Languages Update Issues

Platform Evolution
#22

I tested with the seed 507016145.

If i play against yakalir, the player 1 (second player) always timeout. Always. If i’m the second player, i timeout. If yakalir is second player, he timeout.

Then i tested against reCurse. If i’m player 0 i never timeout. If i’m player 1 i always timeout. reCurse never timeout, no matter the positions.

If i modify my code to prevent it to take all the second for the first turn, i never timeout.

I have no timeout in the arena. It’s IDE specific. And it’s specific to the second player position.

Conclusion: I don’t know what you are doing in the IDE for the second player, but it’s evil and the timeout calculation is wrong.

#23

Please, fix Haskell. No core libraries are found, while trying to compile old code. Note: everything just fine with properly updated Haskell platform on local machine. 10 days till contest…

1 Like
#24

Hello dabino,

Yes we will fix Haskell soon. But to be accurate, all core libraries are already installed, only additional libraries (“full” in Haskell terminology) are not installed. See https://www.haskell.org/platform/contents.html for definition.

So we will replace the core platform with the full platform. I will let you know when it is up.

2 Likes
#25

Hello @dabino

Haskell full platform is now installed. Please check that this is ok.

1 Like
#26

Everything compiles, thanks!

#27

The timeouts as second player should be fixed now. Please confirm.

#28

Hello @eulerscheZahl,

Tensorflow puzzle has been reinstated, though it seems that the new tensorflow consumes more memory, so you may have to adjust your variables (and some function calls are obsolete).

Also we have added:

import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'

in the default code to prevent unwanted warnings.

Please check if your code is still working.

#29

My code doesn’t work anymore (with and without the two lines above), but that might not even be bad: when I solved it, the tensorflow version on CG was outdated, so I had to search for old code in the github history of tensorflow to get it working. Tensorflow made a breaking change in the syntax of how to train it.

#30

I test multiple games in the IDE against @yakalir

Yakalir got a timeout once when he was the second player (once on 5 games).
And i got a timeout once when i was the second player (same, once on 5 games).

It’s pretty weird because for both timeouts it was the first game in the IDE. But i can’t say if it’s related.

#31

My code is no longer working, but after replacing function calls with the new syntax it does.

#32

After fixing my initialization step

init = tf.initialize_all_variables()

init = tf.global_variables_initializer()

I executed and faced this strange error related to SSE

2017-11-15 11:09:49.413614: I tensorflow/core/platform/cpu_feature_guard.cc:137] Your CPU supports instructions that this TensorFlow binary was not compiled to use: SSE4.1 SSE4.2 AVX
MemoryError

CPU execution target has changed?

#33

Hello,

I probably posted in the wrong topic and having a post here seems more appropriate, so please follow this for the detail. => it seems that g++ doesn’t work in C++14 mode ?