CODEWITHSUNDEEP
pythonkeraskaggle
SeanJ
SeanJ

Reputation: 1243

Kernel Threw Exception in kaggle submission

We have six submissions to a kaggle competition

The second submission succeeded but all other failed, four failed for Kernel Threw Exception

Looking at the log there is nothing to point to the failure:

Predicting

2019-09-07 07:41:52.231132: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10.0

2019-09-07 07:41:52.940301: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7

y_pred Shape:  (1928, 5)

[[0.009 0.296 0.398 0.111 0.186]
 [0.015 0.156 0.273 0.276 0.281]
 [0.009 0.293 0.042 0.356 0.3  ]
 ...
 [0.021 0.148 0.283 0.34  0.208]
 [0.007 0.197 0.516 0.136 0.144]
 [0.995 0.002 0.001 0.    0.001]]

(1928, 2)

The log stops here. Any ideas how we can discover what is causing the exception and why it would work once and not again?

Upvotes: 1

Views: 1011

Answers (1)

Daniel Möller
Daniel Möller

Reputation: 86600

Without looking at your kernel it's impossible to know.

But a few hints are:

  • Kernel must finish running before 9 hours
  • Make sure internet is "off" for those kernels
  • Make sure there is an output file called submission.csv
  • Make sure this file is in the correct format for the competition (compare to the output file of the kernel that worked)

Debugging hints:

Before submitting, make a "commit" first. Wait for it to finish and take a look at what is happening.

  • Check each cell for errors in their outputs
  • Check for the existence of the submission file
  • Check what is inside the submission file
  • Take a look at the log page of the kernel (not inside the editor when you commit, but at the kernel's page: https://www.kaggle.com/yourname/kernel_name without the edit part at the end)

This is the best to do if the kernel is private for us.

Upvotes: 2

Related Questions