Chall
Unity Obstacle Tower Challenge
Aicrowd-repo2docker error
Over 6 years agoHmm, not sure if I can deterministically replicate this, it just stopped working on my desktop after some time but starting up a new VM makes it work. I’ll definitely post if I get any further leads.
Aicrowd-repo2docker error
Over 6 years agoYeah, but not sure why it’s happening. But it’s working in another environment so I’ll just use that. Thanks!
Is the evaluation seed truly random?
Over 6 years agoAfter running evaluations for 20+ times, I’ve noticed I always get to 8~10th floor in the fifth round of the evaluation, but not the rest. So it is always like [5, 5, 6, 6, 10], [5, 5, 6, 7, 9], [5, 5, 6, 5, 10], etc.
Are the evaluation seeds fixed?
Aicrowd-repo2docker error
Over 6 years agoI don’t know why but suddenly I’m getting this error.
./build.sh: 3: ./build.sh: aicrowd-repo2docker: not found
It’s weird since when I just type aicrowd-repo2docker in terminal, it works, meaning $PATH isn’t the problem. Any leads?
Successful submissions do not appear on the leaderboard
Over 6 years ago@devops Now I don’t get this error but the submission is stuck for 21 hours after evaluation finished successfully. (Issue #173, Submission #7068) Could you look into this please? Appreciate your help!
Successful submissions do not appear on the leaderboard
Over 6 years ago@devops @mohanty Now I’m getting this error message.
2019-06-19T15:35:13.556346001Z root
2019-06-19T15:35:16.228972099Z INFO:mlagents_envs:Start training by pressing the Play button in the Unity Editor.
2019-06-19T15:35:46.233275211Z Traceback (most recent call last):
2019-06-19T15:35:46.233307689Z File "run.py", line 43, in <module>
2019-06-19T15:35:46.233312631Z env = ObstacleTowerEnv(args.environment_filename, docker_training=args.docker_training, retro=True, greyscale=True)
2019-06-19T15:35:46.233318069Z File "/srv/conda/envs/notebook/lib/python3.6/site-packages/obstacle_tower_env.py", line 45, in __init__
2019-06-19T15:35:46.233322404Z timeout_wait=timeout_wait)
2019-06-19T15:35:46.233326328Z File "/srv/conda/envs/notebook/lib/python3.6/site-packages/mlagents_envs/environment.py", line 69, in __init__
2019-06-19T15:35:46.233330448Z aca_params = self.send_academy_parameters(rl_init_parameters_in)
2019-06-19T15:35:46.233334307Z File "/srv/conda/envs/notebook/lib/python3.6/site-packages/mlagents_envs/environment.py", line 491, in send_academy_parameters
2019-06-19T15:35:46.233338144Z return self.communicator.initialize(inputs).rl_initialization_output
2019-06-19T15:35:46.233341852Z File "/srv/conda/envs/notebook/lib/python3.6/site-packages/mlagents_envs/rpc_communicator.py", line 80, in initialize
2019-06-19T15:35:46.233345601Z "The Unity environment took too long to respond. Make sure that :\n"
2019-06-19T15:35:46.233349459Z mlagents_envs.exception.UnityTimeOutException: The Unity environment took too long to respond. Make sure that :
2019-06-19T15:35:46.233353065Z The environment does not need user interaction to launch
2019-06-19T15:35:46.233356869Z The Academy and the External Brain(s) are attached to objects in the Scene
2019-06-19T15:35:46.233360461Z The environment and the Python interface have compatible versions.
I submitted the exact code that succeeded previously. Maybe the fix introduced a new issue? I’m suspecting the second line, which is something I’ve never seen before.
Successful submissions do not appear on the leaderboard
Over 6 years ago@mohanty My agent has been graded, but the issue wouldn’t close for 4 hours now after saying “generating video” Could you look into this? ( #6984)
Performance degradation after submission
Almost 7 years agoAlso I think more complexity (additional layout, themes, etc) has been introduced in Round 2 so you can’t expect the agent to perform as well as Round 1.
Vector Observation contents
Almost 7 years agoIf everything else is correct, possibly a windows issue- I’m using ubuntu 16.04 btw
Vector Observation contents
Almost 7 years agoWell I do see 8:
array([[1.00e+00, 0.00e+00, 0.00e+00, 0.00e+00, 0.00e+00, 0.00e+00, 2.99e+03, 0.00e+00]])
Maybe check the reset parameters?
Has the submission process changed for round 2?
Almost 7 years ago@arthurj in other words, can we assume the ‘config=None’ for the evaluation environment?
Am I allowed to join from Round 2?
Almost 7 years agoI believe the score of 5 is the minimum requirement to prevent a case where there are less than 50 serious participants, (e.g. if 40th place and below have score of 0) which was the case before the deadline extension.
Also, the “score” in that context is the floor level, not the reward. (This also can be found in Section 11) So you need to get higher than 5 in terms of “Mean number of Floors” but seems like you only scored 4.8… won’t hurt to confirm with the host though.
Am I allowed to join from Round 2?
Almost 7 years agoUp to 50 teams are allowed to participate in Round 2.
You can read the official rules here.
In Section 6,
“Step 4. Pursuant to Section 11, up to 50 Teams from Round 1 will win the option to move onto Round 2 (“Finalists”).”
Question about the “Round 1 of Unity’s Obstacle Tower Challenge” email
Almost 7 years agoTo my knowledge, what Truthmaker said (you pay no tax if you spend that $1100) is not correct for most people with standard deduction schedule. It is true that you may be able to deduct the $1100 spent, if it qualifies for ‘work related expense.’ But unless you have mortgages or other deductions big enough that goes over the standard deduction amount, you won’t be choosing itemized deduction, hence, spending $1100 won’t save you from paying taxes on it.
Again, international laws may differ but in principle that is what I think. At the same time, I’m not a tax professional so I may be wrong, do not depend on this advice for your tax filings.
In the meantime, I think best way would be to “withhold” the tax portion from GCP credits and give us what is remaining after the tax witholding. But that might not be feasible to the sponsor.
Libcuda.so.1 not available
Almost 7 years agoChanging that in your json file will disable gpu usage, which will drastically slow down your evaluation to the point where it won’t be a viable submission. Rather, you should fix your environment file (requirements.txt or environment.yml) so that the docker installs necessary libraries to run Cuda.
Using dopamine trained model
Almost 7 years agoYou need to tailor some of the dopamine files to manually perform inference, at least that’s how I went about it. Try reading through some of the codes in discrete_domains folder as well as agents. The library is very well documented so reading through the files will give you a very good idea.
What does it mean qualifying round?
Over 6 years agoI see that you’ll update this challenge every 3 months. What are the round 1 and 2 ‘qualifying rounds’ ? Is it like, only the people who participated in these qualifying rounds can participate in the final round later? Or are these rounds sort of like a practice?