Evaluation failed

Insurance pricing game
#1

微信截图_20210304205548
微信截图_20210304205548930×708 31.4 KB

Submission #124637

Why does it show rmse=0? Did you calculate rmse for us or you need to define a function to generate rmse,Can you help me see why the evaluation failed

#2

Hi @sweet_bitter,

You need to remove global_imports from your packages list.
It should work after it.

#3

Okay thank you, I now have a new problem, there is an inference failed, if possible, please help me see what is going on

1
11920×585 33.6 KB

#4

Hi @sweet_bitter

So if you click on “Evaluation Status” you will see that this submission at the “generate predictions on training sample” phase. There is a hyperlink with your error traceback right below the diagram that you can click. Labeled “Logs”.

image

If that’s unclear let us know and we’ll dig further :slight_smile:

1 Like
#5

2
2907×560 6.92 KB

I have been investigating for a long time. Is it an error to save the model, but I save it according to the specification, I really don’t know how to find the error, can you help me, thank you

#6

I got this error before. You should check the version of XGBoost you are using. The default in colab is 0.90. The default used by the server is 1.3.1, and trying to generate predictions from a different version of XGBoost can generate this error. You need to specify the version you are using in the configuration.Either pip install XGBoost with version 1.3.1 in colab, or specify version 0.90 in the configuration file so the server knows to use the older version. If you decide to use version 0.90 of xgboost you need to set APT_PACKAGES to include libgomp or you will get an error

Example of config for 0.90:

class Config:
APT_PACKAGES = [
“libgomp1”
]
ADDITIONAL_PACKAGES = [
‘numpy’, # you can define versions as well, numpy==0.19.2
‘pandas’,
‘xgboost==0.90’,
Etc…
]

#7

Ok thank you very much, i will try

#8

Sorry I have a new question, it is about the version, this question has troubled me for a long time, can you help me answer it, thank you

3
31920×905 23.6 KB

#9

Hello, I am very sorry to disturb you, I tried to set the version of xgboost to 1.3.1, but the original error still appeared, do you know why, thank you very much

#10

Hi @sweet_bitter

Based on your error message I see two things:

  1. Can you please make sure that you are using the local functions of xgboost to save and load your models? xgboost and pickle often don’t play nice :slight_smile:
  2. I can see you are trying to install pickle as a package within your required packages. This is not necessary and will raise an error because for some time now pickle has been part of the standard library and has not existed on the PyPI and so when we try to install it via pip install pickle you get that error. So I would just remove that and it should work :+1:

If these two don’t fix it, then please make a submission and cite your submission ID and we can dig further :construction:

1 Like