reset_learning_rate callback: 'NoneType' object has no attribute 'set_param'

[hmishra2250]

Environment info

Operating System: Ubuntu 16.04 LTS

Compiler:

Package used (python/R/jvm/C++): python

xgboost version used: 0.6

Steps to reproduce

1. Define custom learning rate function or provide as list to xgboost.cv method. Attaching the error snippet as image.
   

2. The exact code snippet is

xgtest = xgb.DMatrix(x_test)

def get_learning_rate(i, n):
    if i<=1000:
        return 0.01
    return 0.001

lrl = [1]*5000
for i in range(5000):
        lrl[i] = 0.01 if i<=1000 else 0.00

res = xgb.cv(params, xgtrain, num_boost_round=5000, nfold=5, 
             early_stopping_rounds=50, maximize=False, 
             callbacks=[xgb.callback.reset_learning_rate(lrl)]) 

What have you tried?

1. Read the source code of callbacks.py and it turns out that cvpack object in env.cvfolds is getting to be None
2. Searched through all the issues in this repository and stackoverflow, and couldn't find a similar one, t the best of my reach.

I hope I am doing everything correct, hence I would appreciate feedback if its some bug or human error. Thanks in Advance and cheers 👍

[cinqs]

the newest version from master branch is like this:

So maybe your xgb version is out of date, and you either could wait for their update on pypi, or you can build one yourself.
#2925