Submission error

Learning to Smell
#1

Hello,
I’m trying to make a submission for the learning-to-smell challenge:

I adapted the starter-kit to load and run my model but I keep getting an failure during the prediction setup step.

The error message is the following :
The following containers terminated prematurely. : agent Please contact administrators, or refer to the execution logs.

I think there is a bug in my setup method but since I don’t have access to the logs and I don’t understand the error message I can’t debug my code.

The submission link: http://gitlab.aicrowd.com/Wuala/learning-to-smell-starter-kit/issues/11

Thank you in advance for your help

#2

Hi @Wuala,

I have shared logs with you on the issue page.
It looks like some space vs tab issue.

Meanwhile, do look into debug submissions too. It will help you in accessing logs directly! :smiley:

#3

Hello,
Thank you for your reply.

I have still some questions:

In the starter-kit README there is the following instruction for the debug submission:
debug_submission: true
but in the file aicrowd.json there is the following line:
"debug": false
which one I have to set to true if I want to do a debug submission: debug_submission or debug ?
how many submission are allowed for each player? How many debug-submission are allowed?

(I apologize for the the many questions I’m new to the AIcrowd platform)

Thank you in advance for your help

#4

Hi @Wuala,

Thanks to point it out. It is indeed a typo, I have fixed it now.

You need use "debug": false.

#5

You are allowed to make 10 submissions and 10 debug submissions.

#6

Hello,

Here I’m again with a submission problem (sorry)
I’m making a submission with debug = true in the aicrowd.json

I get the following error during the Prediction Setup:

The following containers terminated prematurely. : agent
Please contact administrators, or refer to the execution logs.

I don’t understand why I can’t see the logs since my submission is with debug: true

Thank you in advance for your help

#7

Hi @Wuala,

You can click on agent-logs in your Gitlab issue to view your code’s logs.

#8

Thank you,
just one more thing:
I’m tying to use conda environment, I added the .yml file in the root directory but it doesn’t seem to work because I get errors related to the absence of modules.
Do you have any suggestions ?

#9

Hi @Wuala,

Your conda environment file name is incorrect, instead of my-xxxx-xxx.yml it should be environment.yml.

More description on different files we allow and their names here:

#10

thank you,
To generate the environment I have followed this post (How to use Conda environment for your submission)

I used the command conda env export --no-builds | grep -v "prefix" > environment.yml in the environment to export it with the right name.
Then I have deleted the two following lines (I’m on mac)

  • clangxx_osx-64=10.0.0
  • compiler-rt_osx-64=10.0.0
    when I make the submission I get the following failure Solving environment: ...working... failed

with this log error:

docker.errors.BuildError: The command '/bin/sh -c conda env update -p ${NB_PYTHON_PREFIX} -f "environment.yml" && conda clean --all -f -y && conda list -p ${NB_PYTHON_PREFIX}' returned a non-zero code: 1

1 Like
#11

Hi @Wuala,

Just above the error you copy pasted, it has the description which is:

ResolvePackageNotFound: 
  - gfortran_osx-64=4.8.5
  - appnope=0.1.2
  - libgfortran=3.0.1

Looks like it is happening because your export contains some packages which are not available on Linux (or the version pinned isn’t available in Linux) v/s osx.

  • In case you don’t need those packages directly (assuming they got added as a dependency of the package you really wanted to use), please remove them from your environment.yml file and make the submission.
  • You want to use the package, in that case, please check for its Linux version and/or remove version pinning.

Please let me know in case you continue to face any issue.

#12

thank you,
I removed all the unnecessary packages that where causing problems but i still have the following error:

Pip subprocess error: ERROR: Could not find a version that satisfies the requirement aicrowd-api==0.1.24 ERROR: No matching distribution found for aicrowd-api==0.1.24

#13

Hi @Wuala,

The version you have pinned to i.e. 0.1.24 which doesn’t exist, causing your submission to fail.

You can unpin the version and submit your code.

#14

Hello, I have another environment related bug:
When I make a submission (https://gitlab.aicrowd.com/Wuala/learning-to-smell-starter-kit/-/issues/75) with a custom environment I get the following error
No CUDA runtime is found
which lead to this error: ERROR: Failed building wheel

Any idea on how to fix this?
Thank you in advance for your help